Economic Regulation of Electricity Distribution Utilities Under High Penetration of Distributed Energy Resources: Applying an Incentive Compatible Menu of Contracts, Reference Network Model and Uncertainty Mechanisms by Jesse D. Jenkins B.S. Computer and Information Science and Philosophy University of Oregon, 2006 SUBMITTED TO THE ENGINEERING SYSTEMS DIVISION IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF ITUE MASTER OF SCIENCE IN TECHNOLOGY AND POLICY AT THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY MAY 2 1 2014 JUNE 2014 LIBRARIES LRAELOG @ 2014 Massachusetts Institute of Technology. All rights reserved. Signature of the Author: Signature redacted 6/ Certified by: Engineering Systems Division May 15, 2014 Signature redacted ___ Ignacio Perez-Arriaga Visiting Professor, Engineering Systems Division Thesis Supervisor Accepted by: Signature red acted Dava Newman Professor of Aeronautids and Astronautics and Engineering Systems Director, Technology and Policy Program Economic Regulation of Electricity Distribution Utilities Under High Penetration of Distributed Energy Resources: Applying an Incentive Compatible Menu of Contracts, Reference Network Model and Uncertainty Mechanisms by Jesse D. Jenkins Submitted to the Engineering Systems Division on May 15, 2014 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Technology and Policy ABSTRACT Ongoing changes in the use and management of electricity distribution systems including the proliferation of distributed energy resources, smart grid technologies (i.e., advanced power electronics and information and communication technologies) and active system management techniques - present new challenges for the economic regulation of electricity distribution utilities. In particular, regulators are likely to face increased uncertainty regarding the evolution of network uses and the efficient cost of network investments and maintenance, as well as an increased informational disadvantage vis-a-vis the regulated utility. These challenges are especially important for regulatory approaches that establish some share of the utility's allowed revenues ex ante (e.g., incentive regulation, also known as revenue or price cap regulation, RPI-X, performance-based regulation, or output-based regulation). This thesis proposes a novel process for establishing the allowed revenues of an electricity distribution utility and demonstrates its application as a practical solution to the imminent regulatory challenges discussed above. The proposed method is a new combination of three established regulatory tools: an engineering-based reference network model (RNM) for forward-looking benchmarking of efficient network expenditures; an incentive compatible menu of contracts to elicit accurate forecasts from the utility and create incentives for cost saving efficiency efforts; and ex post automatic adjustment mechanisms, or "delta factors," to accommodate uncertainty in the evolution of network use and minimize forecast error. 2 Chapter 1 reviews the theoretical economic foundations of the regulation of network monopolies, identifies emerging challenges in the regulation of electricity distribution companies, and introduces the proposed regulatory method. Chapter 2 describes the simulation of a realistic, large-scale urban distribution network used to demonstrate the novel regulatory process proposed in this thesis. Chapter 3 uses the simulated distribution network to demonstrate, step-by-step, the practical implementation of the novel regulatory process, evaluates its performance, and summarizes the advantages for the economic regulation of electricity distribution utilities under increasing penetration of distributed energy resources. Thesis Supervisor: Ignacio Perez-Arriaga Title: Visiting Professor, Engineering Systems Division 3 Acknowledgements This thesis builds on a rich intellectual foundation, without which this work would not have been possible. I would like to acknowledge, in particular, the contributions of Rafael Cossent and Tomas G6mez in developing a practical method for creating an incentive compatible menu of contracts, Carlos Mateo, Tomas G6mez, Alvaro Sanchez-Miralles, Jesus Peco, and Antonio Candela Martinez for their development of the reference network model employed in this thesis, and to the regulatory staff at the UK Office of Gas and Electricity Markets for their pioneering application of so many regulatory best practices (and the numerous, transparent publications detailing their thinking and lessons learned). I would also like to thank my fellow research assistants on the Utility of the Future Project, Ashwini Bharatkumar and Scott Burger, for helping me think through some of the many road blocks encountered on the route to completing this thesis (or at least listening to me explain them!). Above all, I am grateful to two individuals: Ignacio Perez-Arriaga, whose course on the regulation of the electric power sector is the best introduction to the field any student could ask for and whose guidance and advice during this thesis were invaluable; and Claudio Vergara, who developed the network simulation approach employed herein and who generously offered hours of critical assistance as I extended his methods for use in this thesis. Finally, I would like to thank my wife, who supported me through so many long hours. 4 Contents Chapter 1 - Economic Regulation of Electricity Distribution Utilities: Introduction, Emerging Challenges, and New Solutions 6 1.1- An Introduction to Network Regulation 6 1.2 - Emerging Challenges in the Regulation of Electricity Distribution Utilities 15 1.3 - New Solutions for Regulation of Distribution Utilities Under High Penetrations of Distributed Energy Resources 20 1.4 - Structure of the Thesis 29 References 30 Chapter 2 - Constructing a Simulated Electricity Distribution Network 35 2.1 - Specification of Simulation Parameters 35 2.2 - Assigning Load Profiles 37 2.3 - Determining Network Topology and Assigning Geographic Location of Loads 39 2.4 - Constructing the Network with the Reference Network Model 43 2.5 - Simulating Network Expansion Scenarios 47 References 52 Chapter 3 - Demonstrating the Proposed Regulatory Process 54 3.1 - Forecasting the Evolution of Network Uses 54 3.2 - Establishment of Regulator's Ex Ante Estimate of Efficient Expenditures 58 3.3 - Construction of an Incentive Compatible Menu of Contracts 65 3.4 - Calculation of Ex Ante TOTEX and Revenue Baselines and Sharing Factor 70 3.5 - Calculation of Automatic Adjustment Factors to Manage Uncertainty 74 3.6 - The Ex Ante Regulatory Process: Applying Annual Corrections 78 3.7 - Evaluating the Performance of the Regulatory Process 83 3.8 - Summary of Advantages of the Proposed Regulatory Process 88 References 90 5 Chapter 1: Economic -Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) 1 - Economic Regulation of Electricity Distribution Utilities: Introduction, Emerging Challenges, and New Solutions 1.1 - An Introduction to Network Regulation Electricity distribution is a natural monopoly activity. Due to the existence of economies of scale (e.g., average costs decline as output increases), marginal costs for distribution networks are lower than average costs. It is therefore less costly for a single firm to supply network services in a given territory rather than multiple firms (that is, firm costs are subadditive). In addition, the capital-intensive nature of distribution networks and the existence of considerable sunk costs in existing networks create significant barriers to entry for new competitors. This combination of economies of scale and significant barriers to entry make electric distribution activities a poor candidate for market competition, thus requiring regulatory intervention to prevent monopoly abuse of market power (Cossent, 2013; G6mez, 2013a, 2013b; Joskow, 2005).1 Distribution utilities have therefore been subject to various forms of economic regulation since very early in the history of the electric power sector, either via public ownership or oversight of private distribution companies by a regulatory body capable of, at minimum, establishing allowed revenues and/or prices. Furthermore, distribution networks (along with transmission network activities) remain a regulated sector even in electricity markets that have undergone privatization and/or deregulation of generation or retail activities. Indeed, these regulated network activities provide the infrastructure platform for both the generation and retail market activities, whether liberalized or not. Effective regulation of electricity network utilities is thus a critical cornerstone of any well-functioning, competitive market segments and has important welfare consequences for electricity consumers and society (Joskow, 2013). Regulatory frameworks for network utilities in the electric power sector and elsewhere have evolved over time based on both practical experience and advances in economic and regulatory theory. In general, traditional cost-of-service regulation is gradually being supplanted by incentive or performance based regulation, which is designed to create stronger incentives for network utilities to reduce costs through efficiency gains. Incentive regulation of electricity transmission and distribution utilities has become 1 Of course, the costs of imperfect markets must be balanced against the costs of imperfect regulation (Joskow, 2010). It may also be possible to discipline natural monopolies by introducing competition for the market (i.e., via regular franchise auctions) as opposed to competition in the market. However, several considerations make electricity distribution a poor candidate for the introduction of franchise competition, including: the long-lived assets and significant sunk costs associated with distribution networks; considerably advantages afforded to incumbent operators versus other bidders due to information asymmetries; and the currently evolving nature of distribution network technologies and demands for system services (see Cossent (2013) for in-depth discussion of these issues). 6 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) widespread in Europe (Cossent, G6mez, & Frfas, 2009; Jamasb & Pollitt, 2007) and Latin America (Rudnick, et al., 2007) following privatization, restructuring, and liberalization of these markets beginning in the 1990s. In addition, while the United States is still broadly characterized by a mix of municipal or other publicly-owned distribution utilities and cost-of-service regulated private utilities, elements of performance or incentivebased regulation have begun to be adopted with increasing frequency (Joskow & Schmalensee, 1986; Joskow, 2013). Regardless of which method is employed, economic regulation of distribution utilities must confront several general regulatory challenges, including incomplete and asymmetric information, a firm participation constraint, and resulting opportunities for strategic behavior on the part of regulated firms. In addition, regulators must carefully balance fundamental tradeoffs between increasing allocative efficiency and avoiding moral hazard on the one hand and increasing X-efficiency and minimizing adverse selection on the other hand (as discussed further below). The economic regulation of electricity distribution utilities involves two key tasks: first, the regulator must determine the sum of revenues the regulated utility is allowed to collect to remunerate their operating and investment costs (the cost recovery or remuneration challenge); and second, the regulator must determine how the utility should collect these revenues from their network users (the cost allocation or tariff design challenge). The cost allocation challenge lies outside the scope of this thesis and so will not be discusses further herein. Moreover, while fair and efficient tariff design is an essential regulatory undertaking, regulators must typically first determine the firm's total allowed revenues, balancing allocative efficiency, X-efficiency, and the firm participation constraint (Cossent, 2013). This determination of allowed revenues is at least as important as tariff design, and the regulated remuneration of distribution utilities establishes the principal financial incentives for the regulated firm. The practical challenges associated with this key task have spawned considerable advancement in both regulatory theory and practice. The theoretical economic foundations of the cost recovery or remuneration challenge lie in contract theory, particularly the design of regulatory contracts under imperfect and asymmetric information (see Bolton & Dewatripont, 2005; Joskow, 2005, 2008, 2013; Laffont & Tirole, 1993). The first challenge facing the regulator is that of imperfect information: the regulator does not know the utility's cost or service quality opportunities ex ante, nor can the regulator directly observe the utility's managerial effort to capture efficiency 7 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) opportunities (Joskow, 2013). As a result, the utility knows much more about their cost opportunities than the regulator, introducing considerable information asymmetries. In addition, the regulator must assess the prudency and efficiency of capital-intensive utility investments with relatively long asset lives (often measured in decades). This introduces additional challenges associated with uncertainty about future technological change and demand for network services. Together, these challenges create a significant opportunity for strategic behavior on the part of the regulated utility, wherein the firm uses its information advantage in the regulatory process to increase its allowed revenues and profits or achieve other managerial objectives (Averch & Johnson, 1962; Jamasb, Nillesen, & Pollitt, 2003, 2004; Joskow, 2013; Laffont & Tirole, 1993). In particular, the firm would like to convince the regulator that it is a higher cost firm than it really is, taking advantage of the regulator's need to comply with the firm participation constraint (e.g. ensure the financial viability of the regulated firm). To illustrate this challenge, consider a simple case where the utility is either of a "high cost" type or a "low cost" type. 2As illustrated in Table 1-1, if the regulator believes the firm is of a low cost type but it is in actuality a high cost type, the firm will not recover its costs and the regulator will violate the firm participation constraint. In contrast, if the regulator believes the firm is a high cost type, the firm participation constraint will be met regardless of the firm's true cost type, although excess profits (or rents) will result in the case that the firm is in actuality a low cost utility. The social welfare maximizing regulator therefore faces an adverse selection problem as it seeks to adhere to the firm participation constraint in the face of incomplete information about the firm's true cost type and strategic behavior on the part of the regulated utility (Joskow, 2013). Table 1-1: Adverse Selection and the Firm Participation Constraint Regulator's Belief Regarding Utility's Actual Cost Type the Utility's Cost Type Low High Low Break-even Fail to break-even High Excess profits Break-even This leads to a final regulatory challenge: the creation of moral hazard. The regulated firm's actual cost type is dynamic, not fixed, and it depends in part on the exertion of managerial effort to identify and exploit opportunities for cost reduction and/or performance improvement. Yet the regulator cannot directly observe this managerial 2 The discussion that follows is drawn in large part from Cossent (2013), G6mez (2013b), and Joskow (2013). 8 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) effort either (another case of incomplete information). If the regulator establishes allowed revenues that are too high or are set so as to ensure firm participation regardless of the degree of actual managerial effort, the firm's management will have no incentive to pursue cost savings. As Joskow (2013) colorfully puts it, "If the rat doesn't smell the cheese and sometimes get a bit to eat, he may play golf rather than working hard to achieve efficiencies for the regulated firm." This discussion illustrates a broader tradeoff the regulator faces between incentivizing managerial effort by creating financial incentives for the firm to pursue cost savings on the one hand and minimizing economic rents collected by the utility from ratepayers on the other hand (Joskow, 2013). In economic terms, this is a tradeoff between productive efficiency or X-efficiency-that is, ensuring firms are optimizing their production functions to minimizing the inputs used to produce a given level of output-and allocative efficiency-that is, ensuring prices reflect costs and social welfare is maximized by eliminating deadweight losses due to monopoly abuse of market power. Figure 1-1: Fundamental Tradeoffs in Network Utility Regulation Goals increase allocative efficiency vs. Minimize Rent Extraction Increase X-efficiency vs. Incentivize cost savings vs. Adverse selection Challenges Moral hazard Economic theory and regulatory practice have evolved over time to respond to and manage these fundamental regulatory challenges. Two general approaches to the regulation of network utilities have emerged, which, in their "pure" form, reflect alternative approaches to the tradeoffs described above: cost-of-service regulation and incentive regulation. While, in practice, regulatory mechanisms generally depart from these pure forms, discussion of each archetype can help illustrate the core features of these alternative regulatory approaches. We also note that in either case, regulators establish allowed revenues with reference (implicitly or explicitly) to an established level of service quality (typically enacted through minimum standards and/or incentives regarding reliability and losses). In its pure form, cost-of-service regulation (or rate-of-return regulation) is essentially a "cost-plus" contract negotiated between the regulator (on behalf of ratepayers and society) and the utility (G6mez, 2013b). The regulator sets allowed revenues for the utility based on ex post evaluation of recently incurred utility costs (i.e., looking back at 9 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) the preceding year), known as a "historic test year," or based on relatively short-term ex ante estimates of costs (i.e., looking ahead at the following year), known as a "future test year." In either case, the regulator aligns revenues with realized costs plus a regulated rate of return through frequent ex post reviews and adjustments. This approach is designed in such a way as to ensure the regulated firm recovers all of the costs incurred 3 while earning a fair rate of return on their investments. Since allowed revenues are projected ex ante for only a short time period and frequently reviewed ex post, this approach mitigates uncertainties about the evolution of distribution costs and network demands. In addition, the assurance of a fair return on investments ensures the firm participation constraint is met and necessary investment is attracted to the sector. In addition, if the rate of return is set efficiently-that is, high enough to attract sufficient investment into the sector and not too high so as to charge ratepayers more than necessary-then the regulator can maximize allocative efficiency and avoid transferring excess economic rents from ratepayers to the utility. The tradeoff inherent to this pure cost-of-service approach is that regulated firms have little-to-no incentive to pursue cost saving efficiency efforts. That is, the moral hazard problem is entirely unmitigated and significant X-inefficiencies remain. In fact, if the regulator establishes an allowed rate of return that exceeds a fair cost of capital-either due to asymmetric information and adverse selection or deliberate strategic behavior on the part of the utility-then the firm will be incentivized to substitute capital for other inputs (e.g., labor) and over-invest in capital, further exacerbating X-inefficiencies. This over-investment is known as the Averch-Johnson effect (Averch & Johnson, 1962). The pure form of incentive regulation (also known as price cap, revenue cap, or performance regulation) takes the opposite approach (Beesley & Littlechild, 1989; Laffont & Tirole, 1993). The regulator caps allowed revenues or prices ex ante for a set period (e.g., 3-8 years). Firm profitability and returns on investment thus depend on the utility "beating the cap"-that is, reducing realized costs below the price or revenue cap. This approach eliminates the moral hazard problem and creates a high-powered incentive for the exertion of managerial effort to optimize X-efficiencies. Again, there are clear tradeoffs to a pure incentive regulation approach. If the regulator sets the price or revenue cap too low, the firm may be unable to attract necessary investments or remain profitable. Given uncertainties about the evolution of technology and demands for network services after the ex ante cap is established, this approach may therefore risk violating the firm participation constraint. Alternatively, to avoid violating this constraint, the regulator may have a strong incentive to set the cap too In practice, regulators typically limit cost recovery or returns on investment only to "prudently incurred" expenditures, as discussed further below. 3 10 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) high, incurring the full adverse selection problem and undermining allocative efficiency by leaving potentially large rents to the utility (a transfer of welfare from ratepayers). In summary, a pure cost-of-service regulatory approach trades off allocative efficiency, minimization of rents, and assurance of firm participation for unmitigated moral hazard and negligible incentives for X-efficiency. In contrast, a pure incentive regulation approach establishes strong incentives for X-efficiency and eliminates moral hazard yet risks either violating the firm participation constraint or succumbing to adverse selection and leaving considerable rents to the regulated firm (reducing allocative efficiency). Table 1-2: Summary of "Pure" Approaches to Network Regulation Cost-of-Service Regulation (Rate-of-Return) Incentive Regulation (Price/Revenue Cap) * Essentially "cost-plus" contract with firm earning regulated rate of return 0 Caps prices or revenue ex ante for a set period * Prices or revenues set based on realized costs reviewed ex post 0 Firm profitability depends on "beating the cap" * Pros: maximizes allocative efficiency and minimizes economic rents extracted by the firm; ensures firm participation 0 Pros: eliminates moral hazard and creates strong incentives to maximize X-efficiency * Cons: Negligible incentives for X-efficiency, creating moral hazard problem 0 Cons: To avoid violating firm constraint, may incur full cost of adverse selection and leave rents to regulated firm Faced with the tradeoffs and challenges inherent in the regulation of network utilities under information asymmetries, the social welfare maximizing regulator will, in practice, seek mechanisms that balance both the costs of adverse selection and moral hazard and the benefits of allocative and X-efficiency, all while abiding by the firm participation constraint. In addition, the regulator will seek mechanisms to reduce their informational disadvantage as much as possible. Indeed, while the pure forms of cost-of-service and incentive regulation lie at extreme ends of the tradeoff space illustrated in Figure 1-1, the actual implementation of either approach tends to lie somewhere in between. Under cost-of-service-based approaches, regulators can instill some incentive for managerial efficiency by subjecting investments to an ex-post review. That is, the regulator will allow the utility a return only on those investments deemed to be "prudently incurred" as part of periodic regulatory reviews or "rate cases." In addition, as the regulator reviews the utility's investments only periodically rather than continuously, there exists a delay between the time when the utility makes an 11 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) investment or incurs a cost and the time when allowed revenues are adjusted accordingly. This dynamic creates an additional incentive for the firm to reduce costs, as prices or revenues are in practice fixed for some period of time (which may extend to several years). In addition, cost-of-service approaches have frequently been blended with various, targeted performance-based incentives for efficiency or quality (Joskow & Schmalensee, 1986). Similarly, real-world implementations of incentive regulation feature regular "ratchets" of the cap-that is, after a fixed interval of time (i.e., 3-8 years), the regulator will review the utility's actual costs and performance and reset the revenue or price cap based on that review. These price ratchets effectively transfer the economic savings due to exertion of managerial effort from the firm back to ratepayers. The ratchet also reduces incentives for cost saving, as a dollar of cost reduction only entitles the firm to additional revenues for early on in the regulatory period is worth more to the utility than a dollar of savings implemented just prior to the next ratchet of the cap. 4 In reality then, incentive regulation does not provide the maximum incentive for managerial effort, nor does it wholly ignore allocative efficiency in its pursuit of optimal X-efficiency. Indeed, depending on the frequency of ratchets under incentive regulation and the length between rate cases under cost-of-service, real-world implementations of these alternative regulatory approaches may fall quite close together on the spectrum illustrated in Figure 1-1. Given the inherent tradeoffs between regulatory approaches, it is perhaps not surprising that theoretical developments indicate that the preferable regulatory mechanism is a balance between a pure cost-of-service approach and a pure price or revenue cap incentive approach. This approach takes the form of a sliding scale regulatory mechanism in which allowed revenues are partially fixed ex ante so as to create incentives for cost reduction and partially responsive ex post to changes in realized costs to improve rent extraction and mitigate uncertainty (G6mez, 2013b; Joskow, 2013; Schmalensee, 1989). This regulatory regime effectively shares profits and rents as well as risks between the utility and ratepayers based on a profit-sharing factor (or efficiency incentive rate) between one (corresponding to a pure price/revenue-cap incentive regulation) and zero (corresponding to a pure cost-of-service approach). The regulator can choose the precise sharing factor to manage tradeoffs between incentives for efficiency and rent extraction and in response to the degree of uncertainty about future costs and demand. In particular, under lower levels of uncertainty, a higher profit-sharing factor (i.e., the firm is exposed to most of the risks and rewards of cost 4 Cost-saving incentives can be equalized throughout the regulatory period by implementing a rolling efficiency incentive payment that allows a portion of efficiency savings to be earned in the following regulatory period (as demonstrated in Chapter 3). 12 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) savings) performs better, while a lower profit-sharing factor (which shares most risks and rewards with ratepayers) performs better under higher levels of uncertainty (Schmalensee, 1989). Furthermore, the regulator can improve on a single profit-sharing factor by offering a regulated utility a menu of regulatory contracts with a continuum of different sharing factors (Cossent & G6mez, 2013; Laffont & Tirole, 1993). This menu of contracts allows the firm to play a role in selecting the strength of incentives for cost saving. If constructed correctly, this menu will establish "incentive compatibility"-that is, the design of the menu ensures that a profit-maximizing firm will always be better off (i.e., earn the greatest profit and return on equity) when actual expenditures match their ex ante estimate of necessary expenditures. Incentive compatibility thus eliminates incentives for firms to artificially inflate their cost estimates while incentivizing firms to reveal their true cost types to the regulator, helping minimize strategic behavior and overcome information asymmetries. Furthermore, a profit-motivated firm with less opportunity to reduce costs will choose a low-powered incentive, while a firm with large efficiency opportunities will choose a high-powered incentive. Despite strong theoretical advantages, the use of a menu of regulatory contracts in the regulation of electricity distribution utilities has been very limited in practice (Cossent & G6mez, 2013; Cossent, 2013; Joskow, 2013; Ofgem, 2009, 2010b, 2013c). Chapter 3 will discuss the construction and implementation of an incentive compatible menu of regulatory contracts further. In addition to tailoring cost-of-service and incentive regulation approaches to manage these inherent tradeoffs, regulators have also developed various methods to overcome information asymmetries. Regulatory accounting systems are integral to both approaches. These systems involve systematically collecting data from regulated firms regarding their costs, assets, performance indicators, and other information needed in the regulatory process (Cossent, 2013). This includes specifying the regulatory asset value of the firm's capital assets (also known as the "rate base"), tracking depreciation of assets, calculating the weighted average cost of debt and equity capital sources (WACC), allowances for taxes paid, and regular auditing of utility investments and accounts. Regulatory accounting systems must be carefully designed to collect any information necessary for adequate regulatory decisions, while avoiding excess burdens on regulated firms and regulatory staff. Accounting should also be auditable and traceable to minimize opportunities for strategic behavior. In addition, regulators frequently employ outside consultants in this process to help review and assess the regulated firm's performance. 13 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Another way in which regulators can reduce their information disadvantages vis-a-vis the utility (and help reduce the opportunity for strategic behavior) is by comparing the utility's actual costs and performance relative to a reference or benchmark of performance (Cossent, 2013; G6mez, 2013b; Jamasb & Pollitt, 2001, 2003). This reference or benchmark can be the actual performance of one or more similar firms (known as "yardstick competition"), an estimate of efficient performance derived through statistical analysis of the measured performance of a large number of similar firms ("frontier benchmarking"), or an estimate of efficient performance derived from engineering models ("reference network" or "engineering norm" approaches). Consider the simple case where two identical firms serve two identical service territories, but do not directly compete with one another. By establishing allowed revenues for both firms based on some weighted average of both firms' costs, the regulator can force these two firms to effectively compete with one another. If one firm cuts costs faster than the other, it will profit under this scheme, while a firm that lags behind its twin will incur a loss. Shleifer (1985) demonstrates the efficiency of this method of yardstick competition and notes that even in the more realistic case of heterogeneous firms, yardstick competition can compare favorably to setting allowed revenues based on the firm's cost-of-service alone, provided that the heterogeneity can be observed and accounted for. In addition, the regulator does not need to know the precise methods employed by each firm to reduce costs and improve X-efficiency. Rather, regulatory accounting data on each firm's performance is sufficient, thus reducing information asymmetries and regulatory burden. Alternatively, the regulator does not need to limit its comparisons to the performance of a few similar firms. Instead, through statistical and econometric frontier benchmarking techniques, the regulator can analyze the relative efficiency and performance of a large number of utilities in the sector, both within the same country or even internationally (Jamasb & Pollitt, 2001, 2003). Using these benchmarking techniques, 5 regulators can estimate "the efficient frontier"-that is, the optimal ratio of inputs to performance, controlling for various relevant characteristics or cost drivers (i.e., load factors, density of loads, geographic constraints, etc.)-and then benchmark the firm's actual performance relative to this frontier. Finally, regulators can employ engineering reference or norm models to construct an ideal or efficient representative firm, which can be compared to performance of the actual regulated firm (Cossent, 2013; Domingo, et al., 2011; Jamasb & Pollitt, 2008). Using a combination of engineering models and optimization methods, these methods s For reviews of benchmarking techniques for utility regulation, see Cossent (2013), Gomez (2013b), and Jamasb & Pollitt (2003 & 2001). 14 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) can construct a realistic, efficient reference network, including estimated investment and maintenance costs as well as energy losses and quality of service levels, taking into account the particularities of the real regulated firm's service territory (i.e., the location and profiles of network users, cost and performance of available technology components, and geographic constraints on network layout). The key parameters of this reference network can then be employed as indicators of efficient performance for the real, regulated firm. 1.2 - Emerging Challenges in the Regulation of Electricity Distribution Utilities The regulation of the electric power sector has historically co-evolved along with the underlying technical and economic characteristics of the electricity system.6 Economies of scale and scope driven by the particular engineering characteristics of conventional thermal and hydro power plants and electricity transmission and distribution networks gave rise to natural monopoly characteristics and prompted the emergence of the regulatory methods discussed in Section 1.1. Over time, a number of technological changes occurred, including the advent of less capital intensive and smaller scale natural gas-fueled power plants, much stronger and interconnected transmission networks, computerized handling of consumer information, industrial co-generation opportunities, and renewable energy technologies. This new portfolio of technologies reduced economies of scale and lowered barriers to entry, making competitive wholesale generation and retail supply markets possible. This era of technological innovation thus sparked a wave of industry restructuring, liberalization, and regulatory reform that spread across numerous jurisdictions during the 1980s and 1990s 7 (Bushnell & Borenstein, 2000; Perez-Arriaga, 2013). Today, a new era of technological innovation is once again reshaping the electric power sector in many jurisdictions, with much of that change centered around the electricity distribution sector (Bharatkumar et al., n.d.). Previous eras of regulatory reform have focused on market designs (particularly for wholesale generation and retail supply) and regulatory incentives for the network-related activities (transmission and distribution) to improve economic efficiency, affordability, and quality of supply. A new set of reforms must further these ongoing objectives while also addressing the evolving uses of distribution networks and fostering technological innovation and the efficient evolution of electric power systems and markets (Bauknecht, 2011; Bharatkumar et al., 6 This section is based in part on Chapter 3, "Regulatory Issues," in Bharatkumar et al. (2014). 7 Earlier reforms include the restructuring of the Chilean electricity market in 1981, introduction of independent power producers in the United States via the Public Utilities Regulatory Policy Act (PURPA) of 1978. 15 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) 2014; Cossent, G6mez, & Frias, 2009; Cossent, 2013a; Lo Schiavo et al., 2013; Newcomb, Lacy, & Hansen, 2013; Perez-Arriaga et al., 2013). In particular, a broad range of distributed energy resources (DERs), including distributed generation and storage, demand response, and electric vehicles, are proliferating in electricity distribution networks. Information and communication technologies (ICT) are also becoming embedded across the distribution system, enabling new "smart grid" capabilities. Together, these new technologies and capabilities are leading to the emergence of innovative business models utilizing distributed energy systems, or DESs-novel systems combining one or more DERs with ICT capabilities to deliver value to electricity end-users, upstream market actors, and/or system operators (Bharatkumar et al., n.d.). These emerging technologies and associated business models have the potential to drive several paradigmatic shifts in the planning and operation of electricity distribution systems and related markets: " The traditional top-down power flow from centralized generation sources connected to the high voltage transmission grid through lower-voltage distribution to end consumers is challenged by local distributed generation (DG), requiring changes to distribution infrastructure and operations. Multi-directional power flows across distribution networks may soon be the norm, and new opportunities are emerging for local means of electricity trade, including microgrids and on-site DG, that directly compete with wholesale generators. * Distributed storage (DS) could entail profound changes to the real-time operation of electric power systems, offering a buffer between system supply and demand, new ways to provide ancillary services to network operators, and opportunities to temporally shift energy supply to maximize the value of energy production and meet peak demands. At the same time, cost effective storage combined with on-site generation may also reduce electricity end user's reliance on distribution networks. * Demand response (DR) makes electricity loads far more responsive to economic and operational signals than ever before. Conversely, the proliferation of customer-owned DG, particularly variable distributed solar and wind technologies, may make generation less controllable and predictable for system operators. * Widespread adoption of electric vehicles (EVs) would constitute an important new class of electricity system users and loads. EVs also hold the possibility of injecting power back into the grid, delivering so-called "vehicle-to-grid" (V2G) services. Efficient price signals and/or new control systems will be essential to 16 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) manage and coordinate EV charging and V2G services, while new network investments must accommodate and enable EV users. The role of the distribution utility in enabling this new market segment must also be defined. Last, but perhaps most significantly, a cost-effective combination of DG, DS, DR and EVs-whether in the format of an autonomous or semi-autonomous physical microgrid or as a virtual decentralized aggregation of each of these components-may challenge the current centralized paradigm of the electric utility, initiating a transition towards a more decentralized structure and organization of the power sector. Many of these trends are still nascent. Other important changes, foremost the increasing penetration of solar photovoltaics and other DG, are already established facts and concern distribution utilities and regulators in many jurisdictions today. In addition, while many of these trends have so far been driven primarily by supportive public policy measures, utilities and analysts are increasingly focused on the potential for selfsustaining and disruptive forces to take hold. As technology innovation and industry maturation and scale-up steadily improve the price and performance of these technologies, tipping points are possible, after which adoption may proceed quite rapidly (Bharatkumar et al., 2014; Eurelectric, 2013b; Kind, 2013; Newcomb et al., 2013). The ongoing evolution of electricity distribution networks and emergence of new network users therefore require a new era of proactive regulatory innovation and reform. These reforms will likely include: " Updating remuneration methods to ensure distribution utilities are compensated for investments required to accommodate new DERs and DESs and incentivized to make efficient use of these same resources to improve system reliability, reduce losses, and defray unnecessary capital expenditures. * Redesigning distribution access or use of network charges to provide efficient price signals for the location and operation of new DERs while fairly allocating costs to network users and ensuring adequate cost recovery for the distribution utility. * Revisiting the proper industry structure and clarifying the responsibilities of the distribution utility, including vis-&-vis ownership or operation of DER assets, management of and access to data generated by new ICT systems, and installation, maintenance and operation of electric vehicle charging systems. 17 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) * Clarifying the ways in which the distribution utility and adjacent market actors, including retailers, transmission system operators, and new DES-related businesses (e.g. aggregators), will interact and coordinate to dispatch DERs. * Considering the design of mechanisms to incentivize long-term investments in network innovation, technology development, and learning, not just short-run efficiency and cost-savings. While each of these important challenges requires proactive regulatory innovation, only the remuneration challenge lies within the scope of this thesis. The evolving nature of the distribution sector will entail new customer demands and uses for the distribution system, new DER and ICT-related cost drivers for the utility, and new opportunities to harness emerging technologies and services to reduce distribution system costs and improve quality of service. These changes will exacerbate several of the fundamental challenges in the economic regulation of distribution utilities introduced in Section 1.1. First, large-scale penetration of DERs within distribution utilities networks will likely increase the total costs of business-as-usual management of the distribution system (that is, a continued "fit-and-forget" grid management strategy) (Cossent, G6mez, & FrIfas, 2009; Cossent et al., 2010). Substantial future investments will be required to fulfill the distribution utility's open access requirements and connect all new DER system users as well as to enable the system to deal with bidirectional load flows, potentially increased volatility in peak demand, and new DG-related system peaks at various voltage levels. Unless these new costs are adequately remunerated, distribution utilities will have strong incentives to block, delay, or oppose new uses of their network. Regulators must therefore be equipped to proactively identify the impacts of new DERrelated network uses on distribution costs and to incentivize utilities to accommodate DERs through appropriate remuneration methods. Second, distribution utilities can also achieve important cost savings by adopting an active system management approach, especially as DG shares increase (Cossent et al., 2009, 2011; Eurelectric, 2013a; Olmos et al., 2009; Poudineh & Jamasb, 2014; Trebolle, et al., 2010). Setting up ICT and advanced grid management infrastructure that allows distribution utilities to more actively manage distribution network configuration and make use of DESs for their daily grid operations will entail substantial upfront capital expenditures (CAPEX) but can also improve system operating efficiency, avoid or defer other CAPEX, and reduce operational expenditures (OPEX) (Poudineh & Jamasb, 2014; Trebolle et al., 2010). Likewise, coordination and management of EV charging within an active system management approach is essential to avoid heavy investments into low18 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) and medium-voltage lines to compensate for local peak demand resulting from high penetration of EVs (Fernandez, et al., 2011). Remuneration must therefore provide incentives for distribution utilities to take efficient advantage of new DER and ICTrelated capabilities and to manage important tradeoffs between CAPEX and OPEX to find the efficient balance between these expenditures. Third, as the sector evolves, distribution utilities are likely to develop more intimate and immediate knowledge about new cost drivers and opportunities than the regulator, heightening information asymmetries and creating new opportunities for strategic behavior. Regulatory tools to overcome the regulator's information disadvantages will thus become even more critical as the distribution sector evolves. Furthermore, while frontier benchmarking and yardstick approaches are often employed for these purposes, these approaches will become much more challenging as the sector evolves. For example, the growth of distributed generation introduces much more heterogeneity between distribution network costs. The availability of solar, wind, biomass/biogas, and combined heat and power resources differs substantially from location to location and is likely to lead to divergent evolution of distribution networks in different regions. In addition, as network uses and drivers of cost rapidly evolve, benchmarking based on past utility performance or cost will no longer provide an accurate estimate of the forward-looking efficient frontier. New forward-looking methods to assist the regulator in determining appropriate allowed revenues while overcoming information asymmetry are thus essential. Finally, the evolving nature of distribution system technologies and uses will further exacerbate the inherent uncertainty facing the regulator. This is a challenge particularly for ex ante remuneration methods, such as revenue caps. As uses of the network and the new technologies available to utilities may evolve quite rapidly, network costs may deviate substantially from ex ante regulatory estimates, leading to two types of error: forecast error and benchmark error. Costs may rise or fall unexpectedly due to new network uses (e.g., the rapid penetration of newly subsidized or newly cost competitive DG), an example of forecast error. Alternatively, the regulator may fail to anticipate the emergence of new cost saving technologies within the regulatory period, leading to benchmark error. In either case, regulators employing ex ante remuneration methods may be at greater risk of either violating the firm participation constraint if cost recovery is too low or leaving significant economic rents to the utility by being too generous in setting the ex ante revenue cap. More frequent ex post reviews and adjustments to remuneration levels can address these challenges, but not without a cost. Frequent ex post revisions of remuneration levels and reopeners of the regulatory contract can create significant regulatory uncertainty and thus may raise the cost of capital for distribution utilities as well as potentially undermining efficiency incentives. 19 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) 1.3 - New Solutions for Regulation of Distribution Utilities Under High Penetrations of Distributed Energy Resources Addressing the combination of regulator challenges discussed in Section 1.2 will require regulators to more widely adopt best practices in network utility regulation. In particular, this thesis proposes a novel process for establishing the allowed revenues of an electricity distribution utility and demonstrates its application as a practical solution to the imminent regulatory challenges discussed above. The proposed method is a combination of three established best practices or "state of the art" regulatory tools: an engineering-based reference network model (RNM) for forward-looking benchmarking of efficient network expenditures (Cossent, 2013; Domingo et al., 2011); an incentive compatible menu of contracts to elicit accurate forecasts from the utility and create incentives for cost saving efficiency efforts (Cossent & G6mez, 2013; Cossent, 2013; Crouch, 2006); and ex post automatic adjustment mechanisms, or "delta factors," to accommodate uncertainty in the evolution of network use and minimize forecast error.' This section summarizes the overall process at a conceptual level and introduces its advantages. Chapter 3 will demonstrate in detail the step-by-step implementation of this process to establish the allowed revenues of a realistic simulated distribution network utility and evaluate its performance. The proposed regulatory process involves ex ante calculation of allowed revenues and establishment of clear rules for ex post evaluation of actual expenditures and adjustments to final allowed revenues. Figure 1-2 depicts the ex ante regulatory process carried out at the beginning of each regulatory period. The key steps are as follows: 1. The process begins with the utility submitting to the regulator a detailed year-byyear forecast of the evolution of network uses over the upcoming regulatory period (Step 1).9 This forecast should at minimum include a set of appropriately justified scenarios covering a range of the likely load and DER penetration levels, including discussion of the most likely geographic evolution of loads and DERs. 8 This general method is first proposed in Cossent (2013a) Chapter 5 and Cossent and G6mez (2013) and is developed further herein, including demonstration of implementation in Chapter 3. The methods herein, particularly the menu of contracts approach, also draw on the practical experience of the UK Office of Gas and Electricity Markets (Ofgem), as published in Ofgem (2009, 2010b, 2013a, 2013c) and related methods. The use of annual automatic adjustment factors to account for deviations from forecasted load growth and other network uses is also demonstrated. I This general method is agnostic as to the length of the regulatory period, although this length should be carefully determined to balance the appropriate incentives for cost savings on the one hand and adjustments to mitigate uncertainty on the other hand. In general, the profit and risk sharing mechanism established by the proposed menu of contracts along with use of adjustment factors should reduce the impacts of uncertainty and enable a longer regulatory period. 20 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Figure 1-2: Proposed regulatory process - ex ante actions Ex ante process (performed only at beginning of regulatory Regulator 1. Submission of utility's forecast for evolution of network uses 2. Analysis of utility forecast / stakeholder comments 3. Regulator's comments on utilit forecast I 4. Revisions to utility forecast AI 5. Final forecast for evolution of network uses 6. Establishment of regulator's ex ante TOTEX estimate 7. Construction of incentive compatible menu of contracts 8. Submission of utility investment plan and estimated TOTEX 9. Calculation of ex ante TOTEX and revenue baselines and ex post sharing factor Utility a 10. Calculation of automatic adjustment factors ("delta factors") to accommodate uncertainty in forecasted network use 11. Final regulatory contract published 2. The regulator then critically reviews this forecast (Step 2). This review may also include a period of open comment on the preliminary forecast by stakeholders. 3. At the conclusion of this review, the regulator will submit clear comments to the utility on required changes or further analysis needed to construct a final forecast (Step 3). 4. Upon receiving this feedback, the utility will then perform any required updates to their scenarios (Step 4) 5. The utility then re-submits a final forecast to the regulator for use throughout the remainder of the regulatory process (Step 5). 6. After receiving the final forecast from the utility, the regulator will then construct their estimate of efficient network expenditures to meet the forecasted evolution of network use over the regulatory period (Step 6). In this 21 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) stage, the regulator will employ a reference network model (RNM), a large-scale distribution network planning tool capable of producing an estimate of efficient costs of expanding and maintaining a network to serve a specified set of network users at prescribed quality levels (i.e., maximum statistical probability of network disruptions, voltage limits, etc.) and considering incentives for reduction of network losses. The RNM described in Domingo et al. (2010) will be used throughout this thesis to demonstrate application of this method, but any suitably rigorous RNM could be employed. The use of an RNM gives the regulator a forward-looking benchmark for efficient network expenditures, helping reduce information asymmetry and manage uncertainty.10 In this stage, the regulator produces their estimate of efficient TOTEX for the regulatory period. In addition, the regulator can now determine the expected portion of TOTEX associated with both CAPEX (including incremental and replacement investments) and OPEX (network maintenance). These shares are used to fix the portion of realized TOTEX that is capitalized into the utility's regulated asset value (or "rate base") used to determine allowed revenues during the ex post process. This TOTEX-based approach to capitalization equalizes incentives for cost-savings across both CAPEX and OPEX, which can be distorted using a CAPEXonly approach to capitalization (see Chapter 3, Section 3.2 for more). 7. The next step involves creation of an incentive compatible menu of profitsharing contracts for the utility (Step 7).11 A menu of contracts specifies an ex ante regulatory allowance as well as a clear rule for ex post evaluation of actual expenditures and adjustments to final remuneration. The menu outlines a continuum of profit-sharing factors (sliding-scale efficiency incentives) wherein values depend on the ratio of the utility's estimate of network costs over the regulatory period to the regulator's estimate derived via use of the RNM in Step 10 Note that as with any other benchmarking method, the cost figures produced by the RNM should not be used directly to establish the utility's allowed revenues. The RNM is a model that simplifies reality and, as a result, generally produces a network solution that is less expensive than an efficient real-world utility is likely to achieve. For example, the RNM can plan the network with perfect foresight given the input forecast, while utility planners may have to adjust plans over time, incurring sunk costs as real-world conditions change. Other constraints faced by a real utility company may also be ignored by the RNM. Regulators should therefore adjust the cost estimates produced by the RNM accordingly. The RNM's performance can be benchmarked by running it against several real world efficient network cases with known expenditures to estimate an appropriate correction factor, or the regulator can employ consultants to 'spot check' the estimates produced by the RNM. 1 Cossent and G6mez (2013) describe a practical method for creation of an incentive compatible menu of contracts, and this thesis builds on that work herein. Additionally, this general approach has been successfully implemented by the Ofgem since the fourth distribution price control review (DPCR4) enacted from 2005-2010 and is now an integral part of Ofgem's RIO framework (Ofgem, 2010c). The UK's approach, known as the Information Quality Incentive (IQI) is described in Crouch (2006) as well in Ofgem (2009, 2010a, 2013c) and Cossent and G6mez (2013). 22 Jesse D. Jenkins (2014) Chapter 1: Economic Regulation of Electricity Distribution Utilities 6. As discussed in Section 1.1, if the design of the menu preserves incentive compatibility, it will provide incentives for the utility to provide a truthful estimate of expected network costs, helping overcome information asymmetries. In addition, incentives for the utility to engage in strategic behavior by inflating estimates of network costs are eliminated. An example menu of contracts is shown in Table 1-3 below. Table 1-3: Example incentive compatible menu of profit-sharing contracts Shaded cells correspond to those for which the ex ante utility forecast matches actual expenditures, demonstrating the incentive compatible nature of this matrix. For any realized value of network costs (i.e. horizontal row in the bottom half of the matrix), the utility will earn the greatest revenues in the case where their realized cost matches their ex ante forecast. Efficiency incentives are also preserved, as lowering realized costs below the utility's forecast (i.e. moving up in a vertical column) will increase the utility's final revenues (and vice versa). This menu uses the following discretionary parameters (see chapter 3, Section 3.3 for details): weight on regulator's estimate = 0.66; Reference sharing factor = 0.7; Sharing factor rate of change = -0.01; Reference additional income = 1.0%. Ratio of firm's cost estimate to regulator's cost estimate Allowed revenues ex ante %of regulator's cost estimate] Sharing factor [%] Additional Income [% of regulator's cost estimate] Ratio of realized ex past expenditures to regulators ex ante estimate [%J 85 90 954.5 100 105 110 115 120 125 90 95 100 105 110 115 120 X,,.,, 96.6 98.3 100.0 101.7 103.4 105.1 106.8 SF 80.0 75.0 70.0 65.0 60.0 55.0 50.0 Al 3.2 2.2 LO -0.2 -1.5 -2.9 -4.4 L. . R% Final ex post adjustment to allowed revenues [%of regulator's ex ante estimate] 8.1 9.5 10.6 11. 12.1 12.5 5.4 6.5 7.4 8.0 8.4 2.6 3.5 4.1 4.5 -0.1 0.5 0.9 .9 05 2.9 2.5 -2.5 -2.9 -3.5 6 -5.6 -6.0 -6.6 -7.5 -8.5 -8.9 -9.5 -10.4 -11.5 -11.1 -11.5 -12.1 -13.0 -14.1 -15.5 -13.9 -14.5 -154 -16.5 -17.9 -19.5 Ap a 6.5 4.0 1e5 -1.0 -3.5 -6.0 -8.5 -13.5 8. Once the menu of contracts is established and published, the utility will then develop their detailed year-by-year investment or business plan for the regulatory period (Step 8). This plan should justify the utility's expected expenditures in light of the final forecast created in Steps 1-5 including expected responses to departures from the central forecast. This plan should outline not only the expenditures but also the expected impact on network users, including quality of service, losses, reliability, and ability to interconnect DERs, as well as other important performance metrics (i.e. environmental metrics, innovation, etc.). The idea, as articulated by Ofgem, is for the business plan to clearly outline the "value for money" network users should expect over the regulatory period (Cossent & G6mez, 2013; Ofgem, 2010b, 2010c, 2013b, 2013c). 23 Jesse D. Jenkins (2014) Chapter 1: Economic Regulation of Electricity Distribution Utilities 9. Next, the regulator can determine the ex ante TOTEX baseline (the regulator's estimate of the efficient frontier for this utility), calculate the allowed revenue baseline, and publish the ex post efficiency incentive (sharing factor) for the utility. The TOTEX baseline is determined by applying a weighted average of the regulator's estimate of the efficient frontier (based on the RNM in Step 6) and the utility's estimate (submitted in their business plan in Step 8). The weighting of the two estimates is specified by the regulator as part of the creation of the menu of contracts. This TOTEX baseline is the final ex ante estimate of efficient expenditures against which realized expenditures are compared to determine the efficiency incentive the utility earns (or incurs, in the case of over-spending). Next, the ratio between the utility's estimate of TOTEX and the regulator's estimate determines the sharing factor and additional income allowance for the utility as defined by the menu of contracts produced in Step 7. Finally, an ex ante allowed revenue baseline is calculated using financial accounting to determine the appropriate revenues necessary to remunerate the utility for operational expenditures (OPEX) as well as depreciation and cost of capital allowances based on the regulated asset value. The utility then has a clear expectation of how their revenues will evolve over the regulatory period, providing strong regulatory certainty and clear incentives for efficient management of network costs. Table 1-4 Example of TOTEX and revenue baseline calculations Menu of contract corresponds to example in Table 1-3. See Chapter 3, Section 3-4 for detailed methodology. Year1 Year2 Year3 Year4 Year5 NPV Regulator's estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 Utility's estimate $19.69 $20.18 $20.68 $21.20 $21.73 $85.73 TOTEX ESTIMATES TOTEX AND REVENUE BASELINE AND MENU OF CONTRACTS PARAMETERS TOTEX baseline $17.52 $17.96 $18.41 $18.87 $19.34 $76.30 Revenue baseline $32.03 $32.18 $32.37 $32.58 $32.82 $134.5 Ratio 1.2 Sharing Factor 50% Additional income -$3.14 10. While the use of an RNM and menu of contracts produces a clear revenue determination for each utility taking into account forecasted cost of capital, evolution of network uses, and network component costs, the ex ante nature of this regime means there will always be uncertainty regarding the accuracy of these forecasts. To manage this uncertainty and mitigate the impacts of forecast error, the regulator next employs the RNM to estimate network costs across a range of uncertainty scenarios that capture the likely range of potential evolution of load, DG penetration, or other important and uncertain cost drivers 24 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) (Step 10). Regression analysis of the resulting estimated network costs can then be used to determine the relationship between deviations in cost driver values and total network costs. These coefficients, which this thesis calls "delta factors," prescribe simple formulas to adjust the estimated TOTEX baseline ex post based on the realized evolution of network uses. This use of delta factors minimizes the impact of forecast error and thus reduces the risk that the revenue determination will need to be re-opened, increasing regulatory certainty.' 2 These delta factors also reduce incentives for the utility to slow interconnection of DERs by ensuring cost recovery even if DER penetration grows more rapidly than expected. 11. The ex ante regulatory process is now complete, the regulator publishes the final regulatory contract for the utility for the duration of this regulatory period (Step 11). The ex ante regulatory process is now complete and the regulatory period begins. At the conclusion of each year during the regulatory period, an ex ante regulatory commences to adjust the utility's allowed revenues in light of the realized evolution of system uses and utility expenditures. Figure 1-3 illustrates this annual ex post regulatory process, which is comparatively simple. Figure 1-3: Proposed regulatory process - annual ex post actions Regulator Utility A. ULIIILY makes expenditures and reportsto regulator 2. Audit of TOTEX and automatic adjustments to TOTEX baseline computed due to deviation from forecasted network use 3. Efficiency incentive computed as per menu of contracts and adjustments made to allowed revenues 4. Forward-looking revenue adjustments computed to true up revenues 5. Adjusted revenue allowance published TV 2 Similar mechanisms have been employed by the Comisi6n Nacional de Energia (CNE) in Spain to account for deviations in load growth-the "Y factor," calculated using an RNM (Cossent & G6mez, 2013)-and by Ofgem in the UK to account for DG penetration during DPCR4 and DPCR5-the "DG incentive." The method used to calculate of the adjustment factor is not transparent (Ofgem, 2010a). 25 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) 1. First, the utility submits a detailed report on actual investment and operational expenditures (the utility's realized TOTEX) as well as details on the evolution of system uses (i.e., load growth and penetration of DER) (Step 1). 2. Next, the regulator will audit these reports to ensure their accuracy, and then compute the automatic adjustments to the ex ante TOTEX baseline to account for any differences in actual network use as compared to the ex ante forecast (Step 2). The delta factors for each key network use (i.e., load growth and DER penetration) are multiplied by the difference in realized network use from the forecast, producing an adjustment to the net present value (NPV) of efficient TOTEX. Since the utility is not expected to make all of the expenditures to accommodate this deviation from forecast in the immediate year, this total NPV adjustment is then converted into a stream of annual adjustments to the TOTEX baseline for the remaining years in the regulatory period. 3. After calculating the adjusted TOTEX baseline, the regulator then compares the utility's realized TOTEX over the last year with the adjusted TOTEX baseline, and the efficiency incentive is calculated by multiplying the firm's sharing factor (from the menu of contracts) by the difference between the adjusted TOTEX baseline and the realized TOTEX (Step 3). The efficiency incentive is the portion of the realized over/under-spend shared by the utility's shareholders. The utility's ex post allowed TOTEX is thus the utility's realized TOTEX less this efficiency incentive. Next, the regulator computes the revenue allowance based on this allowed TOTEX, including allowed maintenance expenses and allowances for the depreciation of the regulated asset value and payment of debt and equity (cost of capital allowance). 4. Since revenues have already been collected over the course of the recently concluded year, the regulator must adjust the utility's revenue allowance in future years to "true up" the collected revenues and the ex post revenue allowance computed above (Step 4). The deficit or surplus in collected revenues is calculated as the difference between allowed cost recovery and collected revenues. Next, an annual stream of adjustments to true up future allowed revenues is calculated so as to ensure that the NPV of adjustments to future revenues corrects for the surplus or deficit in collected revenues over the recently concluded year. This true up is applied as a stream of annual adjustments, rather than a single lump sum correction, so as to smooth the impact on rates and avoid discontinuous rate increase/decreases. 26 Jesse D. Jenkins (2014) Chapter 1: Economic Regulation of Electricity Distribution Utilities 5. Finally, the regulator publishes the adjusted revenue allowance (Step 5), including the true up adjustments calculated above, and the regulatory period continues. Table 1-5 below illustrates an example of the final profits (as net earnings before interest and taxes) and return on equity for a range of realized network expenditures using this approach. This table demonstrates the incentive compatibility of the full regulatory process, as the utility makes the most profit whenever realized expenditures align with firm's ex ante estimate of efficient costs (the shaded cells). Additionally, the firm can always increase their net profits by reducing costs below their ex ante estimate, demonstrating the efficiency incentives created by this approach. Table 1-5: Example of final profit and return on equity across a range of realized expenditures EX ANTE TOTEX ESTIMATES Regulator Estimate NPV M$ $71.4 $71.4 $71.4 $71.4 Firm's Estimate NPV M$ $64.3 $71.4 $78.6 $85.7 % 90 100 110 120 Ratio (Firm/Regulator) EX ANTE REGULATORY CONTRACT (w = 0.66; SFref = 0.7 SFro = -0.01; Alref = 1.0) TOTEX baseline NPV M$ $69.0 $71.4 $73.9 $76.3 Revenue baseline NPV M$ $136.3 $135.9 $135.3 $134.5 % 80 70 60 50 NPV M$ $2.3 $0.7 -$1.1 -$3.1 Sharing Factor Additional Income EX POST RESULTS Realized TOTEX (NPV M$) TOTEX Ratio (Realized/ Regulator) $64.3 90 $71.4 100 $20.36 / 6.58 $78.6 110 $16.67 / 5.38 $17.44 / 5.60 $85.7 120 $12.99 / 4.18 $14.24 / 4.56 Earnings Before Interest and Taxes (NPV M$) / Return on Equity (%) $23.83/ 7.70 $23.16 / 7.47 $22.00/ 7.10 $20.44 / 6.57 $19.76 / 6.35 $17.53 / 5.61 $15.00 / 4.79 This proposed regulatory method equips the regulator with several tools to confront the regulatory challenges created by the evolution of the electricity distribution sector and the increasing penetration of distributed energy resources. The regulatory regime proposed herein helps overcome information asymmetry by equipping the regulator with a reference network model with which to develop their estimate of efficient network expenditures. The RNM emulates the network planning 27 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) practices of an efficient utility and can help the regulator develop more accurate estimates of future network costs given the expected evolution of network uses. Additionally, as the model can be designed to accommodate expected evolutions in network use, technology costs, and network management practices, the RNM creates a forward-looking benchmark for efficient total network expenditures (TOTEX), reducing the uncertainty facing the regulator. In effect, the RNM gives the regulator a tool with which to "peer into the future," a crucial ability in ex ante regulatory approaches. This forward-looking capability stands in contrast to statistical benchmarking techniques, which rely on backward-looking analysis of realized expenditures during prior regulatory periods and thus cannot capture the dynamic changes now unfolding in the electricity distribution sector. The RNM can also be used to explore a range of possible scenarios for the evolution of network uses (i.e., load growth and DER penetration). The model results can then be used to compute delta factors, simple formulas to automatically adjust the efficient TOTEX baseline in light of the realized evolution of network use. These delta factors effectively minimize the impacts of forecast errors, a significant advantage given increased uncertainty about the likely evolution of network use over the coming years. Combining the use of an RNM with an incentive compatible menu of contracts further reduces information asymmetry by incentivizing the utility to submit their most accurate estimate of future network expenditures. The incentive compatible property of the menu of contracts thus eliminates incentives for the utility to engage in strategic behavior by inflating their estimate of necessary TOTEX, a significant advantage over other ex ante regulatory approaches that do not employ a menu of contracts. The profit sharing parameter established by the menu of contracts creates clear incentives for the utility to seek cost-saving efficiency measures throughout the regulatory period. This profit sharing incentive gives the utility's management and shareholders a direct stake in cost-saving measures and thus overcomes the moral hazard problem that plagues cost-of-service regulation. Furthermore, the regulator has the flexibility to establish the desired strength of the efficiency incentives to manage the tradeoffs between allocative efficiency (extracting rents from the utility) and X-efficiency (providing incentives for cost savings). Finally, by selecting the strength of the profit sharing factor, the regulator also has the ability to mitigate the impacts of benchmark error - i.e., an error in the regulator's estimate of efficient TOTEX (irrespective of the evolution of network use). The lower the sharing factor, the closer the regulatory contract becomes to a cost-ofservice contract, and thus the less sensitive the firm's profits are to differences in forecasted and realized costs, and vice versa. The regulator can thus select an appropriate sharing factor based on their confidence in the accuracy of their forecasts of efficient network expenditures. 28 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Note that the method described above only considers the establishment of allowed TOTEX and the primary allowed revenues. This regulatory process must be combined with careful consideration of performance-based metrics for quality of service and other desired outputs such improvements in reliability, reduction of losses, achievement of environmental objectives. See Cossent (2013a), Malkin & Centolella (2013), and Ofgem (2010c) for more on output or performance-based incentives. Furthermore, while a well-designed menu of contracts provides strong incentives for efficiency and will encourage the utility to pursue novel and innovative approaches to network investment and management, additional incentives for long-term innovation may be necessary. Ex ante regulation of revenues with strong sharing-factors can place the short-run expenditures needed to unlock long-term innovations at odds with the utility's incentives for cost-savings (Bauknecht, 2011). When these dynamics are combined with the considerable knowledge spillovers, long time horizons, and inherent risks associated with the development of new network technologies and demonstration of novel practices (Lester & Hart, 2012; Lo Schiavo et al., 2013), this can lead to a substantial under-investment in network innovation. Regulators may therefore need to employ explicit incentives for innovation, including input-based incentives (such as an R&D cost pass-through), output-based incentives (financial incentives for adoption rates of novel technologies or practices), or competitive innovation funds (such as the UK's Low-carbon Innovation Fund). For discussion of network innovation incentives, see Bauknecht (2011), Lester & Hart (2012), Lo Schiavo et al. (2013), and Ofgem (2010c). Unfortunately, both of these important aspects of network regulation lie outside the scope of this thesis. 1.4 - Structure of the Thesis The remainder of this thesis demonstrates the regulatory process introduced above in detail. First, Chapter 2 describes the simulation of a realistic, large-scale electricity distribution network for use in demonstration of the novel regulatory method proposed herein. Using a reference network model, a base network is constructed for a roughly 120 square kilometer region of Denver, Colorado. A range of scenarios for the evolution of network uses is then developed and the network is expanded to accommodate the realized network uses in each scenario. The estimates of network costs (CAPEX and OPEX) produced by this simulation are then used throughout Chapter 3, which demonstrates, step-by-step, the implementation of this regulatory process. The key stages of the regulatory process are described in detail and the performance of the regulatory process is evaluated. 29 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) References Averch, H., & Johnson, L. L. (1962). Behavior of the Firm Under Regulatory Constraint. American Economic Review, 52(Dec.), 1052-1069. Bauknecht, D. (2011). Incentive Regulation and Network Innovations. Retrieved from http://cadmus.eui.eu/bitstream/handle/1814/15481/RSCAS_2011_02.pdf?sequenc e=1 Beesley, M. E., & Littlechild, S. C. (1989). The Regulation of Privatized Monopolies in the United Kingdom. The RAND Journal of Economics, 20(3), 454. doi:10.2307/2555582 Bharatkumar, A., Burger, S., Jenkins, J. D., Dantec, J. Le, Perez-Arriaga, I., Tabors, R. D., & Batlle, C. (2014, Forthcoming). The MIT Utility of the Future Future of Phase I Report. Cambridge, MA. Bolton, P., & Dewatripont, M. (2005). Contract Theory. Cambridge, MA: MIT Press. Bushnell, J., & Borenstein, S. (2000). Electricity restructuring: Deregulation or reregulation? Regulation, 23(2), 46-52. Cossent, R. (2013). Economic Regulation of Distribution System Operators and its Adaptation to the Penetration of Distributed Energy Resources and Smart Grid Technologies. Comillas Universidad Pontificia. Cossent, R., & G6mez, T. (2013). Implementing incentive compatible menus of contracts to regulate electricity distribution investments. Utilities Policy, 27, 28-38. doi:10.1016/j.jup.2013.09.002 Cossent, R., G6mez, T., & Frias, P. (2009). Towards a future with large penetration of distributed generation: Is the current regulation of electricity distribution ready? Regulatory recommendations under a European perspective. Energy Policy, 37(3), 1145-1155. doi:10.1016/j.enpol.2008.11.011 Cossent, R., Olmos, L., G6mez, T., Mateo, C., & Frias, P. (2011). Mitigating the Impact of Distributed Generation Advanced Response Options. International Transactions on Electrical Energy Systems, 21(6), 1869-1888. doi:10.1002/etep.503 Crouch, M. (2006). Investment under RPI-X: Practical experience with an incentive compatible approach in the GB electricity distribution sector. Utilities Policy, 14(4), 240-244. doi:10.1016/j.jup.2006.05.005 30 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Domingo, C. M., G6mez, T., Sanchez-Miralles, A., Peco, J., & Martinez, A. C. (2011). A Reference Network Model for Large-Scale Street Map Generation. IEEE Transactions on Power Systems, 26(1), 190-197. Eurelectric. (2013a). Active Distribution System Management A key toolfor the smooth integration of distributed generation. Brussels, Belgium. Retrieved from http://www.eurelectric.org/media/74356/asm-full-reportdiscussionpaperfinal2013-030-0117-01-e.pdf Eurelectric. (2013b). Utilities: Powerhouses of Innovation. Brussels, Belgium. Retrieved from http://www.eurelectric.org/media/79178/utiltiespowerhouseofinnovationfull _reportfinal-2013-104-0001-01-e.pdf Fernandez, L. P., G6mez, T., Cossent, R., Domingo, C. M., & Frias, P. (2011). Assessment of the Impact of Plug-in Electric Vehicles on Distribution Networks. IEEE Transactions on Power Systems, 26(1), 206-213. G6mez, T. (2013a). Electricity Distribution. In 1.J. Perez-Arriaga (Ed.), Regulation of the Power Sector. London: Springer-Verlag. doi:10.1007/978-1-4471-5034-3 G6mez, T. (2013b). Monopoly Regulation. In 1.J. Perez-Arriaga (Ed.), Regulation of the Power Sector. London: Springer-Verlag. doi:10.1007/978-1-4471-5034-3 Jamasb, T., Nillesen, P., & Pollitt, M. (2003). Gaming the Regulator: A Survey. The Electricity Journal, 16(10), 68-80. Jamasb, T., Nillesen, P., & Pollitt, M. (2004). Strategic behaviour under regulatory benchmarking. Energy Economics, 26(5), 825-843. doi:10.1016/j.eneco.2004.04.029 Jamasb, T., & Pollitt, M. (2001). Benchmarking and regulation : international electricity experience. Utilities Policy, 9(2001), 107-130. Jamasb, T., & Pollitt, M. (2003). International benchmarking and regulation: an application to European electricity distribution utilities. Energy Policy, 31(15), 1609-1622. doi:10.1016/SO301-4215(02)00226-4 Jamasb, T., & Pollitt, M. (2007). Incentive regulation of electricity distribution networks: Lessons of experience from Britain. Energy Policy, 35(12), 6163-6187. doi:10.1016/j.enpol.2007.06.022 31 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Jamasb, T., & Pollitt, M. (2008). Reference models and incentive regulation of electricity distribution networks: An evaluation of Sweden's Network Performance Assessment Model (NPAM). Energy Policy, 36(5), 1788-1801. doi:10.1016/j.enpol.2008.01.034 Joskow, P. L. (2005). Regulation of Natural Monopolies (No. 05-008). Cambridge, MA. Retrieved from http://web.mit.edu/ceepr/www/publications/workingpapers/2005014.pdf Joskow, P. L. (2008). Incentive Regulation and Its Application to Electricity Networks. Review of Network Economics, 7(4), 547-560. doi: 10.2202/1446-9022.1161 Joskow, P. L. (2010). Market Imperfections versus Regulatory Imperfections. CESifo DICE Report, 8(3), 3-7. Retrieved from http://economics.mit.edu/files/5619 Joskow, P. L. (2013). Incentive Regulation in Theory and Practice: Electricity Distribution and Transmission Networks. In N. L. Rose (Ed.), Economic Regulation and Its Reform: What Have We Learned (Forthcoming). Chicago, IL: University of Chicago Press. Retrieved from http://www.nber.org/chapters/c12566 Joskow, P. L., & Schmalensee, R. (1986). Incentive Regulation For Electric Utilities. Yale Journal on Regulation, 4(1), 1-49. Kind, P. (2013). Disruptive Challenges: Financial Implications and Strategic Responses to a Changing Retail Electric Business. Washington, DC. Retrieved from http://www.eei.org/ourissues/finance/Documents/disruptivechallenges.pdf Laffont, J.-J., & Tirole, J. (1993). A Theory of Incentives in Procurement and Regulation (p. 705). Cambridge, MA: MIT Press. Lester, R. K., & Hart, D. M. (2012). Unlocking energy innovation: how America can build a low-cost, low-carbon energy system. Cambridge, Mass.: MIT Press. Lo Schiavo, L., Delfanti, M., Fumagalli, E., & Olivieri, V. (2013). Changing the regulation for regulating the change: Innovation-driven regulatory developments for smart grids, smart metering and e-mobility in Italy. Energy Policy, 57, 506-517. doi:10.1016/j.enpol.2013.02.022 Malkin, D., & Centolella, P. A. (2013). Results-Based Regulation: A Modern Approach to Modernize the Grid. Atlanta, GA. Retrieved from http://www.gedigitalenergy.com/regulation/ 32 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Newcomb, J., Lacy, V., & Hansen, L. (2013). New Business Models for the Distribution Edge: the Transition From Value Chain to Value Constellation. Boulder, CO. Retrieved from http://www.rmi.org/NEW_BUSINESSMODELS Ofgem. (2009). Electricity Distribution Price Control Review Methodology and Initial Results Paper. London. Ofgem. (2010a). Electricity Distribution Price Control Review 5: Special conditions of the Electricity Distribution License. London. Ofgem. (2010b). Handbook for implementing the RIIO model. London. Ofgem. (2010c). RIIO: A new way to regulate energy networks. London. Ofgem. (2013a). RIIO TI Financial Model (Electric). Network Regulation - the RIIO Model. Retrieved October 05, 2014, from https://www.ofgem.gov.uk/network-regulation-riio-model/price-contros-financial-model-pcfm/riio-tl-financial-model-electric Ofgem. (2013b). Strategy decision for the RIIO-ED1 electricity distribution price control: Business plans and proportionate treatment. London. Ofgem. (2013c). Strategy decision for the RIIO-ED1 electricity distribution price control: Overview. London. Olmos, L., Cossent, R., G6mez, T., Mateo, C., Joode, J. De, Scheepers, M., ... Gerhardt, N. (2009). Case studies of system costs of distribution areas (No. WP 4, Deliverable 5). Petten, the Netherlands. Retrieved from http://www.improgres.org/fileadmin/improgres/user/docs/D5_casestudies-of-sy stemcosts.pdf Perez-Arriaga, I. (2013). Challenges in Power Sector Regulation. In 1.J. Perez-Arriaga (Ed.), Regulation of the Power Sector (pp. 647-678). London: Springer-Verlag. doi:10.1007/978-1-4471-5034-3 Perez-Arriaga, I., Ruester, S., Schwenen, S., Batlle, C., & Glachant, J.-M. (2013). From Distribution Networks to Smart Distribution Systems: Rethinking the Regulation of European Electricity DSOs. Florence. doi:10.2870/78510 Poudineh, R., & Jamasb, T. (2014). Distributed generation, storage, demand response and energy efficiency as alternatives to grid capacity enhancement. Energy Policy, 67, 222-231. doi:10.1016/j.enpol.2013.11.073 Rudnick, H., Arnau, A., Mocarquer, S., & Voscoboinik, E. (2007). Stimulating Efficient Distribution. IEEE Power and Energy Magazine, (august), 50-67. 33 Chapter 1: Economic Regulation of Electricity Distribution Utilities Jesse D. Jenkins (2014) Schmalensee, R. (1989). Good Regulatory Regimes. The RAND Journal of Economics, 20(3), 417-436. doi: 10.2307/2555580 Shleifer, A. (1985). A theory of yardstick competition. RAND Journal of Economics, 16(3), 319-328. Trebolle, D., G6mez, T., Cossent, R., & Frias, P. (2010). Distribution planning with reliability options for distributed generation. Electric Power Systems Research, 80(2), 222-229. doi:10.1016/j.epsr.2009.09.004 34 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) 2 - Constructing a Simulated Electricity Distribution Network This chapter describes the simulation of a realistic, large-scale urban distribution network used to demonstrate the novel regulatory process proposed in this thesis. The methodology employed for this simulation was originally developed in Vergara, et al., (2014) and employs the reference network model (RNM) described in Domingo et al., (2011). This thesis modifies and extends these methods to create a realistic simulated network for a roughly 120 square kilometer (km-sq) section of Denver, Colorado, encompassing more than 27,000 individual load points and approximately 468,000 kilowatts (kW) of peak load (see Table 2-1). Several scenarios for the growth of this network are also simulated, capturing the network expansion, reinforcements, and maintenance costs necessary to accommodate a range of possible increases in loads and the penetration of distributed solar photovoltaic (PV) generators. Table 2-1: Characteristics of simulated network Location Denver, Colorado Simulation area 120.3 km-sq Population density 1,580.5 persons/km-sq Estimated population 190,187 Persons Estimated load power density - Base network Estimated peak power demand - Base network 3,890 kW/km-sq 468,079 kW Load Points - Base Network LV Industrial MV HV Total 0 212 42 254 Commercial 6,263 1,274 63 7,600 Residential 18,788 637 0 19,425 Total 25,051 2,123 105 27,279 2.1 - Specification of Simulation Parameters The creation of the simulated distribution system begins with the specification of key simulation parameters describing the composition and characteristics of network users (loads and DG) and other characteristics of the network. The first set of parameters specifies the power density of loads within the simulated network and the average peak active power and average reactive power factor for loads at each voltage level (Table 2-2). The power density parameters are based on the population density of the simulation location. The population density of Denver is 1,580.5 persons per km-sq (U.S. Census Bureau, 2011) and there are 2.3 persons per habitation in the city (U.S. Census Bureau, 2014). At an average peak demand of 1.84 kilowatts (kW) 35 Jesse D. Jenkins (2014) Chapter 2: Constructing a Simulated Electricity Distribution Network 1 per habitation, consistent with the typical residential customer served by Xcel Energy, the distribution utility serving Denver, this yields a residential demand density of 1,249 kW per km-sq. Residential customers make up 32 percent of Xcel Energy's Colorado retail electricity sales (U.S. Energy Information Administration, 2012), so the residential demand density is scaled up to account for non-residential loads, leading to a total load power density of 3,890 kW per km-sq. This total load density is then allocated across voltage levels as follows: 32 percent low-voltage (LV), 45 percent medium-voltage (MV), and 22 percent high-voltage (HV). Table 2-2: Load power parameters for simulated network Unit LV MV HV kW/km-sq 1,249 1,765 876 kW 6 100 1,000 Load average power factor p.u. 1 1 1 Load StDev power p.u. 0.28 0.2 0.2 Load StDev energy p.u. 0.28 0.2 0.2 Load StDev PF p.u. 0.1 0.125 0.2 Parameter Load power density Load point average power Note that an estimated 7 percent of Xcel Energy's Colorado customers reside in the 120 km-sq region of Denver encompassed by this simulated network.3 The estimated power density derived above yields a total peak power demand of 468,079 kW in the simulated 4 network, which is also 7 percent of Xcel's actual total summer peak demand, verifying that this method yields a realistic power demand density. Both the average peak power and power factor for each individual load point in the simulation is determined by random sampling from a truncated normal distribution with mean and standard deviation specified for each voltage level as in Table 2-2. The average power for each load point are set to be representative of typical LV, MV, and HV 1 Xcel Energy operates in Colorado as Public Service Company of Colorado. Xcel's Colorado residential customers consume an average of 659 kWh during the summer months (Colorado Department of Regulatory Agencies, 2014), for an average power demand of 0.92 kW. At a load factor of 0.5, consistent with the average of residential load profiles used in this simulation, that yields an average peak residential demand of 1.84 kW. 2 Data was unavailable on the actual allocation of power density at each voltage level for Xcel Energy's service territory. These values were selected to match the share of residential, commercial, and industrial loads in Xcel's Colorado territory (U.S. Energy Information Administration, 2012), using residential loads as a proxy for LV customers, commercial loads as a proxy for MV customers, and industrial loads as a proxy for HV customers. 3 Xcel Energy has 1,150,181 residential customers in Colorado (Navigant Consulting, 2010). The simulated network spans 120 sq-km at a population density of 1,580.5 persons/km-sq, for an estimated population of 190,190 persons. Assuming 2.3 persons per habitation, that yields 81,745 residential electricity customers in the simulated network, or 7.1 percent of Xcel's total Colorado residential customers. 4 Xcel's 2010 Colorado summer peak demand was 6,262,000 kW (Navigant Consulting, 2010). 36 Jesse D. Jenkins (2014) Chapter 2: Constructing a Simulated Electricity Distribution Network customers in Xcel's service territory.5 The average power factor for each load point is set to 1.0. Peak power is assumed to vary more significantly for lower voltages, while power factor varies more significantly at higher voltages. A minimum peak power of 1 kW is specified for each load point to prevent unrealistically small loads at the far "left tail" of the distribution. 2.2 - Assigning Load Profiles Three types of loads are considered by this simulation: residential, commercial, and industrial loads. Load points are divided among these customer types based on the shares specified for each voltage level in Table 2-3, and each load point is assigned one of ten different 48-hour load profiles for each customer type. The shares of loads by customer type were selected to ensure that the share of total annual electricity consumption for industrial, commercial, and residential consumers in the simulated network closely matches the real distribution of retail electricity sales in Xcel's Colorado service territory.6 Table 2-3: Distribution of load profiles in simulated network by load type Parameter Unit LV MV HV Industrial profile share p.u. 0 0.1 0.4 Commercial profile share p.u. 0.25 0.6 0.6 Residential profile share p.u. 0.75 0.3 0 Load profiles for residential and commercial customers are derived from simulated hourby-hour annual load profile data from Department of Energy (DOE) reference building models and correspond to TMY3 meteorological database characteristics for Denver (National Renewable Energy Laboratory, 2013), yielding realistic load profiles specific to the simulation area.7 As the DOE dataset does not include industrial load profiles, ten different load profiles are created to approximate industrial loads. 8 The 48 hour profiles selected for the simulation correspond to two non-consecutive days in the annual DOE s Xcel segments commercial and industrial customers into three classes based on peak contracted demand: less than 25 kW; 25-200 kW; and greater than 200 kW. This thesis assumes these values correspond to LV, MV, and HV connections. 6The share of total electricity sales by customer class for Xcel's Colorado territory is as follows: 32 percent residential; 45 percent commercial; 22 percent industrial (U.S. Energy Information Administration, 2012). 7 The DOE reference building load profile database contains four residential building load profiles. Six additional residential load profiles are created by altering these base load profiles to yield a total of ten different residential profiles. Ten commercial load profiles are selected from the 16 available commercial profiles in the DOE dataset. 8 Since industrial load profiles are very dependent of the particular process they supply, a nearly-constant consumption with some hourly variability can be assumed. Industrial profiles are therefore constructed as a random walk around a base demand level. 37 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) dataset selected to match (1) the day of peak net power withdrawal and (2) the day of peak power injection assuming penetration of photovoltaic (PV) generators in the network.9 These two days capture the extremes in power flow to which the distribution network must be designed (see Figures 2-1, 2-2, and 2-3). Figure 2-1: Industrial load profiles Peak Power Withdrawal Day Peak Power Injection Day 1.2 1 0.8 0.6 -Industrial 1 -industrial 2 -Industrial 3 -Industrial 4 -Industrial 5 0,4 -Industrial 6 -Industrial 7 -industrial 8 Industrial 9 0,2 -Industrial 10 ...Average 0 1 2 3 4 5 6 7 8 9 10111213141516171819 20 212223 24 25 26 27 28 29 30 3132 33 34 35 36 37 38 39 40 4142 43 44 45 46 47 48 Hour Hour Figure 2-2: Commercial load profiles Peak Power Withdrawal Day Peak Power Injection Day 1.2 0.8 S0.6 K -Commercial 1 -Commercial 2 -Commercial 3 -Commercial 4 Commercial 5 **. 0.4 Commercial 6 -Commercial 7 Commercial a X'.. Commercial 9 0.2 -Commercial --- 10 Average 0 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Hour 252627282930313233343536373839404142434445464748 Hour 9 Given the DOE reference building loads and annual PV production data Denver, peak net power demand occurs at 18:00 hours on July 2 6 th and peak reverse power flow occurs at 13:00 hours on March 38 1 1 th. Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Figure 2-3: Residential load profiles Peak Power Withdrawal Day Peak Power Injection Day 1.2 0.8 I -Residential 2 - Residential 3 0.6 0.4 -Residential M, X. -.- -Residential 4 -Residential 5 - Residential -Residential 6 7 -Residential B Residential 9 -- Residential 10 *,Average 0 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 25 26 27 28 29 30 3132 33 34 35 36 37 38 3940414243 44454647 48 Hour Hour 2.3 - Determining Network Topology and Assigning Geographic Location of Loads The simulation method ensures a realistic network topography by using real street maps for a 120 square kilometer portion of Denver (Figure 2-4) as a "scaffold" to constrain the location of network users. The street map is first scanned and the layout of streets is recognized (Figure 2-5). The layout of streets is then used as a proxy for the density of network connection points by randomly assigning each load point to a specific geographic coordinate along one of the recognized streets with an equal probability per unit of street length (Figure 2-6). Load points are assigned a location using random sampling without replacement. The location of the primary transmission interconnection substation is assigned as per specification in the configuration files. 39 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Figure 2-4: Street map of simulated distribution network area (Denver, Colorado) W2 I k 231d I A.e Hoh~~ ~ 444U#~412~ 4p E W ss 1C114 r6 2v ?Nho A. 1 A.e Elobt t 31d A A 23rd A. I .4 t-UL 16" E Asvw I tk AT, 011 t(t C' WILST '6T h % of. A t PIT, A" f 14th Aw 13th144 Wea DEVE E 13thA F4h A" 1*122 A1 h M P~ Low.,A Pe -12 CU ®441 f 61h A.e 2 *4 4.. t' Comm A4* E ALAh A$2 ftAv 1.1 Awe I * 1412.441 4.. W III A V.1 E Oth Av A5 1221 :1 FAhamwds A e fa 1Av IFI AWlm Ave E F14.qc 4 -J I viisI A WSA * OI.. IMr id.A,, ~ il* y S.WfiM '5 121144124. 11212*44e I o en ~2I$ 4412121A- t 12*44121 A " tsaff Po* WJ1. I 444 44 IAsh A44 WIf Ae 4c4o .4 PIN * 12.P1 124.2 12 11*12A.e 0 G 4 APA,4 E Y*AW 44412,44444 112 14 J 4 A 12- 44 ~12 40 Ev Yake Awe EY*leA- Jesse D. Jenkins (2014) Chapter 2: Constructing a Simulated Electricity Distribution Network Figure 2-5: Layout of streets is recognized creating "scaffold" for network topology I'IF '. CI C'~F'ih ~ ~~~ F~~ lii "It_"/,- OIF~hII~ il~ 1 ',I11 N11h'hj Ij -1'i li III'l Fl i I' .1. iLI'~ l ii iF~-F-FFPFI L-~l -F -'-FFFILIIFI~ , IlL CD~ 0ii iFI IiILIIFII~iI FF TiF 17 ~~)'.F~ 'IL1I~~' I IIiI' CIIII~IF IIFII- I IFFI ii 'F''"""""F~''F1FF ~iii~ii~II~ hIIi~ F~lIgIIIlF11111I~~~LI 111111.1 I F F111FFIFIIIIF~'___ lFI 1111 1 ii'iIii IFFFIJ 7 I 3 U 1 11F 1, F ~ ~I: ~ ~ IIII 111Fll1 0 UI FLF I-11: ICi [11j 'i~~~~~~~~~~~1 j If'""'lI' hj~'iIIL I1111II11I:1A01 11111FF~ ~~ JI III~~~ Ii'i'i ~ I- II____ 'FI II____ d 111 i F-U 1 0 ~ ~ ~ ~ ~ ~~011 F'1' ni' u '{ ~ IF ~ui ~ 11~1II'F ~ '0~ ' ~1~~'N~ ~~ IINVII 'N bl I F'1111 11110V't W ~ FF~l~FL I]l~ i1IJI I'uNNFNN~I7~F'F 1 1ill " I 1]1 t l 111117 '~ 14J~ I F ~ ~ IFIuiillll ',FF"F II'I ____L~ F II I E-11 ihF~ ''u j F I IJii~l~ Iii Ii,-1 : c 1 -IC -o____ 1'll J -I~LIII~II 10 0111111 111 ll l Fl 1111 f00 I111 JiF 1111 iIIILilILli~ 11 fI' -1 1n oiF 11111111, : ' 'F' 1 - 1 19F 0l l I -i111111 *-0"1 LI(iI F1=( Q1111 LI FIJI111 F 11 1-1- 1i/~u~ I_ J, IFIII, F':11 /i]~ Fl IiF '' Ficu i)i1l)'1L [11 F~ h~nihIlI~ hNhhII I Ii HI "[1''1-F'iFIl 1lflf11.11111 11FF '' III lF ~FIi ~ III' 1 u I 1111111111 i'"0111I/l 11 L __17__l011 1 111 1l 0 lo Hu FiIIIW1I ,NNIIuiNh ~~~~~111 ,l1,nIh 111i-lihNh111 I~ ~~~ ~ ~~ 1011 qIIPI1 11_ 1.1111 011 1:1:0 FF~~~~~~~~~~~~~~i~~~~~~~~~ 1111 II 'F ""FL 2L 1I 001111: :0 F ~ ""I LJINII 1fP N'u"- LII1 1111 71 jiI]~ii1]i C' IQ F'IIII I ]~~ IDl I I iI-)II 0 F CIP~hihIII i Ii -I U~P FIFF CIFFF 1:1F iIlFFIuiI~IFF' Cill 11111 lll F.-F-, "F - 111111011 v 41 -Ii F'FII '''I '' I,-"~ _7t5~hu~l~~~Il'i~i~~I ' 111111 I-lu1 U' 'F1 [1[ ,, fI Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Figure 2-6: Network users are assigned along street map scaffold LV customers shown as small black dots; MV customers medium-sized blue dots; HV customers larger orange dots; primary transmission substation shown as green triangle. *'A .0.JIM -,. I 64 - .4 4 so 4 *?oI it 4 4..4:AV2'F so 1A4 lb of*. : # IT 42 '- 0.I Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) 2.4 - Constructing the Network with the Reference Network Model After the location and profiles of load points are determined, a reference network model (RNM) is used to construct the base distribution network. A description of the model can be found in Domingo et al. (2011). This model was developed by Comillas University in collaboration with the Spanish national regulator to calculate allowed remuneration of electricity distribution companies, and it has been applied to research the impact on distribution networks of large-scale deployment of DG, active network management, and electric vehicle penetration (Cossent et al., 2011; Fernandez et al., 2011; Olmos et al., 2009). The RNM emulates the engineering design process of an electric distribution company by specifying the placement and layout of all major distribution network components connecting one or more primary transmission interconnection substations with all power injection or consumption points (i.e., loads and DG). The network is constructed to minimize total network costs (including capital expenditures, operational expenditures, and ohmic network losses) while meeting three specified quality of service constraints: (1) maximum system average interruption duration index (SAIDI); (2) maximum system average interruption frequency index (SAIFI); and (3) maximum acceptable voltage range at every node. The RNM is able to run in two modes: a "greenfield" mode and a "brownfield" or expansion-planning mode. The greenfield mode builds an efficient network from scratch using the location and maximum contracted power flow of each network user and each transmission substation in conjunction with simultaneity factors to size network components. The simultaneity factors for each network component specify what portion of total power flow downstream of the component contributes to peak power flow for that component and capture the heterogeneity of network users (i.e., not all load points peak at the same hour). See Domingo et al. (2011) for more on these factors. The placement of these simultaneity factors are depicted in Figure 2-7. 43 Jesse D. Jenkins (2014) Chapter 2: Constructing a Simulated Electricity Distribution Network Figure 2-7: Typical distribution network structure with simultaneity factors represented as a bold point (Source: Domingo et al., 2011) Type of faclities Network structure and simultaneity factors 610)kW HV MV LV (1,36)kV MV <1 kW LV W I Customer I k U-Stonler wtomin Transformcr + Simultaneity factor --->Network factor + Simultaneity <::=>Network Transformer J I The brownfield mode takes an established network layout as an input and determines additional network components and reinforcements necessary to accommodate changes in network uses. The brownfield mode also takes into account 48-hour power consumption and injection profiles for each network user (load or DG). To account for the diversity embedded in the thirty load profiles used in the simulation, the simultaneity factors used in the brownfield model runs are calculated by adjusting the factors used in the greenfield runs upwards to ensure consistency between the two modes. 10 Table 2-4: Voltage levels and simultaneity factors used in the RNM model Unit LV MV HV Voltage kV 0.24 12 33/66 Simultaneity Factor p.u. 0.30 0.80 1.0 Parameter In both modes, the RNM builds out a network with four voltage levels (LV, MV, and HV-1 and HV-2) in a balanced three-phase configuration. Network components are selected from a standard catalog file which contains technical and cost information about available equipment and the cost and time burden of maintenance and power restoration actions. The RNM also employs an algorithm that constrains the layout of network components to align with the corresponding street map for the area (see Domingo et al., 2011). This algorithm ensures a realistic layout of the network by avoiding placing overhead or used in the example, a simultaneity factor of 0.92 is already embedded in the mix of LV load profiles is adjusted run Denver simulation. Thus, the LV customer simultaneity factor of 0.3 used in the Greenfield brownfield the upwards by dividing by 0.92 to arrive at the simultaneity factor used for LV customers in runs: 0.3 / 0.92 = 0.33. 10 For 44 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) underground distribution lines through the middle of city blocks and minimizing the number of crossings of major avenues. Finally, the model takes into account the cost of capital, lifetime of assets, discount rate, and the cost of losses. To construct the simulated distribution network for Denver, this thesis first employs the RNM in greenfield mode to build the base network layout. Next, the model is run again in brownfield mode to ensure the network design accounts for the specific power profiles at each load point. The layout generated by the greenfield run is used as the input for the brownfield run which recalculates the optimal network design taking into account the 48-hour power profiles of network users. The resulting base network is depicted in Figure 2-8. The total network investment cost estimated by the RNM is $418.05 million (in overnight costs) as shown in Table 2-5. This is the estimated replacement value of the network and is used in Chapter 3 to calculate the regulated asset value (RAV) or "rate base" of the utility at the outset of the regulatory period. Table 2-5: Estimated efficient network costs for the simulated base network New Network Investment Network components New Quality Equipment Total New Network Investment Preventive Maintenance Corrective Maintenance Total Maintenance Annual costs (US$) Overnight costs (US$) LV feeders $50,096,168 $0 $50,096,168 $792,654 $620,343 $1,412,997 LV/MV transformers $22,706,658 $0 $22,706,658 $1,332,445 $60,829 $1,393,274 MV feeders $120,526,302 $18,746,320 $139,272,622 $703,912 $618,606 $1,322,517 MV/HV substations $147,384,000 $0 $147,384,000 $2,127,960 $589 $2128,549 $58,589,781 $0 $58,589,781 $211,176 $12,987 $224,163 $0 $0 $0 $0 $0 $0 $399,302,909 $18,746,320 $418,049,229 $5,168,147 $1,313,354 $6,481,501 HV lines Transmission substation Total 45 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Figure 2-8: The base simulated network for Denver, Colorado LV lines in black; MV lines in blue; HV lines in red. Insert: zoomed-in view of the portion of the network in vicinity of the primary transmission substation. 46 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) 2.5 - Simulating Network Expansion Scenarios Next, several network expansion scenarios are simulated for use in demonstrating the proposed regulatory process in Chapter 3. These scenarios capture a range of possible evolutions in network use, including variations in load growth and DER penetration. For both load and PV, three forecasts are generated: a central forecast, and both high and low sensitivity scenarios capturing the range of likely deviation from the central forecast (see Table 2-6). Projected load growth is specified as both "vertical" load growth - i.e., a percentage increase in demand at each load point during the regulatory period - and "horizontal" load growth - i.e., a number of new load points at each voltage level connected to the system during the regulatory period. Finally, to account for the impact of DER penetration on distribution networks, the forecast includes projected penetration of solar photovoltaic (PV) generators at each voltage level. Table 2-6: Forecasted evolution of network uses (load growth and PV penetration) Vertical load growth (% increase from base system) Approx. total load LV MV HV (million kWh/year)* Base system - - 2,100.7 Low forecast 3.5 3.5 3.5 2,212.5 Central forecast 4.0 4.0 4.0 2,232.5 High forecast 4.5 4.5 4.5 2,252.5 Horizontal load growth (# of new load points / kW peak) LV MV HV Approx. total peak demand (kW)* Base system - - 467,610 Low forecast 450 / 2,700 35 / 3,500 2 / 2,000 492,460 Central forecast 500 / 3,000 40 / 4,000 3 / 3,000 496,710 High forecast 550 / 3,300 45 / 4,500 4 / 4,000 500,980 * Note: Total values include combined impact of both horizontal and vertical load growth PV penetration (# of new PV connections / kW peak) LV MV - Base system - Approx. total peak generation (kW) - HV Low forecast 1,875 / 22,500 270 / 54,000 5 / 10,000 86,500 Central forecast 2,083 / 25,000 300 / 60,000 6 / 12,000 97,000 High forecast 2,292 / 27,500 330 / 66,000 7 / 14,000 107,500 47 Jesse D. Jenkins (2014) Chapter 2: Constructing a Simulated Electricity Distribution Network The RNM is employed at this stage to estimate the efficient network costs across the range of forecasted scenarios. There are nine cases covering all possible combinations of the low, central, and high forecasts for load growth and PV penetration. For each case, the RNM is run in brownfield mode, using the base network described in Section 2.4 as input. New load points are assigned to eligible connection points along streets using the same random sampling without replacement method used to construct the base network (see Section 2.3). New loads are assigned one of the thirty load profiles described in Section 2.2 with the same distribution of shares across customer types (industrial, commercial, and residential) as in the base network. Power demand at each hour is increased at all load points (existing and new) as per the appropriate vertical load growth in each case. PV generators are assigned one of six PV production profiles shown in Figure 2-9. These profiles are generated by the DOE's PVWatts solar PV production simulator (National Renewable Energy Laboratory, 2014) and correspond to TMY3 meteorological database characteristics for Denver, yielding realistic production profiles. From the annual hourly production data produced by PVWatts, two non-consecutive 24-hour periods are selected to match (1) the day of peak net power withdrawal (load minus PV production) and (2) the day of maximum net power injection given the penetration of PV generators in the network." These two days capture the extremes in power flow to which the distribution network must be designed, given the particular combination of load profiles and PV profiles used in this simulation. Six profiles are produced from the PVWatts calculator corresponding to five possible fixed alignments of rooftop-mounted panels as well as one single-axis tracking system. Individual generators are randomly assigned one of these six Figure 2-91: Solar PV production profiles Peak Power Injection Day Peak Power Witidrawal Day 1.20 1.00 Fixed Roof, South 0- Roof S 00-Fixed Fixed Roof SE Fixed Roof W -Fixed Roof E - *.W I-Axis Tracking ... Average .....- 0.20 0.00--1 2 3 4 5 6 7 8 9 101112131415161718192021222324 Hour 252627282930313233343536373839404142434445464748 Hour Given the DOE reference building loads and annual Pv production data Denver, peak net power demand occurs at 18:00 hours on July 2 6th and peak reverse power flow occurs at 13:00 hours on March 1 1 th 1 48 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) profiles with a probability of 0.5 for a south-facing roof-mounted system (making this the most common alignment) and 0.1 for each of the other five profiles. The location of PV generators connected to LV and MV feeders is determined by matching the generators with an existing load point using a placement algorithm which attempts to minimize the difference between the annual electricity generation of the PV system and the annual electricity consumption of the load point (both measured in kWh/year). This placement algorithm is meant to mimic customer placement decisions under the practice of "net metering," wherein customers with PV systems receive a credit on their electricity bill for each kWh of electricity generated by their PV system, generally with a limit set such that the total credits cannot exceed the customer's total electricity consumption. Large PV systems connected to the HV sub-transmission network are randomly placed at one of several pre-defined locations meant to designate likely connection points for such large systems. Note that the impact of PV systems on network costs is related to their generation profile, their location in the network, and the consumption profiles of load points in the vicinity of the generator (Vergara et al., 2014). Regulators should therefore take care to model realistic placement decisions in this stage of the regulatory process. They may wish to seek input from the utility and other stakeholders on the likely geographic pattern of the penetration of PV and other DERs during the regulatory period (as part of Steps 1-5, related to the forecast, discussed above). To ensure consistency across cases, the set of new load points and PV generators (and their locations and load profiles) used in the high forecast cases contains the full set of loads and PV generators in the central forecast cases, and the set in the central forecasts cases includes the full set of loads and generators in the low forecast cases. That is, for each load Ii c Liow c Lcentrai c Lhigh and 4/c Lcentral c Lhigh and for each PV generator g c Giow c Gcentrai c Ghigh and gj c Gcentrai c Ghigh. Figure 2-10 depicts the location of new network users (loads and PV generators) in the central forecast scenario as well as the network additions necessary to accommodate these new network users, as determined by the RNM. Note that additional reinforcements to existing network lines and protection equipment are also necessary, but are not depicted in Figure 2-10. 49 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Figure 2-10: Location of new network users and necessary network additions in central forecast Existing network and load points in black; new load points in dark blue; new PV generators in light blue; network additions to accommodate new network users in red. Insert: zoomed-in view of a portion of the network 50 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) Table 2-7 summarizes the efficient network expenditures estimated by the RNM. New network investments necessary to accommodate forecasted changes in network use (load growth and PV penetration) are divided into primary network investments and quality-related equipment (protection devices, voltage regulators, etc.) chosen by the RNM to optimize quality of service. This yields the total incremental network investment (expressed as overnight capital costs) required over the regulatory period. Annual operations and maintenance expenditures are also estimated by the RNM and divided into preventative maintenance and corrective maintenance costs. These values are nominal annual costs and include both maintenance of the existing network and new maintenance expenditures necessary to accommodate changes in network use over the regulatory period. Table 2-7: Estimated efficient network costs in central forecast scenario Total New New Network Investment Network components New Quality Equipment Network Investment Preventive Maintenance Overnight costs (US$) Corrective Total Maintenance Maintenance Annual costs (US$) LV feeders $1,625,755 $0 $1,625,755 $814,148 $637,103 $1,451,251 transformers $2,293,146 $0 $2,293,146 $1,467 023 $66,973 $1,533,996 MV feeders $1,178,007 $74,100 $1,252,107 $709,886 $623,117 $1,333,003 substations $0 $0 $0 $2,127,960 $589 $2,128,549 $7,391,355 $0 $7,391,355 $237,752 $14,621 $252,373 $0 $0 $0 $0 $0 $0 $12,488,262 $74,100 $12,562,362 $5,356,768 $1,342,403 $6,699,171 HV lines Transmission substation Total 51 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) References Colorado Department of Regulatory Agencies. (2014). About Xcel Energy's Seasonal Tiered Rates. Retrieved April 26, 2014, from http://cdn.colorado.gov/cs/Satellite/DORA-PUC/CBON/DORA/1251642963374 Cossent, R., Olmos, L., G6mez, T., Mateo, C., & Frias, P. (2011). Mitigating the Impact of Distributed Generation Advanced Response Options. International Transactions on Electrical Energy Systems, 21(6), 1869-1888. doi:10.1002/etep.503 Domingo, C. M., G6mez, T., S nchez-Miralles, A., Peco, J., & Martinez, A. C. (2011). A Reference Network Model for Large-Scale Street Map Generation. IEEE Transactions on Power Systems, 26(1), 190-197. Fernandez, L. P., G6mez, T., Cossent, R., Domingo, C. M., & Fri'as, P. (2011). Assessment of the Impact of Plug-in Electric Vehicles on Distribution Networks. IEEE Transactions on Power Systems, 26(1), 206-213. National Renewable Energy Laboratory. (2013). Commercial and Residential Hourly Load Profiles for all TMY3 Locations in the United States. Retrieved April 26, 2014, from http://en.openei.org/datasets/node/961 National Renewable Energy Laboratory. (2014). PVWatts Calculator (March 2014 Beta Release). Retrieved from http://pvwatts.nrel.gov/ Navigant Consulting. (2010). 2010 Colorado Utilities Report: A Report to the Colorado Governor's Energy Office. Olmos, L., Cossent, R., G6mez, T., Mateo, C., Joode, J. De, Scheepers, M., Gerhardt, N. (2009). Case studies of system costs of distribution areas (No. WP 4, Deliverable 5). Petten, the Netherlands. Retrieved from http://www.improgres.org/fileadmin/improgres/user/docs/D5_casestudiesof-sy stem costs.pdf U.S. Census Bureau. (2011). Table 1. Annual Resident Population Estimates, Estimated Components of Resident Population Change, and Rates of the Components of Resident Population Change for States and Counties: April 1, 2010 to July 1, 2011. 2011 Population Estimates. Retrieved April 26, 2014, from http://www.census.gov/popest/data/counties/totals/2011/files/CO-EST2011Alldata.csv U.S. Census Bureau. (2014). Denver, Colorado: People QuickFacts. State & County QuickFacts. 52 Chapter 2: Constructing a Simulated Electricity Distribution Network Jesse D. Jenkins (2014) U.S. Energy Information Administration. (2012). Colorado Electricity Profile 2010 - Table 3. Top Five Retailers of Electricity, with End Use Sectors. State Electricity Profiles. Vergara, C., Perez-Arriaga, I., Mateo, C., & Frias, P. (2014). Estimating the Aggregate Impact of Distributed Photovoltaic Generation over Distribution Networks (No. lIT14-034A). Madrid. 53 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) 3 - Demonstrating the Proposed Regulatory Process Using the simulated distribution network described in Chapter 2, this chapter demonstrates in detail the application of the novel regulatory process for the remuneration of electricity distribution utilities introduced in Section 1.3 and summarized in Figure 3-1. The proposed regulatory process involves ex ante calculation of allowed revenues and establishment of clear rules for annual ex post evaluations of actual expenditures and adjustments to final allowed revenues. This process provides strong incentives for the regulated utility to implement cost saving efficiencies, equalizes efficiency incentives across operational and capital expenditures (OPEX and CAPEX), and mitigates the impacts of uncertainty on both ratepayers and the regulated firm. In addition, the proposed methods equip the regulator with powerful tools to overcome their informational disadvantages vis-&-vis the regulated firm and minimizes incentives for the firm to engage in strategic behavior during the regulatory process. 3.1 - Forecasting the Evolution of Network Uses The ex ante regulatory process begins with the utility submitting to the regulator a detailed year-by-year forecast of the evolution of network uses over the upcoming regulatory period (Step 1).1 This forecast should at minimum include a set of appropriately justified scenarios covering a range of the likely load and DER penetration levels, including discussion of the most likely geographic evolution of loads and DERs. The regulator then critically reviews this forecast (Step 2). This review may also include a period of open comment on the preliminary forecast by stakeholders. At the conclusion of this review, the regulator will submit clear comments to the utility on required changes or further analysis needed to construct a final forecast (Step 3). Upon receiving this feedback, the utility will then perform any required updates to their scenarios (Step 4) and re-submit a final forecast to the regulator for use throughout the remainder of the regulatory process (Step 5). While in practice, this forecast would be created through several important, iterative steps involving the utility, regulator, and key stakeholders, this demonstration of this regulatory method begins by directly constructing a final forecast, meant to represent the end-product of Steps 1-5. 1 This general method is agnostic as to the length of the regulatory period, although this length should be carefully determined to balance the appropriate incentives for cost savings on the one hand and adjustments to mitigate uncertainty on the other hand. In general, the profit and risk sharing mechanism established by the proposed menu of contracts along with use of adjustment factors should reduce the impacts of uncertainty and enable a longer regulatory period than would otherwise be feasible. To demonstrate this process herein, a five-year regulatory period is selected. 54 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Figure 3-1: Proposed regulatory process for the remuneration of electricity distribution utilities Ex ante process (performed only at beginning of regulatory Regulator 1. Submission of utility's forecast for evolution of networK uses 2. Analysis of utility forecast / stakeholder comments 3. Regulator's comments on utility forecast 4. Revisions to utility forecast 5. Final forecast for evolution of network uses 6. Establishment of regulator's ex ante TOTEX estimate 7. Construction of incentive compatible menu of contracts 8. Submission of utility investment plan and estimated TOTEX 9. Calculation of ex ante TOTEX and revenue baselines and ex post sharing factor Utility IL TI 10. Calculation of automatic adjustment factors ("delta factors") to accommodate uncertainty in forecasted network use 11. Final regulatory contract published Ex post process (performed at the end of each year) Regulator I 1. Utility makes expenditures and reports to regulator 2. Audit of TOTEX and automatic adjustments to TOTEX baseline computed due to deviation from forecasted network use 3. Efficiency incentive computed as per menu of contracts and adjustments made to allowed revenues 4. Forward-looking revenue adjustments computed to true up revenues 5. Adjusted revenue allowance published 55 Utility I Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) The forecast specifies three scenarios for the evolution of network uses: a central forecast, and both high and low sensitivity scenarios capturing the range of likely deviation from the central forecast (see Table 3-1). Projected load growth is specified as both "vertical" load growth - i.e., a percentage increase in demand at each load point during the regulatory period - and "horizontal" load growth - i.e., a number of new load points at each voltage level connected to the system during the regulatory period. Finally, to account for the impact of DER penetration on distribution networks, the forecast includes projected penetration of solar photovoltaic (PV) generators at each voltage level. While a real forecast would involve expected changes in load profiles beyond a simple increase in average demand (vertical load growth) and would encompass a variety of new DER network users, this simplified forecast will suffice to demonstrate the basic application of the regulatory process. Table 3-1: Forecasted evolution of network uses (load growth and PV penetration) Vertical load growth (% increase from base system) Approx. total load (million kWh/year)* LV MV HV Base system - - - 2,100.7 Low forecast 3.5 3.5 3.5 2,212.5 Central forecast 4.0 4.0 4.0 2,232.5 High forecast 4.5 4.5 4.5 2,252.5 Horizontal load growth (# of new load points / kW peak) LV MV HV Approx. total peak demand (kW)* Base system - - - 467,610 Low forecast 450 / 2,700 35 / 3,500 Central forecast 500 / 3,000 40 / 4,000 2 / 2,000 3 / 3,000 492,460 496,710 High forecast 550 / 3,300 45 / 4,500 4 / 4,000 500,980 * Note: Total values include combined impact of both horizontal and vertical load growth PV penetration (# of new PV connections / kW peak) LV MV - HV - Approx. total peak generation (kW) - Low forecast Central forecast 1,875 / 22,500 270 / 54,000 5 / 10,000 86,500 2,083 / 25,000 300 / 60,000 6 / 12,000 97,000 High forecast 2,292 / 27,500 330 / 66,000 7 / 14,000 107,500 Base system 56 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) It is important to note that the utility may have an incentive to engage in strategic behavior during the construction of this forecast (Cossent & G6mez, 2013; Ofgem, 2010b). The utility may believe that inflating estimates of load growth or DER penetration will lead to an increase in ex ante allowed revenues. This is an important concern, but is mitigated through a variety of means. First, critical review of the preliminary forecast by the regulator and comment by stakeholders provides an opportunity to illuminate any strategic inflation in expected cost drivers. Second, use of automatic adjustment factors (see Section 3.4) to account for departures from the final forecast over the regulatory period for each key cost driver minimizes the incentive to engage in strategic behavior: if forecasted load growth is inflated and realized load is much lower, for example, the ex post automatic adjustment accounting for departures of load from the final forecast will reduce final allowed revenues accordingly. Note that constraining the utility's submission at this stage to only the forecast of network uses, as opposed to a full business plan including estimated expenditures (as in the process used by Ofgem, 2010 and 2013b or proposed by Cossent & G6mez, 2013) is intended to minimize opportunities for the firm to engage in strategic behavior overall. To fully ensure the theoretical incentive-compatibility of the menu of contracts designed in Step 7 (see Section 3.3), the regulator's estimate of efficient network costs should be set independently of the utility's cost estimate (see Ofgem, 2010, page 67). If the utility submits its cost estimate in advance of the regulator developing its own estimate, the firm may have an opportunity to influence the regulator. To avoid violating the firm participation constraint and ensure the financeability of the utility, the regulatory has an incentive not to produce a cost estimate too far below that of the utility. This may introduce an incentive for the utility to artificially increase its reported estimate of network costs that is not fully eliminated by the menu of contracts. The process proposed herein avoids this potential problem and preserves the full incentive-compatibility of the menu of contracts. The potential disadvantage of this approach is that the regulator at this stage lacks the detailed information contained in the utility's full business plan, which may be useful in parameterizing the reference network model used in Step 6. Indeed, Ofgem employs a menu of contracts precisely to increase the quality of information contained in these business plans (Ofgem, 2010b). In addition, this lack of information does increase the risk that the regulator may develop a cost estimate far off from that of the firm, raising the potential of leaving significant rents to the utility (if the estimate is too high) or risking the financeability of the firm (if the estimate is too low). However, this risk can be mitigated via the design of the menu of contracts itself: i.e., by setting a lower weight on the regulator's own cost estimate and/or a lower profit-sharing factor. 57 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Ultimately, an independent, confident regulator may decide, as Ofgem has, that the additional information in the full business plan can be obtained without unduly influencing the regulator's estimate of costs. Note however that without the use of an RNM, as proposed herein, the regulator has no practical option other than to accept a full business plan from the utility in advance of calculating their own TOTEX estimate. Ofgem, for example, develops their estimate of efficient costs by reviewing the firm's business plan (with the assistance of analytical models and consultants). Employing the RNM, in contrast, requires the comparatively more limited data inputs contained in the forecast of network use and frees the regulator to set an estimate of efficient expenditures independently of the firm's estimate. This could be a potential advantage over the method employed by Ofgem (2009, 2010b, 2013b). As such, this thesis begins with the forecast alone, demonstrating the practical implementation of this alternative process. 3.2 - Establishment of Regulator's Ex Ante Estimate of Efficient Expenditures Following construction of the forecasted evolution of network uses, the regulator constructs their ex ante estimate of efficient total network expenditures (TOTEX) over the regulatory period (Step 6). To construct this estimate, the regulator will employ a reference network model (RNM), a large-scale distribution network planning tool capable of producing an estimate of efficient expenditures to expand and maintain a network to serve a specified set of network users at prescribed quality levels (i.e., maximum statistical probability of network disruptions, voltage limits, etc.) and considering incentives for reduction of network losses. The RNM described in Domingo et al. (2010) will be used throughout this thesis to demonstrate application of this method, but any suitably rigorous RNM could be employed. Use of an RNM has several informational requirements in addition to the final forecast of network uses for the regulatory period. In order to estimate investment needs as accurately as possible, the network used as the starting point for the RNM should correspond to the actual assets operated by each utility in order to take into account the established layout of the network and sunk investments in network components (Cossent & G6mez, 2013). Regulators must therefore require utilities to report information on their existing networks in a standard format including: the location, voltage level, contracted capacity, and injection/withdrawal profile of all existing network connections (loads and DG); the layout, impedance, and capacity of the high-voltage (HV), medium-voltage (MV), and low-voltage (LV) electrical lines and protection devices; and the capacity and location of transmission interconnection substations, HV/MV substations and MV/LV transformers. Along with the final forecast for the regulatory period, these details on existing network layout are used as inputs for a "brownfield" or expansion-planning RNM that plans the necessary network expansion and maintenance expenditures needed to 58 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) serve forecasted network uses over the regulatory period (Domingo, G6mez, SanchezMiralles, Peco, & Martinez, 2011). Alternatively, the regulator can reduce the information reporting burden by running a "greenfield" RNM model which builds an efficient reference network from scratch using only the location and capacity of HV, MV, and LV network users (loads and DG) and the transmission interconnection substations (Domingo et al., 2011). The outputs of the greenfield RNM can then be used as inputs for the brownfield RNM used to model network expansion to serve forecasted network uses over the regulatory period. However, since the greenfield RNM builds a new simulated network from scratch, the resulting network may differ from the real utility network due to idiosyncratic path dependencies in the real network's prior development, reducing the accuracy of resulting estimate of efficient network costs (Cossent & G6mez, 2013).2 In either case, the regulator must also maintain a detailed library of standard network facilities and equipment for all voltage levels, including: cables, overhead lines, distribution transformers, substation components, and protection devices. These items are used by the model to plan necessary network investments and should adequately characterize the real investment alternatives the utility may face. As such, this library should be updated regularly to reflect the current cost of standard components and expanded to include any new network components recently entering common use, such as new "smart grid" related components (i.e. ICT equipment, advanced power electronics, etc.). To avoid opportunities for strategic behavior via inflation of reported component costs, the regulator should develop costs for library components by benchmarking efficient unit costs across multiple utilities (Cossent & G6mez, 2013; Cossent, 2013). Finally, the regulator should also specify several general economic and technical parameters, including the duration of the planning horizon, the weighted average cost of capital (WACC) for utility investments, the cost of energy losses, and appropriate quality metrics such as voltage limits and continuity of supply requirements. See Domingo et al. (2011) for a detailed discussion of input parameters for this particular RNM. Clearly these information requirements are significant. However, similar requirements have been successfully implemented in Spain, Chile, and Sweden, each of which employ RNMs for benchmarking purposes in the remuneration process (Cossent, 2013; Domingo et al., 2011; Jamasb & Pollitt, 2008). In Spain, utility companies are obligated by law to An intermediate solution is also possible, wherein utilities submit data only on the location and capacity of primary transmission substations, HV/MV substations, and MV/LV transformers along with the location and capacity of loads and DG. The greenfield RNM can then be run using these locations as constraints, forcing the layout of the network to more closely align to the utility's real network while reducing information reporting requirements (i.e. information on the layout and characteristics of the feeders and protection devices would not be necessary). 2 59 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) provide the regulator with necessary data, and similar requirements are likely necessary for this benchmarking method to be adopted. At the same time, given the increasing adoption of electronic equipment inventories and sophisticated geographic information systems by electric utilities, the reporting requirements necessary for the regulator to employ an RNM are likely to become an increasingly negligible hurdle over time. If these information requirements can be met, an RNM equips the regulator with a powerful tool to help overcome information asymmetries by providing a realistic engineering-based benchmark for efficient network expenditures. Unlike statistical benchmarking methods which rely on assessment of past expenditures, RNMs can be forward looking and thus better capable of capturing the evolving nature of distribution network uses. RNMs have already been applied to assess the impact on distribution costs due to large-scale deployment of DG, active network management, and electric vehicle penetration, for example (Cossent, et al., 2011; Fernandez et al., 2011; Olmos et al., 2009; Vergara, et al., 2014). Additionally, as a reference network is constructed for each utility, RNMs can capture the heterogeneity of utility networks, a particularly important feature as DER penetration is likely to increase the heterogeneity between distribution networks (see Section 1.2). Table 3-2: Characteristics of simulated base network Location Denver, Colorado Simulation area Population density Estimated population Estimated load power density - Base network Estimated peak power demand - Base network 120.3 1,580.5 190,187 3,890 468,079 km-sq persons/km-sq persons kW/km-sq kW Load Points - Base Network Industrial Commercial Residential Total LV 0 6,263 18,788 25,051 MV 212 1,274 637 2,123 HV 42 63 0 105 Total 254 7,600 19,425 27,279 For demonstration of this regulatory process, the realistic, large-scale urban distribution network simulated in Chapter 2 will be used as the base network. This network corresponds to a roughly 120 km-sq portion of Denver, Colorado, and is parameterized to closely simulate realistic conditions (see Table 3-2), including geographic distribution of network users using a real street map as a "scaffold" to constrain locations, use of thirty realistic load profiles (ten each for industrial, commercial, and residential users), and a specification of load power density and distribution of load points among user types (industrial, commercial, and residential) and voltage levels that closely matches the real 60 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) distribution of retail electricity sales in Denver. See Chapter 2 for details on simulation methods. Next, using this base network as input, the RNM is run in brownfield mode to compute the efficient network expenditures necessary to accommodate the central forecast for the evolution of network uses as developed in Steps 1-5 above (see Table 3-1). The location, size, and profiles of new load points and PV generators are specified using the simulation methods detailed in Section 2.5. Table 3-3 summarizes the efficient network expenditures estimated by the RNM for the central forecast. Table 3-3: Estimated efficient network costs in central forecast scenario Total New New Network Investment Network components LV feeders New Quality Equipment Network Investment Preventive Maintenance Overnight costs (US$) Corrective Maintenance Total Maintenance Annual costs (US$) $1,625,755 $0 $1,625,755 $814,148 $637,103 $1,451,251 transformers $2,293,146 $0 $2,293,146 $1,467 023 $66,973 $1,533,996 MV feeders $1,178,007 $74,100 $1,252,107 $709,886 $623,117 $1,333,003 $0 $0 $0 $2,127,960 $589 $2,128,549 $7,391,355 $0 $7,391,355 $237,752 $14,621 $252,373 $0 $0 $0 $0 $0 $0 $12,488,262 $74,100 $12,562,362 $5,356,768 $1,342,403 $6,699,171 LV/MV MV/HV substations HV lines Transmission substation Total New network investments necessary to accommodate forecasted changes in network use (load growth and PV penetration) are divided into primary network investments and quality-related equipment (protection devices, voltage regulators, etc.) chosen by the RNM to optimize quality of service. This yields the total incremental network investment required over the regulatory period, expressed as overnight capital costs. Note that these incremental network investments do not include any investments necessary to replace existing network assets. Replacement investments are calculated separately below. The RNM also estimates annual operations and maintenance expenditures, which are divided into preventative maintenance and corrective maintenance costs. These values include maintenance of the existing network and new maintenance expenditures necessary to accommodate changes in network use over the regulatory period. 61 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Note that the current implementation of the RNM used herein does not model active system management or other novel approaches to reduce network costs in the face of increased penetration of DER (see Cossent, et al., 2011; Eurelectric, 2013; Poudineh & Jamasb, 2014; Trebolle, et al., 2010). This is appropriate for regulatory purposes at this stage, as the regulator should establish economic incentives for these novel practices to become commonplace. However, as these techniques become part of the utility's normal repertoire, the RNM must be updated to ensure the efficient frontier estimated by the model aligns with industry best-practices. The efficient investment and annual maintenance expenditures estimated in Table 3-3 must then be converted to the regulator's ex ante estimate of total network expenditures (TOTEX) for the regulatory period. Table 3-4 depicts these calculations. First, the overnight cost of incremental investments is converted into an annual investment schedule by dividing the annual overnight investment cost computed by the RNM ($12.56 million, see Table 3-3) into even annual investments across the regulatory period.3 These overnight cost figures are then adjusted for inflation to current year dollars by applying the producer price index (PPI). In addition to incremental investments to accommodate changes in network uses, some portion of existing network assets reach the end of their useful life and must also be replaced each year. The gross value of retiring assets would be obtained in reality from audits of the firm's booked assets. This demonstration makes the simplifying assumption that the gross asset value of booked network assets at the start of the regulatory period corresponds to the replacement value of the base network simulated in Section 2.4 ($418.05 million), and that the gross asset value is divided evenly into 40 annual vintages of $10.45 million (assuming a 40 year average lifespan for network assets). The oldest vintage in the gross asset base will be fully depreciated each year, and assuming the financial life of assets corresponds to the physical useful life of these assets, this vintage will need to be replaced with new or refurbished assets, to maintain the functionality of the network. This conversion uses a stochastic heuristic Markov-chain particle swarm robust optimization Monte Carlos simulation method. For more, see: http://en.wikipedia.org/wiki/Easter ..egg_(media) 3 (Alternatively, one can divide the total sum by the number of years in the regulatory period...) 62 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Table 3-4: Regulator's ex ante estimate of efficient total network expenditures Assumes 5 year regulatory period and straight-line depreciation of RAV over 40 year average asset life; base network assets are evenly divided among vintages for computation of average age of network assets and gross asset value; pre-tax return on equity is 10%; cost of debt is 5.5%; gearing ratio is 35% equity, 65% debt and W.A.C.C. is 7.08%; inflation is 2.5%; discount rate is 6.5%; extended lifetime factor for replacement investments is 0.67.4 Year 0 CAPEX (million $) Incremental investment Year 1 Year 2 Year 3 Year 4 Year 5 Annual cost (current year dollars) NPV Total overnight cost (incremental network cost from RNM): Overnight cost $2.51 $2.51 $2.51 $2.51 $2.51 Inflation adjusted $2.64 $2.71 $2.77 $2.84 $2.58 Replacement investments Total Total gross asset value (base network cost from RNM): $12.56 $11.22 $418.05 Overnight cost $6.97 $6.97 $6.97 $6.97 $6.97 Inflation adjusted $7.14 $7.32 $7.50 $7.69 $7.88 $31.10 $9.72 $9.96 $10.21 $10.46 $10.73 $42.32 Total investment (CAPEX) OPEX (million $) Network maintenance Base value, no inflation Annual cost (current year dollars) $6.48 Inflation adjusted TOTEX ESTIMATE (million $) Total network expenditures (TOTEX) NPV $6.52 $6.57 $6.61 $6.66 $6.70 - $6.69 $6.85 $7.03 $7.20 $7.38 $29.13 $16.40 $16.81 $17.24 $17.67 $18.11 $71.40 59 59 59 59 59 59 41 41 41 41 41 41 FAST AND SLOW MONEY (% of TOTEX) CAPEX share of TOTEX: "Slow Money" OPEX share of TOTEX: "Fast Money" To accommodate network replacement costs, regulators commonly allow the utility an investment allowance equal to the full replacement value of the assets in the expiring vintage (i.e. $10.45 million per vintage in this case). However, this method is likely to overcompensate the utility. Replacing an existing network asset will almost certainly cost less than the original construction of that asset: trenches and rights of way for underground and overhead cabling have already been dug, permits obtained, connections to other assets installed, etc. In addition, existing assets can often be repaired and repurposed, extending their useful life at lower cost than purchasing a new replacement asset. Public Service Company of Colorado/Xcel's actual allowed pre-tax return on equity was set to 10% in 2012 (cross, 2012). Note that the implied after-tax cost of equity (6.5%) is very close to that calculated by Ofgem for RIO DPCR5 (6.7%) and the range used in ED-1 (6%-7.2%) (Ofgem, 2013, p. 15). Ofgem (2013, p. 12) reports a real cost of debt of about 3%. This is added to the inflation rate to arrive at cost of debt in nominal terms. Inflation rate is based on the producer price index for electric distribution utilities in the United States (U.S. Bureau of Labor Statistics, 2014). The average PPI was 3.5% from 2004-2013 but only 1.9% from 2009-2013. A forward-looking value of 2.5% is used here. The discount rate is aligned with the after-tax return on equity (6.5%). 4 63 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Therefore, regulators should hire an independent auditor to assess the average replacement cost as a percentage of the original asset costs. This percentage can be considered the "extended lifetime factor" and should be applied to the full value of the retiring asset vintage to obtain an estimate of efficient replacement costs that will avoid over-compensating the utility.5 This demonstration assumes an extended lifetime factor of 66.7%. Overnight replacement costs are then adjusted to current year costs by applying the PPI. Annual network maintenance costs (in non-inflation-adjusted terms) are estimated from the RNM. The maintenance costs at the start of the regulatory period equal the expected total maintenance costs for the base network simulated in Section 2.4, while the costs for the final year of the regulatory period correspond to the expected total maintenance costs for the expanded network that corresponds to the central forecast scenario (Table 3-3). Annual values for the interim years are imputed by assuming a compound annual growth in maintenance costs over the regulatory period. These values are then adjusted for inflation by applying the PP. The regulator's estimate of efficient TOTEX is thus the sum of inflation adjusted CAPEX (including incremental and replacement investments) and OPEX (network maintenance).6 In addition, the regulator can now determine the expected portion of TOTEX associated with both CAPEX and OPEX. These shares are important if the regulator wishes to employ a TOTEX-based approach to capitalizing expenditures into the regulated asset value (RAV) used to determine allowed revenues (see Section 3.4 below). If CAPEX and OPEX are treated separately by the regulator with CAPEX added to the RAV and OPEX expensed fully during the regulatory period, this can distort incentives for the utility and encourage over-investment (Ofgem, 2009). Both categories of incentives will face the same efficiency incentives (assuming the incentives are applied to TOTEX, as recommended herein), so a dollar of OPEX savings and a dollar in CAPEX savings will earn the utility the same efficiency-related income (as determined by the sharing factor set by the menu of contracts, see Section 3.3). However, if only CAPEX is capitalized into the RAV, then that dollar in reduced CAPEX will also involve a reduction in the RAV, and thus a reduction in the allowed return on equity and a corresponding decline in net profit for shareholders. This decline in net profit will offset some portion of the efficiency-related s This concept should be credited to Ignacio Perez-Arriaga. 6 Note that other business-related operational expenditures such as business support costs, pensions, etc. are not included in this simulation and are thus excluded from OPEX figures here. These expenditures would have to be accommodated in real revenue allowance determinations. 64 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) income, distorting tradeoffs between OPEX and CAPEX and potentially encouraging overinvestment. This distortion can be very problematic given the increased need for active system management approaches to accommodate DER at minimal cost (see Section 1.2). Active system management strategies will involve greater operational expenditures, including innovative contractual relationships with DER owners and aggregators, in an effort to defer or offset network investments (Cossent, G6mez, & Olmos, 2011; Cossent, et al., 2011; Fernandez et al., 2011; Perez-Arriaga, et al., 2013; Poudineh & Jamasb, 2014; Trebolle, et al., 2010). Addressing this distortion and equalizing incentives for efficiency across CAPEX and OPEX is thus paramount given the evolving nature of electricity distribution systems (Ofgem, 2009, 2013b). Recognizing this important distortion, Ofgem (2009) proposed a TOTEX-based approach to capitalizing expenditures into the RAV which equalizes the incentives for cost-savings across both CAPEX and OPEX. This approach was implemented in the Fifth Distribution Price Control Review period (DPCR5, 2010-2015) and will be continued during the first R1O price control period (RIIO ED-1, 2016-2024) (Ofgem, 2013b). Under this approach, all realized network expenditures are treated in the same way by designating a fixed portion of allowed TOTEX, referred to as "slow money," which is capitalized into the RAV (from which depreciation and cost of capital allowances are calculated) with the remainder of TOTEX designated as "fast money." The share of slow and fast money is fixed ex ante based on the regulator's expected share of CAPEX and OPEX, respectively, in total expenditures. The share of CAPEX and OPEX in actual utility expenditures during the regulatory period is thus free to depart from this expected share without impacting the utility's return on equity. Thus, cost-saving tradeoffs between both types of expenditure can be fully exploited by the utility, and the balance of incentives is restored (see Ofgem, 2009, pages 117-120, and Ofgem, 2013a, pages 30-32 for more). The regulatory burden is also reduced for both the utility and regulator, as neither party has to actively monitor the boundaries between various categories of expenditures. This TOTEX-based approach will be applied in this thesis, so the slow and fast money shares are computed and reported in Table 3.4. 3.3 - Construction of an Incentive Compatible Menu of Contracts Next, the regulator must construct an incentive compatible menu of profit-sharing regulatory contracts for the utility (Step 7). A menu of contracts specifies an ex ante regulatory allowance as well as a clear rule for ex post evaluation of actual expenditures and adjustments to final remuneration. The menu outlines a continuum of profit-sharing factors (sliding-scale efficiency incentives) wherein values depend on the ratio of the 65 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) utility's estimate of total network expenditures over the regulatory period to the regulator's TOTEX estimate derived in Step 6 (Section 3.2). If the design of the menu preserves incentive compatibility, it will provide incentives for the utility to provide a truthful estimate of expected network costs, helping overcome information asymmetries. In addition, incentives for the utility to engage in strategic behavior by inflating estimates of network costs are eliminated. Construction of the menu of contracts here follows the method set out in Cossent & G6mez (2013). Additionally, this general approach has been successfully implemented by Ofgem since the fourth distribution price control review (DPCR4, 2005-2010) and is now an integral part of Ofgem's RIO framework.7 Using the method introduced in Cossent & G6mez (2013), the regulator only needs to establish four discretionary regulatory parameters to create a continuous menu of contracts: 1. The weight placed on the regulator's estimate of efficient network expenditures relative to the utility's estimate, w. This weight should depend on how reliable the regulator believes their estimate of future expenditures is likely to be relative to the accuracy of the firm's estimate. A higher value places more weight on the regulator's estimate, while a lower value places more weight on the firm's estimate. 2. The reference value for the profit-sharing factor (the portion of cost savings/increases to which the utility is exposed, also known as the efficiency incentive rate), SFref, which corresponds to the case where the utility's estimate of future expenditures aligns with the regulator's estimate (Rex ante = 1.0). This value can be set to establish the strength of efficiency incentives faced by utilities in order to manage tradeoffs between incentives for efficiency and rent extraction taking into account the degree of uncertainty about future costs and demand. A value of 1.0 corresponds to a pure revenue cap contract while a value of 0.0 corresponds to a cost of service contract. In general, under lower levels of uncertainty, a higher profit-sharing factor (i.e., the firm is exposed to most of the risks and rewards of cost savings) performs better, while a lower profit-sharing factor (which shares most risks and rewards with ratepayers) performs better under higher levels of uncertainty (Schmalensee, 1989). Figure 3-2 summarizes 7 The UK's menu of contracts approach, known as the Information Quality Incentive (IQI), is described in Crouch (2006) as well in (Ofgem, 2009, 2010b, 2013c) and Cossent and G6mez (2013). See also Ofgem, (2013a) for a financial calculations spreadsheet implementing the IQI. 66 Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process several important considerations in setting the strength of the sharing factor (see Ofgem, 2010, pp 84-87 for further discussion). Figure 3-2: Regulatory considerations in establishing the sharing factor or incentive rate (Source: Ofgem, 2010) Company can operate with higher gearing reducing cost of capital consumers need to fund Stronger efficiency incentives Reduce risk that company spends Reduce risks that company faces financeability problerms, potentially putting outputs at: risk 7-nsumers enjoy more of the unexpected cost iangq achieved duning rne control rerind frnonrey unnieLcess!arily jut to grow RAVl bear smaller share of any additional expenlit-ire incurred by Consumer-s I Redceris roittatorpanesmak wat miyh beseenas indal pofis company durnng pnice :on,-1 penioc I Reduce risk of distortions in expenditureI iand cost Reduce risk of distortions in expenditure from one price control period to the next I ailocaition between regulated III d crpnes business and unrat same corporate group 3. The rate of change in the profit-sharing factor is the ratio between the utility's estimate and the regulator's estimate changes, SFroc. This value can be set so as to control the spread in efficiency incentives faced by different utilities during the regulatory period. A larger value results in a wider range of profit-sharing factors offered while a smaller factor results in a tighter range. Note that the regulator may wish to set this value to ensure that the sharing factor does not fall below a certain value (a lower bound of 0.3 for example). If the sharing factor is too low, a company may not face enough exposure to the costs of overspending and could face perverse incentives to increase their spending unnecessarily to increase their regulated asset value (RAV) and allowed revenues (for more, see Ofgem, 2010, pp 85-86). As a rule of thumb, the sharing factor should not fall below the riskadjusted rate of return the utility would be expected to earn by increasing their regulated asset value (i.e. their allowed returns on equity invested plus depreciation adjusted for relative risk), and it may be advisable to ensure the sharing factor stays a healthy margin above this rate. 4. The reference value for the additional income payment, Alref, used to ensure incentive compatibility of the menu of contracts. This reference value corresponds to the case where the utility's estimate of future costs aligns with the regulator's estimate (Rexante = 1.0), and the selected value can be used to tune expected profit margins for the utility. Table 3-5 describes how each of the remaining initialization parameters are calculated from these four discretionary parameters. The table also describes the formulas to compute the appropriate ex ante regulatory contract and ex post efficient incentive, the portion of realized over- or under-spend shared with the utility's shareholders. 67 Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process Table 3-5: Parameters and formulas for construction of incentive compatible menu of contracts Symbol Formula/constraint Description DISCRETIONARY INITIALIZATION PARAMETERS [0,1] W Weight on regulator's estimate [p.u.] SFref Reference value for sharing factor [p.u. share of over/under-spend retained by firm] SFroc Rate of change of sharing factor with ratio Alref Reference value for additional income [% of regulator's estimate] <0 CALCULATED INITIALIZATION PARAMETERS a 6 Intercept of additional income Alref - 100 * SFref *(w - 1) + 104 *SFroc *(w - 0.5) Alref - 100*a - 1002*6 1st order factor of additional income formula SFref *(W - 1) + 100* SFroc *(1 - 2*w) 2nd order factor of additional income formula SFroc *(w - 0.5) EX ANTE PARAMETERS Xfirm Firm's ex ante TOTEX estimate [$] Submitted by firm Xregulator Regulator's ex ante TOTEX estimate [$] Calculated by regulator using RNM Ratio of firm's estimate to regulator's estimate [%] Xfirm / Xreguiator baseline Ex ante allowed TOTEX [% of regulator's estimate] W *100+(1 - w)*Rexante SF Sharing factor [p.u. share of over/under- SFref + (Rexante - 100)* SFroc Al Additional income [% of regulator's estimate] AIint + a *Rex Xex ante spend retained by firm] ante - 6 *Rex ante2 EX POST PARAMETERS Submitted by firm and audited by regulator [$] Xex post Realized ex post TOTEX Rex post Ratio of realized ex post TOTEX to ex ante allowed TOTEX baseline [%] Xexpost/Xregulator Ex post efficiency incentive [share of overElex post (Xex under-spend retained by firm as % of onte - Xex post)*SF / Xregulator regulator's estimate] Aex post Final ex post adjustment to allowed revenues [% of ex ante regulator's estimate] Elex past + Al Tables 3-6 and 3-7 below present two examples of an incentive compatible menu of contracts. The menu in Table 3-6 offers fairly high-powered incentives (SFref = 0.7; SFroc = -0.01) and corresponds to a case where the regulator is fairly confident in their estimates of efficient costs (w = 0.66). Table 3-7 represents a lower-powered set of incentives (SFref = 0.4; SFroc = -0.005) and the regulator is less confident in their estimates of efficient costs (w = 0.33). Both of these menus will be used to demonstrate the ex post 68 = Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process determination of revenues in Section 3.4 below. Note that while these tables show discrete values in each column, a continuous menu of contracts can be calculated using the formulas in Table 3.5. Table 3-6: Example high powered menu of contracts w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0 Ratio of firm's cost estimate to regulator's cost estimate [%] Allowed revenues ex ante [% of regulator's cost estimate] Sharing factor [%] Additional income [% of regulator's cost estimate] Ratio of realized ex post expenditures to regulator's ex ante estimate [%] 85 90 Rex ante 90 95 100 105 110 115 120 X,,, a, 96.6 98.3 100.0 101.7 103.4 105.1 106.8 SF 80.0 75.0 70.0 65.0 60.0 55.0 50.0 Al 3.2 2.2 1.0 -0.2 -1.5 -2.9 -4.4 Rex,,,s Final ex post adjustment to allowed revenues [% of regulator's ex ante estimate] 8.1 9.5 10.6 11.5 12.1 12.5 5.4 6.5 7.4 8.0 8.4 95 4.5 100 105 110 115 120 125 0.5 -3.5 -7.5 -11.5 -15.5 -19.5 4.5 0.9 -2.9 -6.6 -10.4 -14.1 -17.9 Aexpost 6.5 4.0 4.1 3.5 2.6 1.5 0.9 0.5 -2.5 -0.1 -2.9 -5.6 -1.0 -3.5 -6.0 -8.5 -2.5 -6.0 -9.5 -13.0 -16.5 -5.6 -8.9 -12.1 -15.4 -8.5 -11.5 -14.5 -11.1 -13.9 -13.5 Table 3-7: Example low powered menu of contracts w = 0.33 SFref = 0.4; SFroc = -0.005; Alref = 1.0 Ratio of firm's cost estimate to regulator's cost estimate [%] Allowed revenues ex ante [% of regulator's cost estimate] Sharing factor [%] Additional income [% of regulator's cost estimate] Ratio of realized ex post expenditures to regulator's ex ante estimate [%] 85 90 95 100 105 110 115 120 125 Rexante 90 95 100 105 110 115 120 Xex ante 93.3 96.7 100.0 103.4 106.7 110.1 113.4 SF 45.0 42.5 40.0 37.5 35.0 32.5 30.0 Al 3.8 2.4 1.0 -0.3 -1.6 -2.8 -4.0 Re P, Final expost adjustment to allowed revenues [% of regulator's ex ante estimate] 5.3 6.0 6.6 7.0 7.3 7.5 3.7 4.3 4.7 5.0 5.2 2.1 2.5 2.8 3.0 3.0 0.4 0.8 0.9 0.9 0.8 -1.2 -1.0 -1.0 -1.2 -1.5 -2.8 -2.8 -3.0 -3.3 -3.8 -4.5 -4.7 -5.0 -5.4 -6.0 -6.1 -6.3 -6.6 -7.0 -7.6 -8.3 -7.7 -8.0 -8.4 -9.0 -9.7 -10.5 Aexpjst 4.5 3.0 1.5 0.0 -1.5 -3.0 -4.5 -7.5 Both example menus illustrate incentive compatibility. Shaded cells correspond to the cases for which the ex ante utility forecast matches actual expenditures. For any realized value of network expenditures (i.e., any horizontal row in the bottom half of the matrix), the utility will earn the greatest revenues in the case where their realized cost matches their ex ante forecast. Efficiency incentives are also preserved, as lowering realized 69 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) expenditures below the utility's forecast (i.e., moving up in a vertical column) will increase the utility's final revenues (and vice versa). Employing an incentive compatible menu of contracts can help regulators address both the heightened information asymmetry and uncertainty expected as distribution networks evolve to accommodate new DERs and employ new smart grid capabilities (see Section 1.2). First, the incentive compatible nature of the menu elicits accurate estimates of expected costs from the utility and removes incentives to inflate estimated costs (i.e., engage in strategic behavior). Second, the combination of ex ante revenue determination with clear ex post adjustment rules promotes efficient network investments while minimizing the regulatory uncertainty that can deter investment.8 Finally, the discretionary parameters used to construct the menu of contracts give the regulator flexibility to tune the strength of incentives to mitigate the impacts of uncertainty. For example, the weight placed on the regulator's estimate of network costs and the strength of the efficiency incentive (the sharing factor) can be reduced in the face of greater uncertainty or increased as the regulator becomes more confident in forecasts. 3.4 - Calculation of Ex Ante TOTEX and Revenue Baselines and Sharing Factor With the regulator's estimate of efficient network expenditures and menu of contracts on hand, the regulator can then assess the utility's estimate of network expenditures, which is submitted as part of their detailed business plan in Step 8. First, the utility's annual TOTEX estimates are compared to the regulator's estimates produced in Step 6 and the ex ante TOTEX baseline for each year in the regulatory period is established as per Equation 1: (1) Xex ante = Xregulator X (0 + Xfirm X (1 - w) Where: Xbaseline: annual total network expenditures baseline Xregulator: regulator's ex ante estimate of efficient total network expenditures Xfirm: firm's exante estimate of efficient total network expenditures o: weight placed on regulator's estimate Next, the ratio between the total net present value (NPV) of the utility's TOTEX estimate and the regulator's TOTEX estimate determines the sharing factor and additional income allowances as defined by the menu of contracts produced in Step 7. Regulatory certainty is further improved when this menu of contracts is combined with the automatic adjustment factors discussed below. 8 70 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Finally, the ex ante allowed revenue baseline for the regulatory period is calculated as per Equations 2-8: (2) Slow Money = Xex ante X Slow Money Share (3) Gross Assetsy = Gross Assetsy_ 1 - Expiring Assets + Slow Moneyy (4) RAV = Lif e * Gross Assets (5) Fast Money = Xex ante - Slow Money (6) Depreciation= (7) Gross Assets Lif e Cost of Capitaly = RAVY_ 1 x WACC (8) Revenue Baseline = Fast Money + Depreciation+ Cost of Capital + Additional Income Where: Slow Money: notional CAPEX allowance (capitalized into RAV) Slow Money Share: regulator's expected share of CAPEX in TOTEX Gross Assets: total gross value of in-service assets Expiring Assets: gross value of assets reaching end of useful life y current year in the regulatory period RAV: regulated asset value (gross value of assets less depreciation) Life: regulatory life of assets Age: average age of assets Fast Money: notional OPEX allowance (expensed annually) Depreciation:annual capital depreciation allowance Cost of Capital:annual allowance for repayment of debt and equity WACC: weighted average cost of capital Additional Income: additional income allowance from menu of contracts Revenue Baseline: exante allowed revenues for each year of the regulatory period Note that the gross asset value and average age of assets in the base network at the beginning of the regulatory period must be estimated for use in Equation 3 above. This value would be obtained in reality from audits of the firms booked assets. For this demonstration, a simplifying assumption is made: the total replacement value of the base network simulated in Section 2.4 ($418.05 million) is used as the initial gross asset value, while assets are divided equally and assigned to 40 annual vintages of $10.45 million. This yields an average age of 19.5 years for network assets at the beginning of the regulatory period. Each year, the oldest vintage of assets is fully depreciated, and thus removed from the gross asset value, while new investments (both incremental and replacement investments) made that year are added to the gross asset value as a new vintage. The average age of assets is tracked each year, and expressed as a dollar-weighted average of the age of each vintage in the gross asset base. 71 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Table 3-8 computes an example revenue allowance for a case in which the utility's estimate of TOTEX is higher than that of the regulator (ratio = 1.2) and using the higherpowered menu of contracts in Table 3-7. Table 3-8: Example of TOTEX and revenue baseline calculations w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0; Slow Money Share = 59%; WACC = 7.09%; Avg asset life = 40yrs 1 Year 2 Year 3 Year4 Year 5 NPV Regulator's estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 Utility's estimate $19.69 $20.18 $20.68 $21.20 $21.73 $85.73 $18.41 $18.87 $19.34 $76.30 Additional income -$3.14 Year0 Year TOTEX ESTIMATES TOTEX BASELINE AND MENU OF CONTRACTS PARAMETERS TOTEX baseline $17.52 Ratio 1.2 $17.96 Sharing Factor 50% REVENUE BASELINE CALCULATIONS Capitalization Slow money $10.38 $10.64 $10.90 $11.18 $11.46 $418.05 $417.98 $418.16 $418.61 $419.34 $420.34 19.50 19.50 19.49 19.47 19.44 19.40 $214.25 $214.18 $214.36 $214.81 $215.52 $216.50 Fast money allowance $7.14 $7.32 $7.50 $7.69 $7.88 $31.11 Depreciation allowance $10.45 $10.45 $10.45 $10.47 $10.48 $43.47 Cost of capital allowance $15.16 $15.15 $15.17 $15.20 $15.25 $63.09 Additional income -$0.72 -$0.74 -$0.76 -$0.78 -$0.80 -$3.14 $32.03 $32.18 $32.37 $32.58 $32.82 $134.5 Gross asset value Average age of assets (yrs) Regulated asset value Cost allowances Revenue baseline Together, the ex ante TOTEX and revenue baselines and the ex post sharing factor define the contract between the regulator and the utility for the duration of the regulatory period. This regulatory contract provides the utility with a clear expectation of how their revenues will evolve over the regulatory period and provides clear incentives for efficient management of network costs. Tables 3-9 through 3-11 illustrate the regulatory contracts associated with three different cases: one where the utility and regulator's ex ante cost estimates align (ratio = 1.0), one where the utility estimates a higher cost than the regulator (ratio = 1.2), and one where the utility estimates a lower cost than the regulator (ratio = 0.9). For each case, the regulatory contract describing the ex ante remuneration rules for the utility is described for two menus of contracts: one with a higher powered efficiency incentive (Table 3-6) and the other with a lower-powered incentive (Table 3-7). These cases illustrate a range 72 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) of possible regulatory contracts and demonstrate the flexibility the regulator has in specifying the parameters for constructing the menu of contracts. Table 3-9: Example of ex ante remuneration rules: utility and regulator estimates align Slow money share = 59%; WACC = 7.09%; Avg asset life = 40yrs. Year 1 Year 2 Year 3 Year 4 Year 5 NPV Regulator's cost estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 Utility's cost estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 $17.67 $18.11 $71.44 $32.85 $33.04 $135.90 $17.67 $18.11 $71.44 $32.85 $33.04 $135.90 Year 0 CASE 1: RATIO 1.0 Higher powered contract: w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0 $16.40 TOTEX baseline 70% Sharing factor $16.81 Additional income $0.71 $32.56 $32.69 $32.46 Revenue baseline $17.24 Lower powered contract: w = 0.33 SFref = 0.4; SFroc = -0.005; Alref = 1.0 $16.81 $16.40 TOTEX baseline 70% Sharing factor Revenue baseline $17.24 Additional income $0.71 $32.46 $32.69 $32.56 Table 3-10: Example of ex ante remuneration rules: utility forecasts higher cost than regulator Year1 Year2 Year3 Year4 Year5 NPV Regulator's cost estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 Utility's cost estimate $19.69 $20.18 $20.68 $21.20 $21.73 $85.73 $18.87 $19.34 $76.30 $32.58 $32.82 $134.52 $20.03 $20.53 $81.02 $33.31 $33.62 $137.20 Year0 CASE 2: RATIO 1.2 Higher powered contract: w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0 Sharing factor $17.96 $18.41 Additional income -$3.14 $32.18 $32.37 $17.52 TOTEX baseline 50% $32.03 Revenue baseline Lower powered contract: w = 0.33 SFref = 0.4; SFroc = -0.005; Alref = 1.0 Sharing factor Revenue baseline $19.07 $19.54 Additional income -$2.87 $32.53 $33.02 $18.60 TOTEX baseline 30% $32.76 73 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Table 3-11: Example of ex ante remuneration rules: utility forecasts lower cost than regulator Year 0 Year 1 Year 2 Year 3 Year 4 Year 5 NPV Regulator's cost estimate $16.40 $16.81 $17.24 $17.67 $18.11 $71.44 Utility's cost estimate $14.76 $15.13 $15.51 $15.90 $16.30 $64.30 $17.07 $17.49 $69.02 $32.90 $33.06 $136.3 $16.48 $16.89 $66.66 $32.67 $32.79 $135.44 CASE 1: RATIO 0.9 Higher powered contract: w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0 TOTEX baseline Sharing factor $15.85 80% Revenue baseline $16.24 $16.65 Additional income $2.30 $32.60 $32.77 $32.67 Lower powered contract: w = 0.33 SFref = 0.4; SFroc = -0.005; Alref = 1.0 TOTEX baseline Sharing factor $15.31 45% Revenue baseline $15.69 $16.08 Additional income $2.69 $32.47 $32.57 $32.51 3.5 - Calculation of Automatic Adjustment Factors to Manage Uncertainty Managing uncertainty is a fundamental regulatory challenge, particularly for ex ante approaches to establishing allowed revenues (see Section 1.2). While the use of an RNM and menu of contracts produces a clear revenue determination for each utility taking into account forecasted cost of capital, evolution of network uses, and network component costs, the ex ante nature of this regime means there will always be uncertainty regarding the accuracy of these forecasts. In particular, DER penetration rates are likely to be particularly uncertain and load growth may always be higher or lower than expected. Actual network expenditures may therefore be higher or lower than the ex ante TOTEX baseline (Section 3.4) due to deviations from forecast, irrespective of the utility's managerial efforts. If unmanaged, this uncertainty may either threaten the ability of the utility to recover costs, potentially violating the firm participation constraint, or it may leave large rents to the utility unrelated to the utility's exertion of managerial effort to reduce costs. The longer the regulatory period, the more substantial the effects of uncertainty can be on utility cost recovery or rent extraction (Ofgem, 2010b, 2013e). A range of uncertainty mechanisms can be employed to manage the impacts of uncertainty and account for divergences from forecasted evolution of network uses and costs of inputs. For a discussion of uncertainty mechanisms available for ex ante regulatory approaches, see Ofgem (2010a, 2013c). 9 Uncertainty mechanism options include: 9 This section draws heavily on those texts. 74 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) " Indexation provisions to adjust allowed expenditures based on changes in economy-wide price indexes (i.e., adjustment to account for economy-wide inflation of input costs or changes in average cost of debt). * Full or partial pass-through of the costs of some inputs that the regulator wants to ensure cost recovery for (i.e. smart meter roll-out costs of DG connection costs). * Revenue trigger provisions that specify a rule to increase or decrease revenues by a specified amount if certain trigger events occur (i.e. achievement of certain regulatory goals, substantial departures from forecasted network uses, etc.). * Re-opener thresholds specifying clear conditions under which expenditure and revenue determinations will be revisited and revised (i.e., to account for major new legislative requirements on utilities, changes in the tax code, substantial departures from forecasted network uses, or other major unanticipated cost drivers). * Mid-period reviews of output requirements half-way through the regulatory period. * Automatic adjustment factors or forward-looking volume drivers: pre-determined formula to adjust revenues as the volume of key cost drivers (e.g., the number of customer connections, load growth, DER penetration) depart from ex ante forecasts. The sharing factor established by the menu of contracts also works to manage uncertainty, as utilities and ratepayers share risks associated with divergences in realized costs from ex ante forecasted costs. For example, the higher/lower the sharing factor, the more/less financial exposure the utility faces to departures in costs from the forecast. Appropriate uncertainty mechanisms must be carefully selected and involve a set of important considerations (Ofgem, 2010b, 2013e). Uncertainty mechanisms can increase the complexity of the regulatory regime, and some mechanisms (i.e., re-openers and midperiod reviews) can introduce greater regulatory uncertainty, potentially undermining investment confidence or incentives for efficiency. Uncertainty mechanisms (particularly pass-throughs) can also undermine incentives for the utility to manage uncertainty and risk, reducing incentives for efficiency. The interaction of various uncertainty mechanisms must also be carefully understood, and opportunities for utilities to game mechanisms off one another should be minimized. In general, utilities should remain exposed to any uncertainties for which they are well equipped to manage. Use of uncertainty mechanisms should be limited and focused on 75 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) instances where they deliver clear value for money. Potential rationales for introduction of uncertainty mechanisms include cases where: * Uncertainty may raise concerns that the utility will not be able to finance necessary investments. Uncertainty mechanisms here can thus help reduce the risk that a re-opener of the ex ante expenditure and revenue determinations will be necessary to avoid violating the firm participation constraint. * Uncertainty mechanisms can reduce the cost of capital faced by the utility, thus reducing prices faced by network user. * Key cost drivers are fully outside the utility's control (i.e., inflation indexes). * Uncertainty mechanisms are necessary to ensure utilities do not face disincentives to serve network customers (i.e. automatic adjustment factors to account for costs of DER interconnection). This section focuses on the ex ante calculation of automatic adjustment factors, or "delta factors," simple formulas which will be applied expost to correct the estimate of efficient network expenditures (the TOTEX baseline) to account for any deviations from the forecast for both load growth and DG penetration (Step 9). These delta factors are intended to reduce incentives for the utility to slow interconnection of DERs and to account for variations in the evolution of network uses from loads, thus reducing the risk that the revenue determination will need to be re-opened and improving allocative efficiency. Similar mechanisms have been employed by the Comisidn Nacional de Energia (CNE) in Spain to account for deviations in load growth (known as the "Y factor," which is calculated using an RNM) (Cossent & G6mez, 2013) and by Ofgem in the UK to account for DG penetration during DPCR4 and DPCR5 (known as the "DG incentive;" the method used to calculate of this adjustment is not transparent) (Ofgem, 2010a). This thesis builds on these methods and provides a transparent methodology for calculation of delta factors for both load and DG penetration. First, the RNM is employed to estimate network investment and maintenance costs under the full range of possible load growth and DG penetration specified in Table 3-1. Nine uncertainty scenarios can be constructing corresponding to all possible permutations combining the three forecasts for load growth and three forecasts for DG penetration (the low, central, and high forecasts for each). The RNM is then run in brownfield mode to calculate the efficient network costs under each of these uncertainty scenarios. Table 3-12 illustrates the difference in investment and maintenance costs for the simulated Denver network, as estimated by the RNM, under each of the uncertainty scenarios as compared to the central forecast scenario, as well as the difference in load and PV penetration. 76 Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process Table 3-12: Difference in estimated efficient network costs, load, and PV penetration across uncertainty scenarios Load Total New Network Pv Investment penetration Total Network Maintenance Costs NPV Difference from Central Case Scenario (kWh) (kW) (US$ Overnight Cost) Efficient TOTEX (US$ Annual Cost) (M$) -19,821,979 -10,241 -$2,184,067 -$32,994 $69.47 Low Load, Central PV -19,821,979 0 -$1,708,074 -$27,222 $69.90 Low Load, High PV -19,821,979 10,963 -$1,339,456 -$21,491 $70.23 Central Load, Low PV 0 -10,241 -$520,576 -$6,670 $70.97 Central Load, Central PV 0 0 $0 $0 $71.44 Central Load, High PV 0 10,963 $327,716 $5,026 $71.74 High Load, Low PV 25,481,215 -10,241 $800,338 $13,583 $72.99 High Load, Central PV 25,481,215 0 $1,277,927 $19,926 $73.43 High Load, High PV 25,481,215 10,963 $1,661,622 $24,213 $73.78 Low Load, Low PV Next, an efficient level of TOTEX for each of this scenarios is calculated, using the same method employed to develop the regulator's central TOTEX estimate in Section 3-2. Finally, a two-factor linear regression is performed on the differences in estimated TOTEX, using the difference in load and PV penetration as the explanatory variables. Regression coefficients are obtained describing the change in TOTEX as a function of the divergence in load (in kWh) and PV (in kW) from the central forecast. These coefficients can be considered the automatic adjustment factors, or "delta factors," and prescribe a simple formula for correcting the TOTEX baseline ex post in light of the actual evolution of network uses over the regulatory period. Table 3-13 shows delta factors for the simulated Denver network, along with the R-square values for the regression. As the table illustrates, this regression is quite robust, with an R-square value of 0.998. Table 3-13: Regression coefficients and delta factors for load growth and PV penetration Regression Coefficients / Delta Factors (NPV Efficient TOTEX) Load growth $0.078/kWh PV penetration $36.20/kw R-square value 0.998 77 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) 3.6 - The Ex Ante Regulatory Process: Applying Annual Corrections At the conclusion of each year during the regulatory period, an ex ante regulatory commences to adjust the utility's allowed revenues in light of the realized evolution of system uses and utility expenditures. First, the utility submits a detailed report on actual investment and operational expenditures (the utility's realized TOTEX) as well as details on the evolution of system uses (i.e., load growth and penetration of DER) (Step 1). Next, the regulator will audit these reports to ensure their accuracy, and then compute the automatic adjustments to the ex ante TOTEX baseline to account for any differences in actual network use as compared to the ex ante forecast (Step 2). For each year, the regulator calculates the annual adjustment to the TOTEX baseline in total NPV terms as per Equation 9. This value, AdjustmentNPV, corresponds to the expected total increase/decrease in efficient network expenditures necessary to accommodate the increase/decrease in network use as compared to the TOTEX baseline and is calculated for each of the key network uses for which delta factors have been computed (i.e., in this demonstration, for both load growth and PV penetration). Since the utility would not be expected to make all of the expenditures to accommodate this deviation from forecast in the immediate year, this total NPV adjustment is then converted into a stream of annual expenditures spread across R years, where R is the number of years remaining in the regulatory period. Each annual value is inflated by the discount rate, as in Equation 6, to arrive at a stream of annual values equal in NPV terms to the total adjustment calculated in Equation 5. (9) AdjustmentNPV = Delta Factor x (Deviationy - Deviationy_) (10) Adjustmenti = (AdjustmentNPV x * (1 + Discount Rate) Where: AdjustmentNPV: total adjustment to TOTEX baseline (in NPV$) Delta factor delta factor for network use (in NPV$/kWh for load, NPV$/kW for PV) Deviationy: difference between realized and forecasted network use in year y (in kWh for load and kW for PV) Deviationy_1 : cumulative deviation from forecast through year y-1 y current year in the regulatory period Adjustmenti: annual adjustment to allowed revenues in year iwhere 1= y.y+(R-1) R: number of years remaining in the regulatory period DiscountRate: regulatory discount rate Table 3-14 provides an example of the annual adjustments to the TOTEX baseline for a deviation in load growth from the forecast, as applied over each year of the five year 78 Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process regulatory period. The same calculations would be employed for any deviations in the penetration of PV or other DERs as per the delta factors defined in the ex ante regulatory contract (see Section 3.5). The sum of all such annual adjustments will then be added to each year of the ex ante TOTEX baseline to arrive at the adjusted ex post TOTEX baseline. Table 3-14: Example of annual ex post adjustments to TOTEX baseline due to realized load growth Delta factor for load is $0.078/kWh; regulated discount rate is 6.5% 1 2 3 4 5 Forecasted 2,186.5 2,212.3 2,238.3 2,264.8 2,291.5 Realized 2,182.7 2,204.6 2,226.7 2,249.1 2,271.6 Deviation -3.80 -7.67 -11.64 -15.69 -19.82 -$0.32 -$0.35 -$0.38 -$0.41 -$0.07 -$0.07 -$0.08 -$0.08 -$0.09 -$0.09 -$0.10 -$0.10 -$0.12 -$0.13 -$0.14 -$0.20 -$0.22 Year NETWORK USE: LOAD (M kWh) ADJUSTMENT TO TOTEX BASELINE (M NPV $) -$0.30 AdjustmentNPV ADJUSTMENT TO TOTEX BASELINE (annual streams, in M $) -$0.06 Year 1 Year 2 Year 3 Year 4 -$0.44 Year 5 Cumulative Adjustment -$0.06 -$0.15 -$0.29 -$0.51 -$0.98 After calculating the adjusted TOTEX baseline, the regulator then compares the utility's realized TOTEX over the last year with the adjusted TOTEX baseline, and the efficiency incentive is calculated (Step 3) as per Equation 11. This is the portion of the over/underspend shared by the utility's shareholders. The ex post allowed TOTEX is thus the utility's realized TOTEX less this efficiency incentive, as in Equation 12. (11) Efficiency Incentive = Sharing Factor x (TOTEXbaseline - TOTEXrealized) (12) TOTEXauowed= TOTEXrealized - Efficieny Incentive Where: Efficiency Incentive: portion of over/under-spend shared by utility shareholders SharingFactor the sharing factor specified in the ex ante regulatory contract TOTEXbaseline: adjusted TOTEX baseline TOTEXrealized: total network expenditures realized by the utility This efficiency incentive is the share of any cost savings or cost overruns to which the utility should be exposed, as specified by the sharing factor in the ex ante regulatory contract. If costs are reduced below the adjusted TOTEX baseline, the efficiency incentive 79 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) will be a negative value, indicating that the utility should be allowed to recover more than their realized costs, thus sharing some of the cost savings with the utility's shareholders. If realized costs are higher than the adjusted TOTEX baseline, than the utility has overspent and the efficiency inventive will be a positive value. In this case, the utility's allowed cost recovery should be less than the realized costs, ensuring the utility's shareholders pay for a portion of the cost overruns. After computing the allowed ex post TOTEX, the regulator can calculate the ex post revenue allowance associated with this allowed TOTEX. These calculations employ the same method discussed in Section 3.4 (Equations 2-8). Note that regardless of actual capital expenditures, the portion of allowed TOTEX capitalized into the RAV is determined by the slow money share set ex ante, maintaining balanced incentives for cost-saving efficiency efforts across both CAPEX and OPEX (as discussed in Section 3.2). Since revenues have already been collected over the course of the recently concluded year, the regulator must adjust the utility's revenue allowance in future years to "true up" the collected revenues and the ex post revenue allowance computed above (Step 4). The deficit or surplus in collected revenues is calculated as the difference between allowed cost recovery and collected revenues. Next, an annual stream of adjustments to true up future allowed revenues is calculated as per Equation 13, which ensures that the NPV of adjustments to future revenues corrects for the surplus or deficit in collected revenues over the recently concluded year. This true up is applied as a stream of annual adjustments, rather than a single lump sum correction, so as to smooth the impact on rates and avoid discontinuous rate increase/decreases. The corrected revenue allowance is thus the allowed revenues less the cumulative true up adjustments. (13) True UPK = Allowed Revenues-Collected Revenues x (1+ Discount Rate)K-y Where: True UpK: annual adjustment to allowed revenues in year Kwhere K= y+1:y+N Allowed Revenues: expostrevenue allowance for y Collected Revenues: revenues collected in yeary y current year in the regulatory period N: the number of years in the regulatory period As allowed revenues are adjusted for the next N years, where N is the length of the regulatory period, a portion of the true up corrections will be applied during the next regulatory period. The regulator must therefore track these adjustments and add them to the revenue baseline calculated in the next regulatory period. This N year rolling window of true up corrections also ensures that the utility's incentives for cost savings are equalized across each year in the regulatory period, as no matter what year these 80 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) savings are achieved, the utility will be entitled to collect the agreed share of those savings over the next N years. Table 3-17 (next page) demonstrates the full application of the expost annual adjustment process and formulas described above. In this hypothetical example, the utility initially estimates a higher efficient cost than the regulator (ratio 1.1), but ultimately achieves a reduction in costs bringing TOTEX below the adjusted TOTEX baseline (ratio 0.9). Load ends up growing slower than forecasted, while PV penetration grows more rapidly. This case therefore demonstrates the computation of adjustments to the TOTEX baseline due to realized network use as well as the adjustments to the revenue allowance to account for efficiency incentives and the additional income allowance specified by the ex ante regulatory contract. Table 3-16 shows the financial position of the utility under this example. As illustrated, because the utility was able to achieve significant cost savings, the utility's shareholders earn a final after tax return on equity of 7.5% for the regulatory period, above the target return on equity of 6.5%. Table 3-16: Final financial position of the utility Cost of debt = 5.5%; gearing ratio = 35% equity / 65% debt; tax rate Allowed revenues $132.8 Allowed costs $103.0 Fast money allowance $28.1 Depreciation allowance $43.3 Cost of debt $31.6 Efficiency incentive income $5.7 Earnings before interest & taxes $35.6 Taxes $12.4 Net profit $23.1 After-tax return on equity 7.5% 81 = 35% pter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) e 3-17: Example of ex post annual corrections to allowed revenues nte ratio = 1.1; Sharing factor = 60%; Additional income = -1.54%; Slow money share = 59%; WACC = 7.08; ar NPV 1 2 3 4 5 -3.80 -7.67 -11.64 -15.69 -19.82 +2,193 +4,385 +6,578 +8,770 +10,963 6 7 8 9 VIATION IN NETWORK USE FROM FORECAST ad deviation (M kWh) deviation (kW) TEX (M $) gulator's estimate $71.4 $16.4 $16.8 $17.2 $17.7 $18.1 ility's estimate $78.6 $18.0 $18.5 $19.0 $19.4 $19.9 alized cost $63.2 $14.51 $14.88 $15.25 $15.63 $16.02 ANTE TOTEX AND REVENUE BASELINES (M $) TEX baseline venue baseline $73.9 $17.0 $17.4 $17.8 $18.3 $18.7 $135.3 $32.3 $32.4 $32.6 $32.7 $33.0 MULATIVE ADJUSTMENT TO TOTEX BASELINE DUE TO DEVIATIONS FROM FORECASTED NETWORK USE (M $) justment: load -$1.54 -$0.06 -$0.15 -$0.29 -$0.51 -$0.98 justment: PV +$0.40 $0.02 $0.04 $0.08 $0.13 $0.25 justed TOTEX baseline $72.7 $16.9 $17.3 $17.6 $17.9 $18.0 LCULATION OF EFFICIENCY INCENTIVE AND EX POST ALLOWED TOTEX al over/under-spend -$9.52 -$2.40 -$2.40 -$2.36 -$2.26 -$1.97 iciency incentive -$5.71 -$1.44 -$1.44 -$1.42 -$1.36 -$1.18 post allowed TOTEX $68.9 $16.0 $16.3 $16.7 $17.0 $17.2 Adjustments applied in next regulatory period RRECTION OF REVENUE ALLOWANCE (M $) post allowed revenues $132.8 $31.9 $31.9 $32.0 $32.0 $32.1 venue correction -$1.57 $0.00 -$0.09 -$0.18 -$0.28 -$0.39 -$0.51 -$0.43 -$0.34 -$0.24 -$0 al revenue allowance $132.8 $32.27 $32.22 $32.20 $32.19 $32.18 -$0.51 -$0.43 -$0.34 -$0.24 -$ 82 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) 3.7 - Evaluating the Performance of the Regulatory Process This section evaluates the performance of the regulatory process proposed herein, demonstrating the incentive compatibility of the overall process, the ability of the process to accommodate uncertainty in the evolution of network uses, and the performance of the framework if the regulator errs in establishing their ex ante TOTEX baseline. Tables 3-18 and 3-19 demonstrate the incentive-compatibility of this regulatory method across a range of scenarios for the simulated Denver system. For any ex post realized TOTEX, the utility earns the greatest net profit, expressed as earnings before interest and taxes (EBIT) and after tax return on equity when realized expenditures align with the ex ante estimate submitted by the firm at the beginning of the regulatory period (see the highlighted cells in Tables 3-18 and 3-19). As such, this approach removes any incentive for the firm to engage in strategic behavior by inflating their ex ante estimate of network expenditures in an effort to increase their profits; such an effort would, in fact, reduce the firm's net profit and return on equity under this regulatory approach. This method thus encourages the utility to submit the highest-quality, most accurate business plan they can develop. Along with the use of an RNM to develop the regulator's estimate, the incentive-compatibility of this regulatory approach substantially reduces the information asymmetry between regulator and firm. Table 3-18: Incentive compatibility with higher-powered menu of contracts EX ANTE TOTEX ESTIMATES Regulator Estimate NPV M$ $71.4 $71.4 $71.4 $71.4 Firm's Estimate NPV M$ $64.3 $71.4 $78.6 $85.7 % 90 100 110 120 Ratio (Firm/Regulator) EX ANTE REGULATORY CONTRACT (w = 0.66; SFref = 0.7 SFroc= -0.01; Alref = 1.0) TOTEX baseline NPV M$ $69.0 $71.4 $73.9 $76.3 Revenue baseline NPV M$ $136.3 $135.9 $135.3 $134.5 % 80 70 60 50 NPV M$ $2.3 $0.7 -$1.1 -$3.1 Sharing Factor Additional Income EX POST RESULTS Realized TOTEX (NPV M$) TOTEX Ratio (Realized/ Regulator) $64.3 90 $24.04 / 7.78 $23.83/ 7.70 $23.16 / 7.47 $22.00 / 7.10 $71.4 100 $20.36 / 6.58 $20.63 / 6.65 $20.44 / 6.57 $19.76 / 6.35 $78.6 110 $16.67 / 5.38 $17.44 / 5.60 $17.72 / 5.68 $17.53 / 5.61 $85.7 120 $12.99 / 4.18 $14.24 / 4.56 $15.00 / 4.79 $15.29 / 4.87 Earnings Before interest and Taxes (NPV M$) / Return on Equity (%) 83 Jesse D. Jenkins (2014) Chapter 3: Demonstrating the Proposed Regulatory Process Table 3-19: Incentive compatibility with lower-powered menu of contracts EX ANTE TOTEX ESTIMATES Regulator Estimate NPV M$ $71.4 $71.4 $71.4 $71.4 Firm's Estimate NPV M$ $64.3 $71.4 $78.6 $85.7 % 90 100 110 120 Ratio (Firm/Regulator) EX ANTE REGULATORY CONTRACT (w = 0.66; SFref = 0.7 SFroc = -0.01; Alref = 1.0) TOTEX baseline NPV M$ $66.7 $71.4 $76.2 $81.0 Revenue baseline NPV M$ $135.4 $135.9 $136.5 $137.2 % 45 40 35 30 NPV M$ $2.7 $0.7 -$1.1 -$2.9 Sharing Factor Additional Income EX POST RESULTS Realized TOTEX (NPV M$) TOTEX Ratio (Realized/ Regulator) $64.3 90 $22.46 / 7.29 $22.39 / 7.25 $22.07 / 7.14 $21.51 / 6.95 $71.4 100 $20.47/ 6.61 $20.63 / 6.65 $20.56 / 6.61 $20.24 / 6.50 $78.6 110 $18.47 / 5.94 $18.88 / 6.05 $19.05 / 6.09 $18.97 / 6.06 $85.7 120 $16.48 / 5.27 $17.13 / 5.46 $17.58 / 5.58 $17.69 / 5.62 Earnings Before Interest and Taxes (NPV M$) / Return on Equity (%) In addition, this regulatory method provides the firm with clear incentives to reduce costs to the most efficient level possible (subject to quality of service requirements or incentives): in either table, moving upwards in any column always increases the firm's profits and return on equity. As can be seen by comparing Tables 3-18 and 3-19, the strength of the efficiency incentives are determined by the construction of the menu of contracts, and the regulator thus has substantial flexibility in determining the appropriate level of incentives (see Section 3.3). Because this method adopts the TOTEX-based approach to capitalizing expenditures pioneered by Ofgem (i.e., "fast money" and "slow money" shares are fixed ex ante, see Section 3.2 and Ofgem, 2009, pages 117-120, and Ofgem, 2013a, pages 30-32), incentives for cost-savings across CAPEX and OPEX are equalized. Table 3-20 demonstrates the net profit (earnings before interest and taxes) of the utility under two cases, both of which reduce total net present value of network expenditures by 5 percent: a case where savings are achieved by reducing CAPEX and a case where equal savings are achieved by reducing OPEX. As the table demonstrates, if the utility's assets are capitalized based on weaker the realized CAPEX (the "CAPEX method"), incentives to reduce CAPEX costs are than for OPEX. In this case, the firm would earn $150,000 more over the regulatory period by reducing OPEX instead of CAPEX, distorting tradeoffs between these two expenditure 84 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) categories. If instead, the TOTEX-based method is applied, in which the share of realized expenditures capitalized into the utility's regulated asset value is fixed ex ante by the regulator (the "slow money share"), cost-saving incentives are balanced and the utility's profits are equalized under both savings cases. Table 3-20: Example of balanced and distorted incentives for CAPEX and OPEX cost-savings Sharing factor: 70%; Additional income: 1%; Gearing ratio: 35% equity / 65% debt; WACC: 7.08% Value Unit No savings case CAPEX savings case OPEX savings case CAPEX NPV M$ $42.32 $38.75 $42.32 OPEX NPV M$ $29.13 $29.13 $25.55 TOTEX NPV M$ $71.44 $67.87 $67.87 Ex ante slow money share % 59% 59% 59% Ex post CAPEX share % 59% 57% 62% Distortion EARNINGS BEFORE INTEREST AND TAXES TOTEX method NPV M$ $20.63 $22.23 $22.23 None CAPEX method NPV M$ $20.63 $22.17 $22.32 $0.15 In addition to providing clear incentives for the firm to provide an accurate ex ante business plan and to achieve cost-saving efficiencies throughout the regulatory period, the use of the delta factors described in Section 3.6 also effectively manages uncertainty in the evolution of network use. Table 3-21 illustrates the performance of these delta factors for the simulated Denver system. Across a set of uncertainty scenarios encompassing a deviation of plus or minus 1 percent in realized load and 11 percent in PV penetration, the delta factors correct the TOTEX baseline to within 1.2 percent of the realized network expenditures (as estimated by the RNM). In most cases, the deviation between realized costs and the adjusted TOTEX baseline less than 0.1 percent. As Table 3-20 illustrates, the error appears to increase non-linearly. This may indicate that a non-linear functional form for the delta factors may improve performance. Further research should explore the performance of these delta factors across a wider range of variation in network use and explore alternative functional forms. However, even the simple linear delta factors employed herein substantially reduce the impacts of uncertainty in network use, maintaining efficiency incentives and ensuring the firm can recover reasonably incurred expenditures throughout the regulatory period. This is a considerable advantage for an ex ante regulator approach. 85 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Table 3-21: Performance of delta factors (automatic adjustment factors) NETWORK USE SCENARIOS Load Low Low Low Central Central Central High High High PV Low Central High Low Central High Low Central High NETWORK USE REALIZED VALUES (Load: million kWh; PV: kW-peak) Load 2,272 2,272 2,272 2,292 2,292 2,292 2,317 2,317 2,317 PV 85,100 95,341 106,304 85,100 95,341 106,304 85,100 95,341 106,304 PERCENT DEVIATION IN REALIZED NETWORK USE VS CENTRAL FORECAST Load -1 -1 -1 0 0 0 +1 +1 +1 PV -11 0 +11 -11 0 +11 -11 0 +11 TOTEX (NPV million $) Realized $69.47 $69.90 $70.23 $70.97 $71.44 $71.74 $72.17 $72.60 $72.95 Adjusted $69.53 $69.90 $70.30 $71.07 $71.44 $71.84 $73.06 $73.43 $73.82 +0.1 +0.0 +0.1 +0.1 0.0 +0.1 +1.2 +1.1 +1.2 Baseline Error (%) While the delta factors successfully address forecast error, this method is still susceptible to benchmark error: that is, if the regulator errs in establishing their ex ante estimate of the efficient frontier for network costs due to a defect in the RNM or another error on the regulator's part, the TOTEX baseline may be set too high or too low, distorting efficiency incentives. Ideally, the combination of the regulator's estimate and the firm's estimate of efficient TOTEX will blunt the impacts of benchmark error. The incentive compatibility of the menu of contracts will encourage the utility to submit their most accurate TOTEX estimate. As this estimate is averaged with the regulator's estimate as per the weighting factor, w, the firm's estimate can counteract a portion of any benchmarking error on the regulator's part when setting the ex ante TOTEX baseline. However, the additional income computed to ensure the incentive compatibility of the menu of contracts also depends on the weighting factor placed on the regulator's estimate as per the equations in Table 3-5. This counteracts the impact of incorporating the firm's estimate into the TOTEX baseline on the error in final allowed profits and return on equity, as illustrated in Table 3-22. Two scenarios are depicted in this table: one in which the regulator under-estimates efficient costs by 20 percentage points and the other where the regulator over-estimates efficient costs by an equivalent percentage. In both cases, the utility's ex ante forecast is treated as accurate. Three different weighting factors are also shown, with the error in profit and after tax return on equity computed. As Table 3-22 demonstrates, the error is barely impacted by the choice of weight placed on the regulator's TOTEX estimate. Counterintuitively then, the weighting factor cannot help mitigate the impacts of benchmark error 86 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) using this method. Future research may wish to identify a method for constructing an incentive compatible menu of contracts which does not have this feature. Table 3-22: Performance of regulatory method under benchmark error with different weighting factors SFref = 0.7 SFroc = -0.01; Alref = 1.0 +20 percent -20 percent Regulator Error 1.0 0.66 0.33 1.0 0.66 0.33 Allowed Profit / Return on equity $15.23 / 4.9 $15.29 / 4.9 $15.35 / 4.9 $28.03 / 9.0 $27.93 / 9.1 $27.82 / 9.1 Accurate Profit / Return on equity $20.98 / 6.6 $20.98 / 6.6 $20.98 / 6.6 $20.29 / 6.7 $20.29 / 6.7 $20.29 / 6.7 Error in profit / Return on equity -$5.74 / -1.8 -$5.69 / -1.8 -$5.63 / -1.8 +7.74 / +2.3 +$7.64 / +2.4 +7.53 / +2.4 Weighting factor (o) While the weighting factor cannot help mitigate the impacts of benchmark error, the regulator can select the reference sharing factor parameter (SFref) to reduce the impacts of any errors in the regulator's forecast. As Table 3-23 demonstrates, the error in allowed profits and return on equity declines as the sharing factor declines (as anticipated by Schmalensee (1989)). This result is expected, as the lower the sharing factor, the closer the regulatory contract becomes to a cost-of-service contract, and thus the less sensitive the firm's profits are to realized costs. Table 3-23: Performance of regulatory method under benchmark error with different sharing factors w = 0.66; SFroc = -0.01; Alref = 1.0 +20 percent -20 percent Regulator Error 1.0 0.7 0.4 1.0 0.7 0.4 Allowed Profit / Return on equity $12.44 / 4.0 $15.29 / 4.9 $18.15 / 5.8 $30.78 / 10.0 $27.93 / 9.1 $25.07 / 8.2 Accurate Profit / Return on equity $20.98 / 6.6 $20.98 / 6.6 $20.98 / 6.6 $20.29 / 6.7 $20.29 / 6.7 $20.29 / 6.7 Error in profit / Return on equity -$8.54 / -2.7 -$-5.69 / -1.8 -$2.83 / -0.9 +10.49/ +3.3 +$7.64 / +2.4 +4.78 / +1.5 Sharing factor (SFref) 87 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) 3.8 - Summary of Advantages of the Proposed Regulatory Process This chapter demonstrates, step-by-step, the practical implementation of a novel regulatory method combining several state-of-the-art regulatory methods, including: use of an engineering-based reference network model (RNM) for forward-looking benchmarking of efficient network expenditures; an incentive compatible menu of contracts to elicit accurate forecasts from the utility and create incentives for cost saving efficiency efforts; and ex post automatic adjustment mechanisms, or "delta factors," to accommodate uncertainty in the evolution of network use and minimize forecast error. This approach yields several important advantages for the economic regulation of electric distribution utilities under increasing penetrations of distributed energy resources, which are summarized below. Reduces information asymmetry and minimizes opportunities strategic behavior " The regulatory regime proposed herein helps overcome information asymmetry by equipping the regulator with a reference network model with which to develop their estimate of efficient network expenditures. The RNM emulates the network planning practices of an efficient utility and can help the regulator develop more accurate estimates of future network costs given the expected evolution of network uses. * Combining the use of an RNM with an incentive compatible menu of contracts further reduces information asymmetry by incentivizing the utility to submit their most accurate estimate of future network expenditures. The incentive compatible property of the menu of contracts thus eliminates incentives for the utility to engage in strategic behavior by inflating their estimate of necessary TOTEX, a significant advantage over other ex ante regulatory approaches that do not employ a menu of contracts. Helps manage systemic uncertainty * Because the RNM creates a forward-looking benchmark for efficient total network expenditures (TOTEX), the model can be designed to accommodate expected evolutions in network use, technology costs, and network management practices, reducing the uncertainty facing the regulator. In effect, the RNM gives the regulator a tool with which to "peer into the future," a crucial ability in ex ante regulatory approaches. This forward-looking capability stands in contrast to statistical benchmarking techniques, which rely on backward-looking analysis of realized expenditures during prior regulatory periods and thus cannot capture the dynamic changes now unfolding in the electricity distribution sector. 88 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) " The RNM can also be used to explore a range of possible scenarios for the evolution of network uses (i.e., load growth and DER penetration). The model results can then be used to compute delta factors, simple formulas to automatically adjust the efficient TOTEX baseline in light of the realized evolution of network use. These delta factors effectively minimize the impacts of forecast errors, a significant advantage given increased uncertainty about the likely evolution of network use over the coming years. " By selecting the strength of the profit sharing factor, the regulator can also help mitigate the impacts of benchmark error - i.e., an error in the regulator's estimate of efficient TOTEX (irrespective of the evolution of network use). The lower the sharing factor, the closer the regulatory contract becomes to a cost-of-service contract, and thus the less sensitive the firm's profits are to differences in forecasted and realized costs, and vice versa. The regulator can thus select an appropriate sharing factor based on their confidence in the accuracy of their forecasts of efficient network expenditures. Incentives cost savings and overcomes moral hazard * The profit sharing parameter established by the menu of contracts creates clear incentives for the utility to seek cost-saving efficiency measures throughout the regulatory period. This profit sharing incentive gives the utility's management and shareholders a direct stake in cost-saving measures and thus overcomes the moral hazard problem that plagues cost-of-service regulation. In addition, these cost savings can help offset the investments that are likely necessary to accommodate changes in network use, improve network resilience, and implement smart grid and active system management techniques, all of which are growing priorities for various regulatory jurisdictions. Furthermore, the regulator can establish the strength of the efficiency incentives as desired through the design of the menu of contracts (i.e., setting the SFref and SFroc discretionary parameters). * The "slow money/fast money" approach to capitalization of allowed ex post network expenditures also equalizes incentives for the firm to optimize costsaving tradeoffs between network investments (CAPEX) and operational expenditures (OPEX). Without this approach, the utility may face distorted incentives that encourage over-spending on network assets in lieu of cost saving operational expenditures, including innovative contractual arrangements with the owners of distributed energy resources. Removing this distortion and equalizing cost-saving incentives across both categories of network expenditures is thus an important step to encouraging cost-saving active system management approaches and encouraging an evolution in the distribution utility business model. 89 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Balances fundamental regulatory tradeoffs * The regulator has significant flexibility and discretion to set the strength of the sharing factor parameters used to create the menu of contracts in order to balance the fundamental regulatory tradeoffs between allocative efficiency (extracting rents from the utility) and X-efficiency (providing incentives for cost savings). " Furthermore, the incentive " compatible nature of the menu of contracts will encourage firms with significant cost-saving opportunities to select a higher-powered incentive (thus improving Xefficiency) while firms closer to the efficient frontier will select a lower-powered incentive (improving allocative efficiency). Firms thus reveal their cost type, and the resulting regulatory contract appropriately balances the moral hazard and adverse selection challenges. Finally, it is important to note that this regulatory approach only considers the establishment of allowed TOTEX and the primary allowed revenues. These methods must be accompanied by appropriate incentives for the utility to maintain and improve quality of service, meet other performance expectations (including customer service quality and environmental performance) and engage in long-term innovation efforts (which are not adequately incentivized by the relatively short-term cost-savings incentives created herein. (For more on these important considerations, see: Bauknecht, 2011; Cossent, 2013; Lo Schiavo, et al., 2013; Ofgem, 2010c, 2013c, 2013d; Perez-Arriaga et al., 2013). References Bauknecht, D. (2011). Incentive Regulation and Network Innovations. Retrieved from http://cadmus.eui.eu/bitstream/handle/1814/15481/RSCAS_2011_02.pdf?sequenc e=1 Cossent, R. (2013). Economic Regulation of Distribution System Operators and its Adaptation to the Penetration of Distributed Energy Resources and Smart Grid Technologies. Comillas Universidad Pontificia. Cossent, R., & G6mez, T. (2013). Implementing incentive compatible menus of contracts to regulate electricity distribution investments. Utilities Policy, 27, 28-38. doi:10.1016/j.jup.2013.09.002 Cossent, R., G6mez, T., & Olmos, L. (2011). Large-scale integration of renewable and distributed generation of electricity in Spain: Current situation and future needs. Energy Policy, 39(12), 8078-8087. doi:10.1016/j.enpol.2011.09.069 90 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Cossent, R., Olmos, L., G6mez, T., Mateo, C., & Frias, P. (2011). Mitigating the Impact of Distributed Generation Advanced Response Options. International Transactions on Electrical Energy Systems, 21(6), 1869-1888. doi:10.1002/etep.503 Cross, P. S. (2012). How Much is Enough? Utilities face rate pressure as financing costs hit rock bottom. Public Utilities Fortnightly. Retrieved April 05, 2014, from http://www.fortnightly.com/fortnightly/2012/11/how-much-enough Crouch, M. (2006). Investment under RPI-X: Practical experience with an incentive compatible approach in the GB electricity distribution sector. Utilities Policy, 14(4), 240-244. doi:10.1016/j.jup.2006.05.005 Domingo, C. M., G6mez, T., S nchez-Miralles, A., Peco, J., & Martinez, A. C. (2011). A Reference Network Model for Large-Scale Street Map Generation. IEEE Transactions on Power Systems, 26(1), 190-197. Eurelectric. (2013). Active Distribution System Management A key toolfor the smooth integration of distributed generation. Brussels, Belgium. Retrieved from http://www.eurelectric.org/media/74356/asmjfull-report_discussion-paperfinal2013-030-0117-01-e.pdf Fern ndez, L. P., G6mez, T., Cossent, R., Domingo, C. M., & Frias, P. (2011). Assessment of the Impact of Plug-in Electric Vehicles on Distribution Networks. IEEE Transactions on Power Systems, 26(1), 206-213. Jamasb, T., & Pollitt, M. (2008). Reference models and incentive regulation of electricity distribution networks: An evaluation of Sweden's Network Performance Assessment Model (NPAM). Energy Policy, 36(5), 1788-1801. doi:10.1016/j.enpol.2008.01.034 Lo Schiavo, L., Delfanti, M., Fumagalli, E., & Olivieri, V. (2013). Changing the regulation for regulating the change: Innovation-driven regulatory developments for smart grids, smart metering and e-mobility in Italy. Energy Policy, 57, 506-517. doi:10.1016/j.enpol.2013.02.022 Ofgem. (2009). Electricity Distribution Price Control Review Methodology and Initial Results Paper. London. Ofgem. (2010a). Electricity Distribution Price Control Review 5: Special conditions of the Electricity Distribution License. London. Ofgem. (2010b). Handbookfor implementing the RIO model. London. Ofgem. (2010c). RIIO: A new way to regulate energy networks. London. 91 Chapter 3: Demonstrating the Proposed Regulatory Process Jesse D. Jenkins (2014) Ofgem. (2013a). RHO T1 Financial Model (Electric). Network Regulation - the RHO Model. Retrieved October 05, 2014, from https://www.ofgem.gov.uk/network-regulation-- riio-model/price-controls-financial-model-pcfm/riio-t1-financial-model-electric Ofgem. (2013b). Strategy decision for the RIIO-ED1 electricity distribution price control: Financial issues. London. Ofgem. (2013c). Strategy decision for the RIIO-ED1 electricity distribution price control: Outputs, incentives and innovation. London. Ofgem. (2013d). Strategy decision for the RIIO-ED1 electricity distribution price control: Overview. London. Ofgem. (2013e). Strategy decision for the RIIO-ED1 electricity distribution price control: Uncertainty mechanisms. London. Olmos, L., Cossent, R., G6mez, T., Mateo, C., Joode, J. De, Scheepers, M., ... Gerhardt, N. (2009). Case studies of system costs of distribution areas (No. WP 4, Deliverable 5). Petten, the Netherlands. Retrieved from http://www.improgres.org/fileadmin/im progres/user/docs/D5_casestudiesof-sy stem_costs.pdf Perez-Arriaga, I., Ruester, S., Schwenen, S., Batlle, C., & Glachant, J.-M. (2013). From Distribution Networks to Smart Distribution Systems: Rethinking the Regulation of European Electricity DSOs. Florence. doi:10.2870/78510 Poudineh, R., & Jamasb, T. (2014). Distributed generation, storage, demand response and energy efficiency as alternatives to grid capacity enhancement. Energy Policy, 67, 222-231. doi:10.1016/j.enpol.2013.11.073 Schmalensee, R. (1989). Good Regulatory Regimes. The RAND Journal of Economics, 20(3), 417-436. doi:10.2307/2555580 Trebolle, D., G6mez, T., Cossent, R., & Frias, P. (2010). Distribution planning with reliability options for distributed generation. Electric Power Systems Research, 80(2), 222-229. doi:10.1016/j.epsr.2009.09.004 U.S. Bureau of Labor Statistics. (2014). PPI Databases -Electric power distribution PCU221122221122. Producer Price Indexes. Vergara, C., Perez-Arriaga, I., Mateo, C., & Frias, P. (2014). Estimating the Aggregate Impact of Distributed Photovoltaic Generation over Distribution Networks (No. lIT14-034A). Madrid. 92