Meter Data Management: The Key to Unlocking the Benefits of Advanced Metering Aclara Software White Paper March, 2008 © Aclara Software Inc. 2008 1 Introduction: Why MDMS? Until recently, even though most utilities have had to cope with some amount of interval data, the category of systems known as Meter Data Management Systems (MDMS) essentially did not exist. The recent exploding interest in Advanced Metering Infrastructures (AMI), along with the introduction of systems that require interval data, such as settlement systems, has driven the need for an entirely new generation of class of Meter Data Management Systems. Given current industry drivers of system efficiency, including demand-side management, the potential of carbon credits, and the linking of retail to wholesale pricing, the industry is driving to deploying AMI. But making the most of AMI data is not a simple matter. Over the past several years, the concept of an MDMS as the key software enabler for AMI has begun to evolve, as utilities have begun to understand that AMI cannot achieve all of the desired benefits unless the data can be effectively cleansed, processed, stored, and applied to address and enhance key business processes. The long list of benefits of implementing an AMI system is compelling. If managed properly, the availability of AMI data can generate significant benefits and enhance a range of processes, including not only internal-facing processes such as revenue protection, outage management, and distribution system planning, but perhaps even more so customer-facing processes that result in enhanced customer satisfaction and reduced customer service costs. The “customer side of AMI” is where some key benefits are gained; for example, the combination of AMI/MDM should be able to empower customers to truly leverage creative pricing programs that have been shown to effectively reduce resource requirements. Thus, many of the benefits promised by AMI relate to how metered data is managed after it is collected, and how it is stored, formatted and deployed across the enterprise. In other words, collecting the data is just one part of the AMI savings equation; the utility must also be able to use its AMI data to support decisionmaking throughout the organization to achieve the maximum return on its AMI investment. A critical step, therefore, to maximizing the potential of AMI is the creation of a central repository that is the single source for all metered and interval data. Further, given the ever-increasing amounts of AMI data, such a repository must have the ability to scale in volume and in data complexity to accommodate future needs, without the risk of long-term costs for database changes. This system - the MDMS - must be built not only with these AMI business processes and data needs in mind, but also with the understanding that several other converging trends within the utility industry are transforming most utilities even more into data management organizations in addition to the traditional “wires management” role. This is being driven by regulatory demands and rapidly evolving technology and standards that affect both the quantity and quality of data that are becoming available, all the way from inside the substation fence to inside the homes of residential customers. © Aclara Software Inc. 2008 2 An effective MDMS must recognize and support all of these evolving trends and requirements. MDMS’ Place in the Range of Utility Enterprise Systems Before describing what constitutes an effective MDMS, let’s explore both why there is a need for a new system, and what an MDMS is not. CIS vs. MDMS One question often asked is whether an entirely new system is needed, or whether these new needs can be met by expansion of the Customer Information Systems (CIS). The CIS has evolved into the full cycle “meter-to-cash” engine in the utility and, as such, the key system used to leverage meter data. However, CIS system design pre-dates the needs that are driving the current interest in Meter Data Management. Energy companies looking to fully exploit interval data need to consider the roles that can be played by their CIS systems and the role that should be addressed by separate MDMS systems. A CIS ingests meter data, typically via a monthly consumption read (although there is always some more granular data that is incorporated into the appropriate buckets). When this data has been stored, monthly bill cycles are run to create invoices, generally based on simple billing determinants. Once the customer has been invoiced, related areas are addressed by the CIS, such as credit and collections, payment arrangements, financial functions, etc. CIS systems are often the launching point for CSRs to facilitate other customer activities that are managed by downstream systems, including taking service orders and inputting outage reports. However, even before the advent of MDMS systems, CIS systems were generally insulated from the raw details of interval metering data by other systems, such as MV90. For example, CIS systems were not designed to perform validation on interval data; such validations were performed elsewhere. In addition, most CIS systems only handle simple rates for the majority of the customer base. Commercial and Industrial customers with more complex rates requiring interval data are often handled by specialty complex billing systems or manual processes that send information to the CIS in an aggregated manner for invoicing and posting to the general ledger. The MDMS is being increasingly seen as the logical broker between the AMI systems and any enterprise systems that may need access to AMI data or AMI actions – such as on demand reads or connect/disconnect – in real or near-real time. By leveraging an MDMS, interactions with multiple AMI (and other) systems can be accomplished via many-to-one rather than many-to-many relationships. CIS systems were never intended to play this type of role. © Aclara Software Inc. 2008 3 Today’s CIS systems generally do what they do very well, but they were simply not designed to play a central role in processing and managing large quantities of interval data, nor managing a set of real time interactions with meter data collection systems. Adding that capability to a CIS will increase complexity in what is an already overly complex system in most cases, and increase overall risk of system problems. In fact, an all encompassing CIS system would run counter to the current trend of moving to best of breed, componentized systems. As a pundit once observed, CIS systems know very little about the “Customer” nor have a great deal of “Information”; they are essentially a form of accounting system that performs billing. Extending them into a full featured meter data processor and repository stretches them well beyond what they were intended to be. The diversity of systems resulting from using a separate MDMS enables innovation and flexibility, reduces risk, and centralizes the issues of complex interval data management in a system where it can be accessed through the increasing array of standards. Meter Data vs. Meter Asset Management There is often also confusion about Meter Data Management Systems vs. Meter Asset Management Systems. It is sometimes presumed that a new MDMS will also serve as a Meter Asset Management system. The two functions are, of course, closely related, and share a fair amount of common information. However, these two systems play very different roles with very different needs. A Meter Data Management system is all about usage at the Service Point. One needs to know what meter is at a Service Point at any point in time, but the key metrics being tracked deal with usage and activity, regardless of what meter is installed. Conversely, a Meter Asset Management system, or more correctly a Meter Device Management system, is all about the meter or other device. In this case, the key metrics being tracked deal with device performance, no matter where it is located, as well as device location at any particular point in time. Thus, while these 2 systems need to be joined at the hip, trying to cram both functions into a single system can result in sub-optimal performance for both. Once again, the role of the CIS is an issue. In many cases, utilities did not have effective Meter Device Management systems in place. As a result, the CIS has been made the system of record on meter assets and their location. But like other functions, this was never a capability intended for a CIS. An effective Meter Device Management System tracks all information about meter assets, including test results and other indications of performance, device associations, and cradle to grave location, not just location at © Aclara Software Inc. 2008 4 premises. If there is such a system in place, it makes much more sense for the Meter Device Management System to be the system of record. Device Events Device Messages Asset Communications ID’s AMI Event Codes CIS ADMS SOA MDMS Device Configuration Device Groups Device Class Device Associations Firmware Versions Firmware Upgrade Location and GPS Information Site Configuration Information Device History Information Meter Changes – Change Outs, New Meters and Removed Meters A Meter Device Management System – much like the MDMS - becomes particularly important with the roll out of an AMI system. First of all, with an AMI rollout there will be tens of thousands or even hundreds of thousands of meters and other AMI devices being installed in a year that need to be tracked and supported. Secondly, there will be much more complex device associations than in the past – as well as the need to test and calibrate not just meters but an integrated set of devices. An AMI Device Management System – or AMDS 1 to differentiate it from an MDMS – becomes an important prerequisite for AMI. Thus, while a single system should not be expected to perform both functions, the MDMS and ADMS need to be planned together, and designed to be tightly integrated. This should go so far as to allow users looking at meter asset information in the ADMS, for example, to easily link to and pull up relevant usage information in the MDMS, and vice versa. 1 We have switched to the ADMS term from the more commonly applied term Meter Asset Management because that is sometimes confused with a Financial Asset system. In most cases, utilities have an enterprise Financial Asset System covering all capital assets, including meters. That system needs to be interfaced with the ADMS, but an ADMS should not be expected to perform this function. © Aclara Software Inc. 2008 5 Elements of an Effective MDMS An effective MDM solution consists of multiple elements: 1. First and foremost, the underlying MDM repository and infrastructure must be able to effectively: x ingest AMI data from multiple sources in a timely manner; x process AMI data as required - including performing Validation, Estimation and Editing (VEE) to cleanse the database, and creating billing determinants, not only for traditional mass market rates but for evolving time-based rates as well; x store data for all commodities in a way that supports smaller utilities and those just focused on C&I customers, but scales to meet the needs of larger utilities with millions of customers and hourly or sub-hourly data; x distribute the data to the appropriate systems, including the CIS. To do this effectively, the core MDM architecture must be designed with full recognition of the way in which the data will be used by other systems; without that perspective from the outset, there is a risk that large inefficiencies will be introduced, or that some potential benefits are not achieved; x manage events relating to AMI data and equipment, serving as the broker between other systems – such as OMS – and the AMI systems in order to take full advantage of the availability of AMI on a real-time basis, and; x publish the data for use by downstream systems, and for customer and other stakeholder or partner access, and make it available for ad hoc reporting. 2. Second, an overall solution should offer a set of integrated business applications that address key operational processes which benefit most from the availability of AMI data, such as revenue protection, load research, load forecasting, settlement, complex billing, and distribution asset optimization. 3. Finally, the overall solution should also address the customer side of AMI. This needs to go well beyond the simple presentation of load profiles to customers - many of whom will not be able to readily leverage such information – to include the appropriate systems and tools that enable customers to understand their options and the impact of these options, and to take actions. Key issues in delivering these types of capabilities are explored in subsequent sections below. Critical Characteristics of an MDM Data Repository An MDMS must be built around the needs of managing large quantities of data. To best do this requires more than just using the types of databases generally found in CIS systems, which are typically organized around customer and premise. The MDMS must © Aclara Software Inc. 2008 6 bridge the gap between incoming interval meter data and enterprise-wide business functions, such as CIS, in an optimized, scalable, and efficient manner. To accomplish this, an MDMS needs to look more to best practices in data warehouse than to CIS data structures. For example, the star schema architecture has been shown within the data warehousing community as an effective design methodology for large volumes of infrequently changing time series data, in terms of both performance and scalability. Further, “wide storage” – where data for interval units of 1 hour or less are stored as columns in a daily record rather than as 24 or more individual records – can significantly reduce storage requirements and increase performance. Consider the case of 15 minute data, for example. In a traditional “tall storage” configuration, a day’s worth of 15 minute data would require 96 records. In a wide storage environment this is a single record, resulting in a dramatic savings in storage requirements. Moreover, since an MDMS is a very I/O intensive application, this also directly translates into significantly faster performance. These types of architectural issues are critical to scalability – the ability for the MDMS to process and store data for utilities with millions of meters and hourly or sub-hourly data for all. An N-Tier architecture that enables the system to leverage multiple application and database servers is also critical to achieving scalability. In addition, an MDM needs to support load balancing and leverage multi-threading capabilities in order to effectively take advantage of scalability provided by the inexpensive addition of application servers. Data in the repository should also be optimized for integration with analytical engines that can be used to drive other functionality. For example, segmentation and aggregation are critical to such functions such as load forecasting, settlement, distribution planning etc. Furthermore, the repository needs to be able to store multiple versions of data resulting from different processing steps, and readily apply the right data to the right business process. The other key question is the overall data structure architecture. Many CIS systems have hard-wired data schemas that are very difficult to change. Given the fact that AMI and MDM are still in their infancy, and markets are in a state of flux, it is critical that the MDMS be able to quickly adapt to changing market conditions. Thus, a rigid data schema that requires extensive coding or scripting and table changes every time there is a need for a new data element will have significant implications on cost and time to market. Instead, it is important for an MDMS to have a flexible data structure that lends itself to rapid change, without the need for programming. Ideally, the system should be designed, perhaps with the use of a meta-data layer, to enable the business users to add new database objects and attributes and to change relationships without the need for database table or structure changes. © Aclara Software Inc. 2008 7 Effective Data Processing Validation Estimation and Editing (VEE) One of the most basic functions expected of an MDMS system is the Data Cleansing process, commonly referred to as Validation, Estimation and Editing, or VEE. Processing interval data collected via AMI is much more complex than processing consumption data collected via manual or semi-manual methods. An effective MDMS should not only deal with this level of complexity, but also provide significant flexibility in how data cleansing is accomplished. In fact, flexibility is the key to delivering effective data cleansing. The following must be considered in the design of an effective VEE process: x There are a number of generally accepted validation tests that any MDM needs to perform, such as Missing Interval Check, Zero Check, Negative Value Check, Static Value Check, Spike Check and Sum Check. (These are all part of the standard California VEE rule set). However, an effective MDMS also needs to provide users with the ability to easily create custom validation – and estimation – rules for a variety of purposes. Requiring programming or scripting to create special rules only adds to the cost and time associated with system deployment and operations. x The system should allow the user to create different VEE work flows for different groups of customers; for example based on rate class or perhaps competitive supplier in a deregulated market where the distributor manages the MDMS. There needs to be a simple way to group customers together for VEE, and even the ability to include customers in different groupings for different purposes. x In fact, it is easy to envision how different uses of the data may require different VEE rules, or even different sets of data. Revenue Protection analysis, for example, clearly requires the use of raw data for analysis purpose, rather than data where 0 values have been estimated. On the other hand, a Settlements system might require an additional level of VEE on the already processed data, based on the local market rules. An effective MDM will provide full flexibility in establishing VEE work flows. x As suggested above, VEE is often driven by the regulatory environment. There may be radically more complex VEE requirements imposed in a deregulated area where data needs to be delivered to multiple parties. A system needs to be able to support different VEE work flows in different jurisdictions. x The availability of interval data provides an unprecedented ability to create accurate customer load profiles. The MDM needs to be able to readily leverage these load profiles, certainly for the Estimation process where they can reliably fill in missing data with the appropriate weather modeling, but perhaps for use in Validation testing as well. © Aclara Software Inc. 2008 8 x A VEE workflow must also consider the overall business process. Under what circumstances are data automatically estimated, and under what circumstances should there be manual intervention? How are exceptions routed and handled? x Finally, the VEE process must recognize the importance of timing. When should a work flow be scheduled and when should it be triggered by another event, such as data loading? How much if any VEE should be done as the data is being loaded, and how much needs to wait until the end of the day so that sum checks can be run and estimations performed with all available data points? Again, an MDM must be able to demonstrate significant flexibility in order to maximize effectiveness. Billing Determinants Historically, the creation of billing determinants, at least as it relates to mass market customers, has been the domain of the CIS. As AMI systems began to be implemented this essentially did not change. Typically the CIS would use monthly consumption reads to create billing determinants, even in cases where daily data was available. Utilities are now consistently looking for the MDMS to create the billing determinants. There are a variety of reasons for this. First and foremost, utilities are looking to isolate the CIS from the AMI systems – particularly where they may be more than one. The MDMS, therefore, becomes the logical choice for this function. Secondly, with the movement to expand the use of Time-of-Use rates, many CIS systems simply do not have the ability to create billing determinants for all customers. Introducing the MDMS in the mix, however, creates a major timing risk. Many utilities are already challenged to fit the billing process into the defined overnight time windows. Now there must be synchronization with a separate system. Moreover, that system is busy collecting interval data from all customers on a daily basis, not just from those customers on the daily billing cycle, and these data loads can take a considerable amount of time. For an MDMS to play this role effectively, it needs several things: 1. A user controlled method for easily creating any type of billing determinant, so that the system can readily compute the appropriate determinants, and rapidly respond to changes over time, particularly as new rate classes are introduced. 2. A high level of performance, to ensure that all of the required processes are performed in the available time windows. 3. A robust scheduling system to ensure that different processes can scheduled to avoid contention. It is too early in the evolution of the MDMS to determine how well different systems perform this critical function. © Aclara Software Inc. 2008 9 MDMS as a Broker of Requests to the AMI System and Manager of Events As utilities consider the business case of rolling out an AMI system, it quickly becomes apparent that the AMI system needs to interact with numerous enterprise systems, in many cases in real time. Is it necessary to provide direct interfaces between these systems and the AMI system? A recent IBM Utility Integration Whitepaper included the following diagram and noted: “The point-to-point approach results in numerous complex connections that need to be maintained” In fact, as the concept of an MDMS has evolved over the last 2 years, it has become increasingly clear that an alternative to direct linking between all systems and each AMI head-end directly is to leverage the MDMS to broker requests between the AMI systems and the other systems. There are several reasons that this makes sense: x First, in many cases utilities are putting in AMI systems from 2 or more vendors, in order to spread risk, leverage different technologies better suited to specific geographical areas (e.g., utilize Power Line Systems in rural areas while relying on RF in denser parts of the service territory), or take advantage of different technologies for different commodities. Whatever the reason, multiple AMI systems would require many-to-many interfaces between the enterprise systems and the AMI systems. Alternatively, leveraging the MDMS as a broker requires only many-to-one interfaces with the MDM, letting the MDM worry about identifying and communicating with the right AMI system. x Not only does the existence of multiple AMI systems complicate interface requirements, but it also requires that there be appropriate logic to know which system to interact with. This type of logic would need to be added to each enterprise system interface. Instead, with the MDMS as a broker, only the MDMS © Aclara Software Inc. 2008 10 need maintain this type of logic, and a well designed MDMS will support this outof-the-box. x Finally, other than perhaps the case of the newest systems, most enterprise applications were not designed with real time interaction in mind. While even with the MDMS acting as broker there will still need to be real time integration with the MDMS itself, the MDMS can worry about the greater complexities of requesting and retrieving data from the AMI system to address On Demand Reads, Connect/Disconnect Requests, Voltage Checks, and other services made feasible by the availability of AMI. With this in mind, there are a range of events for which an MDM could become a central component. CIS AMI OM Manual Reads\ MDMS WM SCADA Other For example, the availability of AMI can support outage management in a variety of ways. If an outage is reported, the AMI system can be accessed to determine if, in fact, an outage exists, and potentially the extent of the outage. To accomplish this, either the OMS or the MDM ideally requires an effective mechanism to be able to determine which end points should be contacted, and how the information being retuned should be interpreted. Where the AMI system is sending outage flags in a wide-scale outage situation, the MDMS needs an effective filtering mechanism to keep from overwhelming the OMS with messages. The MDM can play a role during the restoration process as well, selectively pinging meters that are deemed to have been restored, and checking downstream assets. While some newer OMS systems have an effective predictive modeling capability to optimize decisions about where to dispatch crews and what parts of the network are likely restored, it is conceivable that an MDM which incorporates a circuit model can dramatically enhance this type of capability. While no MDM has yet played this © Aclara Software Inc. 2008 11 type of role, this is a future model that should be considered, particularly for those utilities that continue to use legacy OMS systems. Another key event that can be managed by an MDM is demand-response. This can take a variety of forms. At the simplest level, an MDM system can be told to implement direct load control at specific end point points, and then provide those requests to the appropriate AMI head-end systems that support load control. However, it is conceivable that a utility may have multiple demand-response programs, including direct load control, Critical Peak Pricing (CPP), and a program that pays participants for load shedding, perhaps managed by a third party. An MDMS: x will either need to know what customers are on what program, or be able to accept that information from an external source such as a CIS; x may need, in the case of CPP, to initiate an alert to the customer and perform ondemand reads at the start and end of the event, or more frequently depending upon the way the customer is being billed; x might provide, for Real Time pricing, day ahead pricing notification through both an in-home display and through Internet presentment, offer an analysis of options for reducing costs during high price periods, and provide after the fact presentation of both load and rates; x needs to know; for 3rd party managed load shedding programs, which parties to contact, be able to perform on-demand reads and calculate baselines, and create the appropriate billing determinants, and; x perhaps longer term, have the intelligence to accept only the level of demand reduction required, and then determine the optimum way to achieve that load reduction through all of the available options. Again, there are no MDM systems in place today that even approach this level of functionality in managing demand-response events, but the potential is there for MDM to play a broad role. Complex Event Processing These types of transactions are examples of complex event processes which require the handling of transactions, events and exceptions identified by the MDM or other systems and must process data received from - and needing to be sent to - the various AMI headend systems as well as other source systems. While the processes described above may sound fairly basic, in fact the number of potential events, exceptions, and resulting © Aclara Software Inc. 2008 12 workflows generated from MDM can vary greatly, and some of the workflows can grow to be quite complex. As a result, an effective MDM should provide some basic Business Process Management (BPM) or workflow capabilities that allow the systems integrator initially, and utility staff ultimately, to readily model work flows and implement processes based on them. An effective BPM implementation can: x Trigger system-to-system business processes without requiring manual intervention to address exceptions. This is extremely important given the potential volumes of transactions that will need to be managed and filtered. x Intelligently schedule and assign those activities that do require manual intervention to the right users based on their roles and workload. This may include the need to communicate these tasks using several channels – the MDMS application, e-mail, SMS, PDAs, etc. For some events, the appropriate escalation may need to occur – for example when tasks have not been completed within a certain time period, a supervisor may need to be notified. x Provide operational reporting to understand the health and effectiveness of each business process. Data Integration In the previous section, the need to integrate an MDM with AMI systems and enterprise systems was addressed. In addition to these real time interface requirements, there is also a need for bulk loads of data, including both AMI data and data that will be needed from systems such as the CIS. An MDM needs an effective strategy for dealing with both types of integration challenges Newer generation systems leverage a Service Oriented Architecture (SOA) to support integration. SOA has rapidly become an overused industry buzzword that is defined differently based on different points of view, but in general it refers to an architecture that packages business processes into services. Services communicate with each other by passing data from one service to another. SOA allows services to be combined to create © Aclara Software Inc. 2008 13 new application, and different applications to exchange data as part of a business process. In an SOA, data is exposed so that it can be shared with other systems. To facilitate integration between applications, most utilities are in the midst of rolling out either an Enterprise Service Bus (ESB) or Enterprise Application Integration (EAI) architecture. While these architectures are tightly related, they are in fact different, and there is considerable confusion about what each represents and what the differences are. This is not the forum to explore the differences, so the discussion will only address the ESB. An ESB generally provides some level of abstraction on top of a messaging system layer, which allows the value of messaging to be gained without having to write code. An ESB does not by itself implement a SOA, but provides features that can be used to implement one. Similarly, ESB is not necessarily web-services based. Based on EAI rather than SOA patterns, it tries to remove the coupling between the service called and the transport medium. Some MDMS vendors have built their systems around a particular ESB application. This has concerned some utilities that are reluctant to introduce multiple ESB architectures in a particular implementation. Instead, it may make sense for MDMS vendors to focus on the lowest common denominator, which is the messaging layer itself. There is active work ongoing to create comprehensive messaging standards relating to AMI. For example, the MultiSpeak initiative has created a set of standards that have been well adopted by vendors supplying solutions to rural cooperatives in particular. The IEC Technical Committee 57, Working Group 14 has been actively developing standards (more commonly known as CIM - Common Information Model), which are similar but cover a broader range of processes and specifically intended for larger utilities. By adopting these types of messaging standards – to handle functions such as On Demand Reads etc. - an MDM vendor can easily integrate with any EAI architecture or ESB. Nonetheless, ESB functionality within an MDMS will not only support environments where no ESB is present, but may also simplify integration even where the utility has deployed an ESB architecture. A robust integration strategy will require multiple elements. © Aclara Software Inc. 2008 14 Current Elements of IEC 61968-9 Historically, messaging standards primarily fell into a category known as “Request-Reply”, where System A would make a request of System B (e.g. an on-demand read), and then receive and process a reply from System B (e.g. the current register read). Some AMI interactions may take advantage of a “Publish-Subscribe” model. This model is being deployed where many entities may need to know about certain events. A broker is introduced between the “Publisher” and a series of “Subscribers”, so that many to many interfaces are not required. This form of messaging may be appropriate in the event of outage flags being generated by the AMI system, so that the OMS, CIS and other appropriate systems are notified. An effective MDMS should be able to leverage both Request-Reply and Publish-Subscribe Messaging. While all of these concepts could be important for batch data loads, there are in fact many situations in which a more effective strategy would be a simple batch file transfer. For example, an ESB might introduce overhead that will affect performance, which is critical when loading millions of records. An MDM needs to have an effective approach for © Aclara Software Inc. 2008 15 loading large batch files. This requires capability to Extract, Transform and Load data., including robust routines for processing data as it is being loaded. For example, dealing with DST might be most effectively handled as part of an interface. While comprehensive VEE cannot be performed during data loads, as suggested earlier, some amount of validation processing should be done as data is being collected so that exceptions can be dealt with. An effective MDM integration strategy needs to consider all the ways in which the MDM must communicate with other systems. Integrated Business Applications Providers of Meter Data Management Systems come from several different backgrounds. Some previously focused on the process of collecting interval and other meter data. Others are trying to move from the CIS world Another group of MDM providers have come from the application side; this group focused on areas such as load research, settlements and complex billing, or other applications that required the creation of a database that could ingest and manage interval data. An advantage that this background brings is that it required that there be a technical architecture, data model, and analytic engines created that enable sharing of useful information, and not just focused on the storing of data. For an MDMS to be effective in leveraging the data, it should be able to seamlessly integrate and synchronize multiple data types, including customer information, network data, system level metering, and weather information, and aggregate data to the appropriate levels. MDMS systems that are not designed with these types of capabilities in mind may not be able to readily support key business applications that can leverage this data. In addition, such consideration in the design of an MDM may yield advantages in the core MDM functionality. For example, consider load profiling. Certainly, load research, forecasting, and settlement systems must rely heavily on load profiling capabilities. But profiling is a key ingredient in Estimation, and could play a role in Validation as well. An MDM that incorporates a load profiling engine to support business applications is likely to have a more robust VEE capability. One of the types of business applications that is not even practical without AMI data is Revenue Protection. AMI data enables the Revenue Protection group to better analyze loss of revenue opportunities. An effective Revenue Protection application will identify potential areas of revenue loss due to theft, malfunctioning meters or other factors. This may involve standard and custom tests that leverage interval data as well as AMI event data, such as an indication of an outage. Conceivably, load profiles could be leveraged here as well. A system could include Case Management capabilities to help the Revenue Protection group manage the process. An MDMS that has an integrated Revenue Protection application is more likely to support this critical process than one focused © Aclara Software Inc. 2008 16 exclusively on data management, and such an application itself will generate significant value. Similarly, electric utilities deploying AMI often look to Transformer Load Management (TLM) as an area of potential benefits. However, with the appropriate tools it is possible to leverage AMI data for a broader set of distribution analysis applications. Leveraging segmentation, profiling and aggregation, as well as integration with the GIS system, an application should be able to accurately estimate hourly load at any point on network, even when hourly data is not available. This information can help utilities avoid device failures and outages caused by overload, while also reducing capital expenditures by enabling more precise sizing and timing of planned system additions. In addition, such a system should be able to improve phase balancing and circuit utilization, as well as voltage regulation and capacitor placement. An MDM that has these types of uses in mind as a result of an integrated Business Application is more likely to have the underlying engines as well as data structure to ensure that these types of analyses can readily be fed. The Customer Side of AMI One desired benefit of AMI is to reduce customer service costs by making detailed usage information available to customers, through both customer self-service and the call center. This can help more quickly resolve billing issues, while at the same time educating customers about the relationship of their behavior to energy costs. Moreover, a key driver for AMI is interest in improving the cost-responsiveness of consumer behavior. Meter systems are needed to provide the hourly cash registers for time-based pricing - which holds the potential to be a cheaper, cleaner, safer way of meeting future system load requirements. But deciding how to deliver meter data to customers in a meaningful manner is a key issue. © Aclara Software Inc. 2008 17 This goes beyond simply choosing whether to display the data on the Web or through an In-Home Display, described further below. An effective and comprehensive approach requires that the information be provided in a way that customers can understand and act upon. The availability of interval data makes it possible to provide customers with a rich set of information to help them better manage their energy use. In too many cases, however, load data has been viewed by utilities as a separable item, not integrally tied with other information of interest to customers, and not provided with tools to help the customer make sense of the information. While this approach may work for larger companies that have available energy expertise, it has not worked effectively for mass-market customers who will now have access to interval information for the first time. To deliver information that will work, it is important to tightly integrate usage information with billing information, which is at the core of utility-customer communications, and provide customers with a rich set of tools they can use to take best advantage of this valuable new information. For example, the types of tools that become possible with an AMI system include x A “bill-to-date” function that estimates a customer’s bill at any time in the month based on the AMI data x Rate Analysis that not only allows the customer to play what-if games to compare alternative rates, but also automatically notifies the customer viewing their bills online how they would have done on alternative rate scheme, and can combine the effects of rate changes with efficiency actions. x E-mail and On-Line Alerts to customers when they appear to be at risk of going over pre-set monthly usage/costs, and alerts for time-of-use customers of shifts in usage from off-peak to peak with an estimate of impacts. x Load Shift Analyses that can be used to help customers understand how they can shift load to the off-peak and how much they might save. x Targeted promotion of key programs based on analysis of AMI data. © Aclara Software Inc. 2008 18 x Analysis of the impact of different actions to mitigate higher pricing, and presentment of load overlaid with actual prices so customers can see their usage during periods with different price points. These types of tools are examples of what can be accomplished with an MDMS architecture that provides value-added analytics, not just data storage, and is designed with the full range of potential uses of data in mind. Extending MDM into the Home or Business There has been much talk in the past year about the need to access Home Area Networks (HAN) for both customer communications and support of demand-response programs. Indeed, the Open HAN Committee has been working on another set of standards to support messaging with HAN devices. There has also been a sometimes spirited debate about how to communicate to the HAN. On one side are the proponents of the meter network as the communications channel, and the use of In Home Displays to reach customers. This has led to debates about the bandwidth that will be required for the meter communication network. On the other side have been proponents of using the Internet as a means of communications, with the PC as the display device; this approach would, by definition, be significantly less expensive. Arguments against this channel range from the need to leave computers on in the home to the fact that not all homes have Internet connections, As with many such debates, the reality probably lies somewhere in the middle, with there potentially being a place for both approaches. From an MDM perspective, the key is to make sure that the MDM system adheres to the rapidly developing HAN messaging standards to ensure that it can play a role as HANs evolve. One of the arguments for the use of the meter network and an In Home Display is that this is the only way to get up-to-the minute information to the customer. This is exacerbated by the fact that, given the way data has been collected in some AMI systems, by the time the customer sees the data on the utility’s portal it may be 2 to 2 days out of date. The extent to which this will actually be required, however, is still a matter of debate. Sometimes lost in this discussion, however, is the fact that data coming directly from the meter has not gone through any VEE process. In the end, this may represent a serious concern to utilities who are, rightfully, worried about the quality of data customers see. A better solution might be for AMI systems to deliver data to the MDM on a more frequent basis, so that customers see at least the previous day’s data. More real time data could then be routed directly from the meter to a PC on the HAN, to be integrated with the data available over the Web from the central repository. Basic validation can then be performed via the portal on this small amount of more recent data, so that customers can be assured of getting clean information. © Aclara Software Inc. 2008 19 The Future of MDM The term Meter Data Management is already a misnomer. As noted earlier, typically an MDM will store lots more than just meter data, and include weather data, asset data, GIS data, and more. In the future, an MDM may be called on to do a lot more. As Smart Grid initiatives expand, an MDM may be asked to collect data from sensors located all across the network. In fact, some MDM systems already are being used to collect summary SCADA data. Data synchronization requirements will expand as a broader range of data is ingested and processed. Built in MDM analytics will need to expand to include a range of predictive modeling capabilities. The Meter Data Management System may evolve into a full featured Network Data Management System, or NDMS. Network + Cost + Asset + Customer + Meter Data 9 Failure prediction Cost + Asset+ + Asset Customer+ + Customer MeterData Data Meter 9 Detection & 9 ROI and ROA Based Decision-Making 9 System Loss Detection Response Profitability Profitability Optimization Optimization 9 Customer Care 9 Outage 9 Pricing Design Detection & 9 Response Forecasting 9 Settlement 9 System Loss 9Detection Unbilled Revenue Reporting Revenue Prot. 9 Settlement Complex 9 Pricing Design 9 Interval Meter Validation, Estimation, Billing 9 Unbilled Revenue Editing 9 Forecasting Reporting 9 Customer Revenue Prot. 9 Centralized 9Database Repository Care 9 9 Meter Data Only 9 Increasing Value 9 Complex Billing T&D Asset 9 Customer Meter Data + Meter Only Data Management of demand resources Outage 99Asset 9 Customer 9 T&D Asset Asset + + Customer Customer Meter Data+ Meter Data 9 Operational optimization Interval Meter Validation, Estimation, Editing Centralized Database Repository Meter Data Management May Evolve into Network Data Management Many of the characteristics cited in this paper that define an effective MDMS – e.g. scalability, flexible data schema, focus on synchronization of multiple data sources, and leveraging of underlying analytics – become critically important if the MDMS is to evolve into an NDMS. However, care needs to be taken to make sure we do not have a repeat of the CIS experience, where many CIS systems evolved into bloated, inflexible systems that performed functions well beyond what they were intended to do, and became extremely expensive to implement and adjust as needs changed. MDM systems need to be flexible, componentized, easy to integrate with, and continue to focus only on those things that the © Aclara Software Inc. 2008 20 MDMS can do best, leaving it to other systems to perform functions that they are better suited to perform. In this way, the MDMS is best positioned to evolve into one of the handful of critical utility IT systems - alongside the CIS, ERP, DMS etc. - and allow the company to best evolve into a 21st century utility. © Aclara Software Inc. 2008 21