Coolidge Connector Need and NTA Analysis Prepared for the Vermont Department of Public Service by Energy and Environmental Economics, Inc. 6/6/2008 Brian Horii Doug Allen Dr. CK Woo Eli Kollman Michele Chait Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Table of Contents INTRODUCTION ..............................................................................................3 BACKGROUND ..................................................................................................3 SUMMARY OF FINDINGS ....................................................................................4 PEAK DEMAND FORECAST .........................................................................8 BACKGROUND ..................................................................................................8 LOAD FORECAST PROCESS................................................................................9 INCORPORATING ENERGY EFFICIENCY ........................................................... 10 LOAD FORECAST RESULTS.............................................................................. 13 WEATHER SENSITIVITY OF VERMONT LOADS ................................................. 16 COMPARISON TO LA CAPRA AND ISO-NE FORECASTS ................... 18 LA CAPRA FORECAST ..................................................................................... 18 ISO-NE FORECAST ......................................................................................... 19 ISO-NE 2007 VERSUS 2008 CELT FORECAST ................................................ 22 ENERGY EFFICIENCY ................................................................................. 24 STATUS QUO EE ............................................................................................. 24 INCREMENTAL EE ........................................................................................... 24 MEASURE PENETRATIONS FOR INCREMENTAL EE .......................................... 26 POTENTIAL IMPACT OF THE INCREMENTAL EE ............................................... 26 DEMAND RESPONSE .................................................................................... 28 LA CAPRA ESTIMATE ...................................................................................... 28 CURRENT DEMAND RESPONSE ENROLLMENT ................................................. 29 DEMAND RESPONSE IN ISO-NE FORWARD CAPACITY MARKET..................... 30 E3 DR FORECAST ........................................................................................... 32 PEAKER UNITS IN VERMONT ................................................................... 34 GENERATION UNITS ........................................................................................ 34 RELIABILITY METRICS .................................................................................... 35 MODELING ASSUMPTIONS AND RESULTS ........................................................ 37 GENERATOR SIMULATION ............................................................................... 37 CAPACITY DISTRIBUTION APPROACH ............................................................. 40 INCREMENTAL PEAKING UNIT CAPACITY ....................................................... 42 Energy and Environmental Economics, Inc. Page 1 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 COMBINED HEAT AND POWER (CHP) .................................................... 43 ECONOMIC FEASIBILITY.................................................................................. 43 POWER AND GAS PRICE PROJECTIONS ............................................................ 43 CHP UNIT CONFIGURATION ........................................................................... 44 CHP COST EFFECTIVENESS UNDER ALTERNATE ELECTRICITY COST SCENARIOS ..................................................................................................... 45 TRANSMISSION NEED DATE ..................................................................... 47 COST OF DELAY OR DEFERRAL .............................................................. 50 BASE CASE ..................................................................................................... 50 DELAY CASES ................................................................................................. 50 E3’S DEFERRED CONSTRUCTION CASE ........................................................... 52 CONCLUSION ................................................................................................. 54 APPENDIX A: 1-IN-10 VERSUS 90TH PERCENTILE PEAK LOAD FORECAST ...................................................................................................... 55 APPENDIX B: PEAKING GENERATION ANALYSIS .............................. 57 APPENDIX C: GENERATOR SYSTEM CAPACITY DISTRIBUTION .. 58 Energy and Environmental Economics, Inc. Page 2 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Introduction Background On November 7, 2007, Vermont Transco, LLC, and Vermont Electric Company, Inc. (collectively, “VELCO”) and Central Vermont Public Service Corporation (“CVPS”, together with VELCO, “Petitioners”) filed an application with the Vermont Department of Public Service (“Vermont DPS”) for a Certificate of Public Good for the “Southern Loop Project.” The Petitioners claim that this project is needed to address reliability concerns arising in Vermont in the near future due to load growth in Vermont and the greater New England region. The filing indicates that under VELCO’s projections, the Vermont transmission system will not satisfy mandatory NERC reliability criteria in 2009 under certain contingencies, putting the system at risk for voltage collapses and thermal overloads that have the potential to affect service not only in Vermont, but in New York and the rest of New England as well. The VELCO transmission system relies heavily on imports from outside the state, through major ties with Quebec, New York, and New Hampshire, to serve Vermont load and the Connecticut Valley Electricity Company (“CVEC”) service territory in New Hampshire. Within the state, as much as 45% of the state’s power is carried on Line 340 from Vernon to Coolidge, even in the absence of outages. The heavy dependence on this line puts VELCO at risk for reliability problems if it fails, especially when either the Highgate Converter (200 MW of import from Canada) or the PV20 line (an underwater line providing imports from New York) is out of service. Load flow models prepared by VELCO show that their system is at risk at load levels as low as 1,155 MW in 2010, only 37 MW higher than the all-time peak on the VELCO system, which occurred on August 2, 2006. As load outside of Vermont grows, however, the threshold load level at which the system is at risk in the event of a contingency decreases. To solve these reliability issues, the Petitioners have proposed a second 345- Energy and Environmental Economics, Inc. Page 3 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 kV line running parallel to Line 340 from Vernon to Coolidge, along with a substation into which that line will connect and other upgrades to solve local reliability problems. La Capra Associates, Inc. was hired by VELCO to examine feasible nontransmission alternatives (“NTAs”) that could act as cost-effective alternatives to the Southern Loop Project in the appropriate time frame. In looking at these NTAs, La Capra performed analyses of the potential for energy efficiency, distributed generation, and demand response to achieve reductions in peak load and mitigate the reliability problem. The La Capra analysis identified four NTAs consisting of demand response, energy efficiency, and distributed generation schemes, as well as a fifth hybrid alternative that combined the proposed line with achievable energy efficiency. These alternatives were compared to VELCO’s proposal in terms of societal cost, capital cost, capacity provided, and risk level. Overall, the La Capra analysis concluded that the hybrid option, combining the aggressive pursuit of peak reductions through energy efficiency with the Southern Loop Project, had the most favorable tradeoff between risk and cost. Summary of Findings Energy and Environmental Economics, Inc. (“E3”) was hired by the Vermont Department of Public Service to “provide a detailed review of petitioner’s analysis of the feasibility and cost-effectiveness of targeted non-transmission alternatives to the transmission proposal.”1 E3 divided this analysis into four areas: (1) load forecasting; (2) distributed generation resources (including combined heat and power); (3) energy efficiency; and (4) other demand-side resources. In each category, E3 analyzed the information presented by VELCO and La Capra for any deficiencies, and produced alternate estimates of resource availability where appropriate. The report examines the effect of any alternate 1 Vermont Department of Public Service Request for Proposals Energy and Environmental Economics, Inc. Page 4 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 estimates of resource availability on the need for the project, and presents E3’s findings on the value of deferring the need for the line. E3’s biggest concern regarding the La Capra analysis is their load forecasting methodology and projected load levels. La Capra’s projections for future load are based on adjustments made to a DPS load forecast produced in the 2006 Draft Update to the Vermont Electricity Plan. When compared to forecasts produced independently by E3 and ISO-NE, the La Capra forecast predicts peak load to be 32 MW higher in 2011, when the proposed line is projected to be in service. Due to the relatively small Vermont load, a difference of 32 MW represents about two years of load growth. The difference in the load forecasts leads to a difference of a few years in the timing of the problem; this, in turn, affects the analysis of NTAs. La Capra estimates that demand resources will be able to reduce the summer peak load by 14 MW in 2010, based on adjustments to FERC and ISO-NE estimates of achievable peak reductions as a percentage of the summer peak. However, since the La Capra report was released, ISO-NE has conducted the first of their Forward Capacity Auctions as part of their Forward Capacity Market, which contracted resources for the 2010 – 2011 delivery year. E3 incorporates the results of that auction, and finds that the Forward Capacity Auction attracted less demand response for the summer of 2010 (after adjustments for response rate, location, etc.) than was assumed in the La Capra analysis. The analysis of generation resources centers on issues regarding the availability of peaking generation resources. In their analyses, VELCO and La Capra assumed that many of the smaller diesel and distilled fuel oil generators in Vermont would be unavailable to run during capacity emergencies, and so were “held in reserve for loss of McNeil,”2 the largest generator in Vermont. However, many of these same generators bid into the Forward Capacity Auction held in February, which indicates 2 Prefiled Testimony of Dean L. LaForest and Christopher Diebold, p. 44 at 8. Energy and Environmental Economics, Inc. Page 5 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 that they intend to be available at least until the summer of 2010. Though these units are unreliable when taken individually, E3 believes that the aggregate dependable capacity of the generators should be counted toward meeting peak demands and lowering the net peak loads. E3 estimates that the small generators plus the two planned 20MW Swanton peaker units can firm-up McNeil and provide an additional 45MW of dependable summer peak capacity with a high level of reliability. Under the assumptions used by La Capra in their analysis of combined heat and power (“CHP”) options, E3 agrees that there are no costeffective opportunities for CHP installations in the relevant study area. However, this result is driven by La Capra’s natural gas and electricity price forecasts, which project that nominal electricity prices will decrease while nominal gas prices remain at or near their current level. E3 expects that as the current long-term contracts with Entergy and Hydro Quebec begin to expire in 2012, electricity prices will increase as those contracts are replaced with new generation or market-priced power. While higher electricity prices could result in cost-effective CHP applications in the future, the amounts are small and the Coolidge Connector project would need to be deferred several years before one could confidently forecast CHP adoptions. With regards to the testimony of Mr. Ralph Roam on the costs of the project, E3 disagrees with Mr. Roam’s finding that delays of the project will necessarily result in high inflation in the real cost of the project. While E3 agrees with Mr. Roam’s finding that starting the project on time and delaying or suspending it mid-construction will lead to significant cost increases, E3 also believes that there is value to deferring the start date of the project. Though the deferral value calculated by E3 is small due to the high rate of inflation in the costs of materials, it implies that there is no increase in the present value cost of the project if it is deferred, which is contrary to Witness Roam’s findings because he did not examine a straight deferral case. Energy and Environmental Economics, Inc. Page 6 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 E3 conservatively estimates that using the load thresholds provided in the La Capra report the transmission project is needed in 2012, but could be pushed out to 2014 or 2015 with aggressive incremental energy efficiency activities. However, if the PSB determines that the Department’s lower peak load thresholds are the appropriate levels to use for planning purposes, then the transmission project cannot be deferred and should proceed on its current schedule. Department witness Smith discusses the reasoning and analysis behind the lowering of the 2011 peak load threshold from 1155MW to 945MW.. Energy and Environmental Economics, Inc. Page 7 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Peak Demand Forecast Background This section presents E3’s assessment of the reasonableness of the 1-in10 peak demand forecast used by La Capra to assess NTAs. Based on the forecast described below and supported by a comparison to the ISONE 2008 CELT report, E3 concludes that La Capra’s summer peak3 load forecast for planning purposes overstates the peak demand for Vermont and CVEC4 by approximately two years of demand growth. E3 has prepared a summer peak demand (MW) forecast for Vermont for 2008-2017. The forecast includes the CVEC load served by the Vermont transmission infrastructure, and is at the transmission level (i.e. includes losses). E3 used two forecast methods The first follows the 1-in-10 extreme weather methodology used by ISO-NE, and the second is a 90th percentile forecast that is more conservative (results in a higher peak forecast) than the 1-in-10 method. E3’s analysis utilizes the second (higher) forecast as a conservative assumption. In general terms, E3’s forecast is an ordinary least squares (OLS) regression using weather, a trend variable for the year, and dummy variables for day of the week and month of the year. The regression is performed using daily summer peak data from 1991 through 2007, and captures the increasing sensitivity of peak demands to weather that Vermont has been experiencing due to increased air conditioning usage in the state. The forecast also includes the demand reductions that are expected to occur due to continued energy efficiency activities by Efficiency Vermont (EVt) and the Burlington Electric Department (BED). 3 The southern portion of Vermont experiences its peak loads in the winter. However, the service territory as a whole has evolved over the past ten years from winter peaking to summer peaking. Accordingly, this forecast focuses on the summer peak demands for Vermont. 4 The transmission system serving the state of Vermont is a network of 345kVa and 115kVa VELCO infrastructure that also serves a small section of Central Valley Electric Cooperative (CVEC). Unless otherwise noted, any references to Vermont load include the portion of CVEC load served by VELCO. Energy and Environmental Economics, Inc. Page 8 of 58 Coolidge Connector Need and NTA Analysis Load Forecast Process Exhibit BKH-2 To make the net peak demand forecast, E3 used the following steps: 1. Apply ordinary least squares (OLS) to a sample of 799 daily peaks recorded on weekdays with average weighted temperature humidity index (WTHI) at or above 65 in the summer months of June – September during 1991-2007.5 2. Use the estimated regression to make a 1-in-10 forecast that reflects the 90th percentile of average WTHI recorded on Vermont’s annual summer peak days during 1965-2007, see Table 1 below. This forecast is a conditional forecast, aiming to capture an extreme weather event of 76.9 average WTHI6 expected to occur with a 10% probability. This 1-in-10 methodology is consistent with the ISONE Extreme Weather Peaks and La Capra 90/10 forecasts. The forecast also includes projected levels of energy efficiency activities at current levels (status quo EE) as describe below. Table 1: Annual maximum Average WTHI for Vermont 1965-2007 Summary statistic Average WTHI Number of observations 43 Mean 74.9 Variance 2.71 Median 74.9 90th percentile 76.9 3. Use the estimated regression to establish the distribution of a peak MW forecast based on the mean and variance of average WTHI shown in Table 1. The 90th percentile forecast is then defined as the MW level that an actual peak may exceed with only a 10% 5 The analysis uses 1991 through 2007 daily peak loads for Vermont plus the portion of CVEC load served by VELCO (VT+CVEC) from VELCO’s power accounting data. E3 removed four days where the reported daily peak loads were in excess of the annual peak values, and one day (6/11/99) where the peak load was 200 MW lower than the peak load on the prior or subsequent day . E3 also excluded all non-summer days, weekends, holidays (including days just prior to or following July 4th where a visual inspection of the data indicated extended holiday usage on those days), and days where the average WTHI was below 65 degrees. This left 799 peak demands on days with temperatures high enough to reflect weather sensitive usage. 6 Average Weighted Temperature Humidity Index (WTHI) is the average of the WTHI values for the 24 hourly values for the day. Each hourly WTHI value is the temperature humidity index weighted 59% for that day’s value, 29% for the hourly value in the prior day, and 12% for the hourly value two days prior. (Temperature Humidity Index = 0.5* dry bulb temperature + 0.3 * wet bulb temperature + 15) Energy and Environmental Economics, Inc. Page 9 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 probability. The 90th percentile forecast recognizes multiple sources of forecast uncertainty, including the regression’s coefficient estimates, weather, and forecast error. Like the 1-in10 forecast, the 90th percentile forecast also includes projected levels of energy efficiency activities at current levels (status quo EE) as described below. Table 2 presents the two peak demand forecasts. The table shows that the 90th percentile forecast is slightly higher, and therefore more conservative for purposes of estimating transmission capacity needs. The two forecasts are very similar, however, indicating that the forecasts are insensitive to how the forecasting process accounts for uncertainty. Note that the forecasts are based on VELCO power accounting data that does not include losses. The actual peak load, including losses, would be about 12.4MW higher than what is shown in the table7. Table 2: Peak demand forecasts (net of status quo energy efficiency, but unadjusted for losses) Year Peak demand 1-in-10 weather Peak demand 90th percentile forecast forecast 2008 1124.24 1126.02 2009 1135.88 1137.50 2010 1148.34 1149.84 2011 1161.49 1162.90 2012 1175.62 1176.94 2013 1190.78 1192.02 2014 1206.15 1207.33 2015 1221.73 1222.86 2016 1237.51 1238.61 2017 1253.58 1254.65 Losses would add 12.4MW to each year. Incorporating Energy Efficiency E3’s load forecast reflects load reductions associated with historic and projected energy efficiency (EE) activities. Figure 1 below reports the estimated MW size of EE programs for the historic period of 1995-2007 and forecast period of 2008-2017. The historic EE data is used to inform 7 12.4 MW is the difference between the Vermont peak load value for 2007 quoted in A.DPS:PET.1-4, less the 2007 peak load value from A.DPS:PET.1-3. Energy and Environmental Economics, Inc. Page 10 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 the regression model (it is a right hand side variable) and the forecast period EE data is included in the forecast, scaled by the regression coefficient. To be applicable to the forecasting of summer peak MW, the EE values should reflect only the reductions from EE in place at the beginning of summer, rather than the reductions from EE in place at the end of the calendar year. Accordingly, E3 has adjusted the historical reported values provided by the Vermont utilities and Efficiency Vermont to reflect this mid-year convention by including only 50% of the EE in the year of installation.8 In the year following the installation year, E3 includes 100% of the EE, and then adjusts the load reductions by applying the annual decay factors developed by Optimal Energy for subsequent years. For the forecast status quo EE, E3 started with Optimal Energy’s estimates, removed 2007 (as that is part of the historical EE activities) and adjusted the EE reductions for the mid-year convention and decay factors. 8 Table FN1 uses a stylized three year EE program to illustrate how E3 has implemented the mid-year EE adjustment described above. Columns A through D show “end of year” values that are reported by entities like Efficiency Vermont. The end of year values show the peak reductions attributable to the EE measures installed in years 1 through 3, regardless of whether the measures were in place at the time of the summer peak. Each column follows a tranche of EE measures. The peak load reductions are highest in the year of installation, then decline each year due to performance decay and measure persistence. The summer peak reductions derived by the mid-year adjustment are shown in columns E through H. The adjustments are only applied in the year when the measures are installed. The adjusted values are highlighted in bold. For example, looking at line 2 shows that the end of year value for measures installed in year 2 in column B is 80MW, and the mid-year corrected value in column F is only 40MW (half of the end-of-year value). After year 2, the measures installed in year 2 are not adjusted. This is clearly illustrated by rows 4 and 5 where the total mid-year values in column H and the end of year values in column D are the same. Table FN1 Stylized Mid-Year EE Adjustment A 1 2 3 4 5 Yr1 Yr2 Yr3 Yr4 Yr5 B C End of Year Values Yr1 Yr2 Yr3 100 95 80 90 80 110 89 78 105 85 75 100 D Total 100 175 280 272 260 E F G H Mid-Year Values Yr1 Yr2 Yr3 Total 50 50 95 40 135 90 80 55 225 89 78 105 272 85 75 100 260 Energy and Environmental Economics, Inc. Page 11 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 1: Status Quo EE Inputs (reductions counted in the forecast are lower) Year 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 180 Summer Peak Reductions (MW) 160 140 120 100 80 60 40 20 2015 2013 2011 2009 2007 2005 2003 2001 1999 1997 1995 0 EE Smr Peak Reduction (MW) 0.9 2.0 2.6 3.5 4.1 5.2 8.7 13.6 19.6 27.1 35.3 44.6 56.2 70.3 84.4 97.3 109.2 119.7 128.7 137.4 145.8 153.9 In the regression analysis, the Status Quo EE input has a coefficient of 0.69. This indicates that each 1.0 MW of Status Quo EE reduces the summer peak by 0.69 MW. Applying that factor to EE values from Figure 1 yields the de-rated EE that is included in the E3 forecasts. The de-rated values are indicated by the triangle marker in Figure 2. Figure 2: De-rated EE reductions included in E3 peak forecast Unadjusted EE Derated EE 2001 9 6 2002 14 9 2003 20 14 2004 27 19 2005 35 24 2006 45 31 2007 56 39 2008 70 49 2009 84 58 2010 97 67 2011 109 75 2012 120 83 2013 129 89 2014 137 95 2015 146 101 2016 154 106 2017 162 112 180 160 22 23 Total EE in Place Mid-Year (MW) Derated EE included in E3 Forecast 120 100 80 61.0% 60.3% 60 40 20 Energy and Environmental Economics, Inc. 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 0 1995 Summer Peak MW 140 Page 12 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Using de-rated EE peak reductions results in a higher load forecast than would be obtained from using an unadjusted Status Quo EE forecast. E3 has not investigated the reasons for the difference between EE reported and verified savings, and the regression results, but believes that it is appropriate to use de-rated EE peak reductions. Possible explanations for the differences include the fact that the peak reductions verified and reported by Efficiency Vermont are based on the timing of the ISO-NE summer peak. Vermont experiences its peak demand earlier in the day than ISO-NE, so the peak reduction of residential EE measures in particular, would likely be lower at the time of the Vermont peak. The less than 100% regression coefficient factor may also reflect the lower contribution to peak load reductions provided by some types of DR measures at times of extreme weather9 . Load Forecast Results E3’s summer peak load forecast is listed in Table 2 above and shown graphically in Figure 3 below. The 2008 forecast is only slightly above the 2006 recorded peak demand, which is likely the result of the ramp-up of energy efficiency activities since 2006 and the extreme weather experienced at the time of the 2006 summer peak. 9 A two-stage air conditioner is an example of a device that would provide large energy savings and reasonable load reductions during time of moderate temperatures, but would have higher peak demand during times of extreme weather. Energy and Environmental Economics, Inc. Page 13 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 3: Recorded and Forecast Vermont + CVEC Summer Peak Demand (with Status Quo EE and losses) Summer Peak MW (VT + CVEC) 1,300 1,250 1,200 1,150 1,100 1,050 1,000 950 900 850 Load with losses (VT+CVEC) 2016 2014 2012 2010 2008 2006 2004 2002 2000 1998 1996 1994 1992 1990 800 E3 90th %tile Forecast w/ Losses & SQ EE Table 3 reports E3’s daily peak demand regression model. The data sample contains 799 daily observations of peak MW recorded on hot summer (June – September) weekdays with average weighted temperature humidity index (WTHI) at or above 65 in 1991-2007.10 E3 tested other weather metrics, such as the maximum WTHI value for each day, but found that the average WTHI provided the best statistical fit and provides the best indication of a multiple day heat storm. The regression is statistically significant (as shown by the F-statistic of 1163), has high adjusted R2 of 94%, and has statistically significant coefficient estimates at the 5% level, except for those of the June, July and Wednesday binary indicators. The interpretation of the coefficient estimates is as follows: The 06/11/09 observation is 598.65 MW, well below the prior day’s 806.62 MW. As a result, it is deleted from the estimation sample by reason of data error. 10 Energy and Environmental Economics, Inc. Page 14 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 The coefficient estimates for the month-of-year and day-of-week binary indicators show that August Mondays tend to have the highest daily peak demands when compared to other weekdays in the summer. The coefficient estimate for Mid-year EE is -0.69 and has a standard error of 0.08. The estimate’s 95% confidence interval is (-0.53, 0.85) (= -0.69 ± 1.96 * 0.08), rejecting the hypothesis that each midyear EE MW reduces daily peaks by 1 MW. E3’s forecast counts 69% of the estimated mid-year EE for peak reductions. For the average WTHI of 75, Vermont’s peak demand will grow at 19.7 (= -43.9 + 0.847 * 75) MW per year. For year 2007 (t = 17), the weather response is 19.8 MW (= 5.39 + 0.847 * 17) for each increment of average WTHI. This weather response grows at a rate of 0.847 MW per year. Table 3: Peak demand regression for 799 summer (June-September) weekdays with average WTHI ≥ 65 in 1991-2007 Variable Coefficient estimate with standard errors in ( ) p-value Intercept 351.0 (35.76) <.0001 = 1 if June, 0 otherwise -5.02 (2.67) 0.0607 = 1 if July, 0 otherwise -2.99 (2.53) 0.2369 = 1 if August, 0 otherwise 13.90 (2.54) <.0001 = 1 if Monday, 0 otherwise 7.78 (2.16) 0.0003 = 1 if Tuesday, 0 otherwise 5.96 (2.09) 0.0045 = 1 if Wednesday, 0 otherwise 2.46 (2.13) 0.2476 = 1 if Thursday, 0 otherwise 5.54 (2.15) 0.0101 EE size (MW) -0.692 (0.08) <.0001 Time trend t -43.86 (3.40) <.0001 Average WTHI 5.39 (0.52) <.0001 Time trend * Average WTHI 0.847 (0.049) <.0001 F-statistic with (11, 787) degrees of freedom 1164 < .0001 Adjusted R2 0.9413 Mean square error 347.2 Energy and Environmental Economics, Inc. Page 15 of 58 Coolidge Connector Need and NTA Analysis E3’s regression equation contains the variable (Time trend * Average WTHI). E3 has included that variable in the model specification to reflect the increasing importance of weather sensitive loads such as air conditioners in driving the Vermont system peak. Figure 4 shows the summer load factor for Vermont plus CVEC from 1991 through 2007, based on VELCO power accounting data. The figure shows a clear downward trend in load factors, which indicates that the loads are becoming more peaked, which is consistent with weather sensitive end uses. Figure 4: Annual Summer Peak Load Factors 85% Annual Load Factor (Smr Peak) Weather Sensitivity of Vermont loads Exhibit BKH-2 80% 75% 70% 65% y = -0.0065x + 13.762 R2 = 0.6905 60% 55% 50% 1990 1995 2000 2005 2010 This increasing level of weather sensitive loads is further confirmed by an examination of the daily summer peak loads and temperatures. The two figures below compare the sensitivity of Vermont plus CVEC summer loads in 1996 and 2006. The trend line in each figure shows the average relationship between average WTHI and summer peak demands for summer weekdays with average WTHI temperatures above 65 degrees. The steeper the slope, the more the loads vary with weather. The figures clearly demonstrate that over the past ten years, the Vermont plus CVEC loads have become about twice as sensitive to weather. Energy and Environmental Economics, Inc. Page 16 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 5: Vermont plus CVEC daily peak response to temperature (1996) 1200 y = 9.9359x + 126.39 R2 = 0.6874 1100 Daily Peak MW 1000 Weekday Load Weekend Load 900 Weekday above 65 Linear (Weekday above 65) 800 700 600 50 55 60 65 70 75 Avg WTHI Figure 6: Vermont plus CVEC daily peak response to temperature (2006) 1200 y = 19.436x - 414.52 R2 = 0.8357 1100 Daily Peak MW 1000 Weekday Load Weekend Load 900 Weekday above 65 Linear (Weekday above 65) 800 700 600 50 55 60 65 70 75 80 Avg WTHI Energy and Environmental Economics, Inc. Page 17 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Comparison to La Capra and ISO-NE Forecasts Figure 7 compares E3’s 90th percentile forecast with La Capra’s 90/10 La Capra Forecast forecast. Both forecasts reflect status quo energy efficiency reductions. The E3 forecast is lower than the La Capra forecast by about two years of growth, and would be even lower had E3 included the full EE reductions that La Capra included in their forecast. Figure 7: Comparison of E3 and La Capra forecasts11 1,250 1,200 1,150 1,100 1,050 1,000 950 900 850 2016 2014 2012 2010 2008 2006 2004 2002 2000 1998 1996 1994 1992 800 1990 Summer Peak MW (VT + CVEC) 1,300 Load with losses (VT+CVEC) E3 90th %tile Forecast w/ Losses & SQ EE La Capra 90/10 w/ SQ EE E3’s forecast includes less that half of the EE peak reduction used by La Capra. Figure 8 shows the Status Quo EE included in the La Capra forecast (diamond marker),12 the Status Quo EE numbers used as inputs to the E3 load forecast (square marker), and the de-rated EE reflected in the peak loads estimated by the E3 forecast (triangle marker). Absent E3’s conservative treatment of EE reductions, E3’s forecast would easily 11 La Capra 90/10 w/ SQ EE is from Reliability Needs Assessment table on page 25 of RSH-2, "Total VT." 12 The values are the difference between the La Capra 90/10 with CVEC forecasts that (a) include SQ EE and (b) exclude future EE. See Exhibit Petitioners RSH-2, p. 24. Energy and Environmental Economics, Inc. Page 18 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 be below La Capra’s forecast by an additional two years’ growth in 2016, shown by the 30 MW gap between the middle and lowest lines. Figure 8: EE Peak Reductions Included in the Forecasts for Status Quo EE installed in 2007 or later (reductions at the generator level) 160 Summer Peak Reductions (MW) La Capra SQ EE 140 E3 SQ EE Inputs 120 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Derated SQ EE in E3 Forecast 100 80 60 40 La Capra SQ EE 14.0 33.0 51.0 69.0 87.0 100.0 112.0 124.0 136.0 147.0 Derated E3 SQ EE in SQ EE E3 Inputs Forecast 11.6 8.0 25.7 17.7 39.8 27.5 52.8 36.4 64.6 44.6 75.2 51.9 84.1 58.0 92.8 64.1 101.2 69.9 109.3 75.4 20 ISO-NE Forecast 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 0 A comparison of the E3 and La Capra forecasts with ISO-NE’s most recent forecast for Vermont plus CVEC from their 2008 CELT is shown below. At first glance, the ISO-NE forecast seems to confirm the La Capra forecast. However, the ISO-NE forecast reflects a level of EE activity consistent with past levels, and does not reflect the higher status quo EE activity levels assumed in the La Capra analysis. Energy and Environmental Economics, Inc. Page 19 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 9: Comparison of E3 and La Capra Forecasts with Unadjusted ISONE 2008 CELT 1,250 1,200 1,150 1,100 1,050 1,000 950 900 850 2016 2014 2012 2010 2008 2006 2004 2002 2000 1998 1996 1994 1992 800 1990 Summer Peak MW (VT + CVEC) 1,300 Load with losses (VT+CVEC) E3 90th %tile Forecast w/ Losses & SQ EE ISO NE 2008 CELT 1-in-10 Weather (Tab 7) La Capra 90/10 w/ SQ EE Table 4: E3, La Capra, and Unadjusted ISO-NE 2008 CELT Forecasts (MW) Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 E3 Forecast SQ EE and losses 1,138 1,150 1,162 1,175 1,189 1,204 1,220 1,235 1,251 La Capra SQ EE 1,161 1,180 1,192 1,207 1,224 1,240 1,251 1,260 1,266 ISO-NE (unadjusted for SQ EE) 1,140 1,165 1,185 1,200 1,220 1,235 1,245 1,255 1,265 The ISO-NE ignores energy efficiency in their analysis, allowing it to affect projections only inasmuch as it has affected peak loads in the recorded data. Thus, the ISO-NE forecast incorporates a trend of energy efficiency, which assumes that energy efficiency efforts continue along the trend seen in the past. To make an apples-to-apples comparison of the ISO-NE forecast to those prepared by La Capra and E3, E3 adjusted the ISO-NE forecast to reflect Energy and Environmental Economics, Inc. Page 20 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 status quo EE activity levels. The adjustment involves first determining the amount of EE implicitly included in the ISO-NE forecast, and then adjusting the ISO-NE forecast downward by the difference between the SQ EE forecast and the implicit EE in the ISO-NE forecast. To estimate the amount of energy efficiency captured in the ISO-NE forecast, E3 trended the historical mid-year EE values for the years 1991 through 2007. E3 used the derated EE values because E3’s regression analysis indicates that the de-rated values reflect the EE reductions that are observed in the recorded peak loads. E3’s estimate of the de-rated peak reductions from EE are indicated by the pink boxes. The trended amounts that E3 believes are implicitly included in the ISO-NE forecast are shown in by the dashed line with the diamond markers, and the difference is indicated by the green triangles. 120 100 80 Trend of Derated Pre-2008 EE (MW) 60 Derated average EE in E3 Forecast Difference 40 20 2016 2014 2012 2010 2008 2006 2004 2002 2000 EE Summer Peak Reductions (MW) Figure 10: Comparison of E3 EE Forecast, and EE Implicit in ISO-NE Forecast The ISO-NE forecast net of status quo energy efficiency is derived by subtracting the difference from the ISO-NE 90/10 forecast. This is shown as the solid circles in Figure 11 below. The figure clearly demonstrates that after adjusting for status quo EE, the E3 and ISO-NE peak forecasts are very close. Table 6 shows the projected load levels under the three forecasts, and the MW difference relative to the E3 forecast for both the La Capra and ISO-NE forecasts. Energy and Environmental Economics, Inc. Page 21 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 11: A Comparison of EE included in the E3 and ISO-NE models Summer Peak MW (VT + CVEC) 1,300 1,250 1,200 1,150 1,100 1,050 1,000 950 900 850 2016 2014 2012 2010 2008 2006 2004 2002 2000 1998 1996 1994 1992 1990 800 Load with losses (VT+CVEC) E3 90th %tile Forecast w/ Losses & SQ EE ISO NE 2008 CELT 1-in-10 Weather (Tab 7) La Capra 90/10 w/ SQ EE ISO-NE less New EE above the trend Table 5: 90/10 forecasts from E3, La Capra, and ISO-NE (MW) Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 ISO-NE 2007 versus 2008 CELT forecast E3 Forecast 1138 1150 1162 1175 1189 1204 1220 1235 1251 La Capra - SQ EE ISO-NE 2008 net of (Difference) EE (Difference) 1161 (+23) 1132 (-6) 1180 (+30) 1152 (+2) 1192 (+30) 1168 (+6) 1207 (+32) 1180 (+5) 1224 (+35) 1198 (+9) 1240 (+36) 1212 (+7) 1251 (+31) 1221 (+1) 1260 (+25) 1230 (-5) 1266 (+15) 1239 (-12) The testimony of Dean L. LaForest and Christopher Diebold discusses VELCO’s belief that the ISO-NE 2007 forecast “is lower than it should be by at least 30 to 35 MW per year over the first few years” (p. 37 at 16). This conclusion is based on the fact that the ISO-NE system peak load in 2007 corresponded to the forecasted 30/70 peak value forecasted in the ISO-NE 2007 CELT, which LaForest and Diebold claim indicates that the weather in New England at the time was 30/70 weather. ISONE’s 30/70 forecast for Vermont’s 2007 peak load was 1045 MW, 35 Energy and Environmental Economics, Inc. Page 22 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 MW lower than VELCO’s recorded value of 1080 MW, prompting VELCO’s concern. However, in their response to DPS:PET.2-24.a, VELCO states that the WTHI on the day of the 2007 Summer Peak “fell just below the expected value of 80.1.” Weather data from the Vermont DPS indicates that the WTHI was actually above the expected value, not below, at roughly 80.2. Rather than corresponding to VELCO’s assumption that the weather during the 2007 peak was indicative of 30/70 weather, a WTHI of 80.2 indicates that the weather was closer to 50/50 or even 60/40 weather. For this weather, ISO-NE forecasted a Vermont peak load of between 1070 MW and 1075 MW. Rather than being 35 MW low, as LaForest and Diebold claimed, ISO-NE’s forecast appears to have been low by about 5-10 MW. The fact that ISO-NE and E3, using more recent data, independently generated consistent forecasts indicates that the E3 forecast may be more accurate than the forecast used by La Capra. This is supported by the fact that La Capra did not itself generate a forecast, but rather made a series of adjustments to an existing forecast. The existing forecast that they modified was developed in mid-2006, and appears to be based on data that did not include 2006. Thus, E3 believes that the La Capra load forecast is too high and that making decisions based on this load forecast would be inappropriate. Energy and Environmental Economics, Inc. Page 23 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Energy Efficiency VELCO retained Optimal Energy, Inc. to develop and coordinate the study of energy efficiency to be included in the analysis of NTAs for the Southern Loop and Coolidge Connector. As part of this work, Optimal Energy developed forecasts of (1) the summer peak demand reductions that would occur under status quo EE funding comparable to Efficiency Vermont 2008 levels, and (2) the incremental cost-effective EE peak reductions that could be achieved in the Vermont Central and Northwest zones. Status Quo EE E3 reviewed Optimal Energy’s estimates of status quo energy efficiency reductions and considers them to be reasonable estimates of summer peak reductions as reported for the tracking of Efficiency Vermont goals. Note that this assessment is based on the kW claims, and not the actual impact on the Vermont peak. As discussed above, the actual reduction reflected in the load forecast is about 70% of the EE input values. E3 assessed the reasonableness of the status quo estimates by comparing the forecast budgets and reductions to past EE budgets and reported program achievements. Using the 2005 through 2007 average summer kW reduction per dollar of expenditures, E3’s estimate of annual summer peak reductions for reporting purposes is within one percent of Optimal Energy’s forecast. Using the 2007 average summer kW reduction per dollar of expenditure, E3’s estimate is within 10% of the Optimal Energy summer demand reduction estimate. Given the uncertainly surrounding future budgets, potential declining effectiveness in attaining summer peak kW reductions, and variations in decay rates, E3 does not consider this difference in peak kW estimates to be an issue. Incremental EE Optimal Energy also estimated an additional achievable potential of 88MW of cost-effective EE summer peak reductions in Vermont’s Central and Northwest regions. Optimal Energy estimates that it would cost $256 million (2007 dollars) over ten years to attain this additional Energy and Environmental Economics, Inc. Page 24 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 88MW. 13 This would result in net savings for customers, however, as the EE measures are cost-effective. While it is true that average rates could increase if the additional EE were implemented, the average customer’s energy expenditures would likely decrease. Table 6 compares Optimal Energy’s estimate of achievable potential for Vermont with estimates for New Jersey, Connecticut, and California, three states with relatively aggressive efficiency programs. The potential MW demand savings projected by Optimal Energy for VT as a percentage of the forecast load are higher (but not dramatically so) than those projected for NJ and CA, while they are lower than the estimates for CT. This is the case despite Optimal Energy’s somewhat conservative assumptions regarding market penetration (see next section). Table 6: EE Potential Estimates in High EE Jurisdictions VT NJ CT CA Author Optimal KEMA14 GDS15 Itron16 Year 2016 2020 2012 2020 MW Savings 8.0% 5.9% 13.4% 7.0% 12.5% 3.4% GWh Savings 4.5% 10.5% Savings as percent of forecasted load 13 Prefiled Testimony of Jonathan Kleinman, Page 7 at 1. KEMA (2004). “New Jersey Energy Efficiency and Distributed Generation Market Assessment.” Final Report to Rutgers University Center for Energy, Economic and Environmental Policy. August 2004. 14 GDS Associates (2004). “Independent Assessment of Conservation and Energy Efficiency Potential for Connecticut and the Southwest Connecticut Region.” Prepared for the Energy Conservation Management Board. June 2004. 15 Itron (2007). “Assistance in Updating the Energy Efficiency Savings Goals for 2012 and Beyond. Task A4.1 Final Report: Scenario Analysis to Support Updates to the CPUC Savings Goals.” Prepared for the California Public Utility Commission. March, 24, 2007. 16 Energy and Environmental Economics, Inc. Page 25 of 58 Coolidge Connector Need and NTA Analysis Measure Penetrations for Incremental EE Exhibit BKH-2 For the Commercial and Industrial sectors, the Optimal Energy study assumed penetration rates of 50% for lighting and 25-65% for other end uses. The report acknowledges that these penetration rates are lower than those presented in other studies and reports, which have targeted penetration goals as high as 85%. The Northwest Power and Conservation Council (NWPCC) calls for a target penetration of 85% for retrofit measures and 65% for lost-opportunity measures over 12 years. In supporting these targets, the NWPCC cites a targeted Hood River Conservation Project which achieved 83% penetration in two years with 100% incentives. In California, the IOU’s have targets set based on an 80% penetration rate over 10 years. In a 2007 study supporting the CPUC’s update of energy efficiency goals, Itron assumed penetration rates similar to those used by Optimal Energy, though Itron’s low end penetration rates were somewhat higher. For the base case, Itron assumed penetration rates of 40-45% for the lighting, commercial cooling and commercial refrigeration measures that represented the majority of the potential savings. Penetration for other end uses was estimated in the 50-60% range. For the more aggressive “Full Market” scenario, the penetration rates were 60-70% for lighting, commercial cooling and commercial refrigeration, and 70-80% for other end uses. Potential Impact of the Incremental EE If one assumes that the annual incremental peak kW reductions could be obtained in proportion to the status quo EE reductions (that is, no significant additional ramp-up lag for the incremental EE), 46MW would be achieved by 2011. Factoring in the EE de-rating factor of 69% from E3’s regression analysis would reduce the reductions to about 32MW. The annual peak reductions from incremental EE are shown in Table 7. Energy and Environmental Economics, Inc. Page 26 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Table 7: Annual Incremental EE Peak Impacts (MW) A Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 B C Expected peak Status Quo EE Incremental EE impacts Inputs (Col A * 88/107.7) (Col B * 69%) 8.7 7.1 4.9 25.3 20.7 14.3 41.2 33.6 23.2 56.6 46.3 31.9 71.6 58.5 40.4 81.2 66.3 45.8 90.3 73.7 50.9 99.1 81.0 55.9 107.7 88.0 60.7 88MW is the incremental EE potential from Optimal Energy. While this is a significant reduction in peak demand, E3 is concerned that it is not clear that Efficiency Vermont could scale up implementation efforts quickly enough to attain this level of savings. This would require a doubling of the expected Efficiency Vermont effort in 2008. In addition, E3 is concerned that it will become increasingly difficult to achieve the full level of peak reductions as saturation increases. Program implementations typically follow an S-curve where adoptions are slow to start, then hit a period of sustained success before hitting the period of diminishing returns. It is unclear when those diminishing returns would significantly hamper the peak kW reduction effort. Because of these concerns, E3 has included incremental EE as the last load reduction option in its forecast of peak loads. To the extent that the Coolidge Connector can be deferred by other means, Vermont will gain more experience and confidence with implementing sustained high effort EE programs in the state. This would inform future expectations of the ability of Vermont to cost-effectively attain the incremental EE reductions and further defer the need for the Coolidge Connector. Energy and Environmental Economics, Inc. Page 27 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Demand Response La Capra Estimate The discussion of demand response (DR) in the VELCO filing is found mostly within Appendix B to the La Capra Report. The information presented there is derived primarily from two sources: (1) “Assessment of Demand Response and Advance Metering,” a FERC staff report from August 2006; and (2) ISO-NE’s 2004 Demand Response Evaluation Report. La Capra made adjustments to the DR estimates based on two assumptions. First, they assumed that most residential DR would come from a reduction in air conditioner use. Since Vermont has very little penetration of residential air conditioners, they assumed that residential DR would not be available in Vermont. Second, they excluded all generation DR resources, deeming them ineligible for consideration based on a belief that DR generators would not be able to obtain the air permits necessary to run when called upon. La Capra estimated that generation made up 50% of the non-residential DR for each of the studies. Applying these assumptions to the 6% DR estimate made by FERC and the 4.3% estimate made by ISO-NE results in estimates for available DR in Vermont of 2.4% and 1.1% for the FERC and ISO-NE studies, respectively.17 La Capra concluded that a DR potential of 2%, or 21 MW, represented a reasonable estimate for the VT load area. This 21 MW number was then shared to the different zones proportional to load, reducing it to 17 MW of DR in areas where load reductions would decrease the loading on Line 340. Finally, the 17 MW was scaled by the Subscriber Performance Index, which represents the proportion of load enrolled that typically responds to calls to reduce demand. Overall, this resulted in an estimate of 14 MW of DR for the relevant areas of VT. 17 The FERC study assumes 20% residential DR, so La Capra counted 40% of the FERC estimate (half of the 80% non-residential is assumed to be emergency generation) for 40% * 6% = 2.4% available DR in Vermont. The ISO-NE study assumes 50% residential DR, so La Capra counted 25% of the ISO-NE estimate for 25% * 4.3% = 1.1%. Energy and Environmental Economics, Inc. Page 28 of 58 Coolidge Connector Need and NTA Analysis Current Demand Response Enrollment Exhibit BKH-2 The following table shows the demand response resources enrolled with ISO-NE at month’s end for the past year, classified as “Ready to Respond” or “Approved.”18 Table 8: VT Demand Response Resources Enrolled with ISO-NE Ready to Respond Approved March 30, 2007 32.6 MW 0.2 MW May 1, 2007 26.6 MW 0.0 MW June 1, 2007 27.2 MW 0.0 MW July 2, 2007 22.4 MW 2.6 MW July 31, 2007 22.5 MW 5.0 MW August 31, 2007 29.7 MW 0.2 MW October 1, 2007 32.1 MW 0.0 MW November 1, 2007 33.4 MW 0.0 MW November 30, 2007 54.6 MW 8.1 MW December 31, 2007 62.6 MW 0.5 MW January 31, 2008 64.3 MW 0.7 MW February 29, 2008 60.3 MW 0.0 MW March 31, 2008 41.4 MW 0.5 MW Though there is a significant amount of load available to respond in the winter months, this amount drops sharply (by over 60% from the maximum to the minimum) during the summer months. This is consistent with the discussion of Demand Response in Appendix B to the La Capra report on page 2. This section cites the 34 MW that ISO-NE listed as either “ready to respond” or “approved” as of December 31, 2006, but points out that the 34 MW value “represents both generating and non-generating assets and assets that are likely capable of responding in only one season such as ski area snowmaking.” These single season “Ready to Respond” means the registration process is complete and the resource is eligible to participate in an Event, while “Approved” means the application for registration has been approved by ISO-NE. Source: http://iso-ne.com/genrtion_resrcs/dr/stats/enroll_sum/index.html 18 Energy and Environmental Economics, Inc. Page 29 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 assets are likely a reason for the decrease in available demand response between the winter and summer months. However, ISO-NE has also recently changed the way that they calculate payments in their Day-Ahead Load Response Program (DALRP), which also may have caused a significant decline in participation. In February, ISO-NE filed a proposal with the Federal Energy Regulatory Commission (FERC) to raise the price threshold at which demand response providers were paid to reduce load, which was approved by FERC on April 4, 2008. Before this proposal, the Locational Marginal Prices were so high relative to the baseline that demand response resources could collect money for reducing load 80% of the time, which was preventing the development of accurate baselines from which to measure the reductions. ISO-NE believed that this was resulting in substantial payments for load reductions that weren’t actually happening. Since ISO-NE filed for this change, there has been a significant decline in the participation of demand response resources in the DALRP, but as this corresponds with the time of year during which there is normally a decline in enrollment anyway, it is hard to determine the extent to which the decline is due to seasonality of demand response resources and the extent to which it is due to the new payment scheme. Demand Response in ISO-NE Forward Capacity Market The first Forward Capacity Auction (FCA) for the ISO-NE Forward Capacity Market (FCM) was held in February 2008, and is described in a presentation made by the ISO-NE Demand Resources Department to the Demand Resources Group on April 2, 2008.19 This presentation shows the resources bid into the first FCA by resource type. The resources for Vermont are shown below: 19 http://www.isone.com/committees/comm_wkgrps/mrkts_comm/dr_wkgrp/mtrls/2008/apr2200 8/fca-1_results_measure_type_040708.ppt Energy and Environmental Economics, Inc. Page 30 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Table 9: VT Demand Resources Cleared in the 2010/2011 FCA Demand Resources in VT Resource Type Cleared MW – Existing20 Cleared MW – New Total MW Dist Generation – Fossil Fuel 0 0 0 Dist Generation - Renewable 0 0 0 Energy Efficiency 2.0 55.7 57.7 Real-Time Demand Response and Load Management 8.1 15.6 23.6 Real-time Emergency Generation 20.3 0 20.3 Total 30.4 71.3 101.6 The resources described above are committed for the 2010 – 2011 Power Year. When considering the DR resources that could serve to defer the need for the given project, the energy efficiency measures described above should be excluded, because they are accounted for separately in the analysis.21 The concern of the La Capra report discussed above regarding seasonal resources is not relevant for the forward capacity market because the demand response resources offer capacity bids for a specific month, and all DR resources are required to “demonstrate or certify that [they have] the capability to deliver the targeted load reduction”22. Though the resources above have bid into the capacity market in ISONE, there is no guarantee that they will drop load when called upon. Payments in the Forward Capacity Market are based on performance under shortage events, but the only penalty that the FCM has for a failure to respond is forfeiture of payment for capacity the resource did not 20 Demand Response resources that are commercial by March 1, 2007 or are under construction and either (1) have an in-service date before February 1, 2009, (2) are required by a contract, or (3) elect to be treated as such are considered “Existing.” All other Demand Resources are treated as “New” DR resources. 21 It is interesting to note, however, that the La Capra report assumes that there will be 69 MW of Energy Efficiency under the status quo, which seems reasonable given the amount bid into the first FCM auction. 22 A.DPS:PET.1-111 Energy and Environmental Economics, Inc. Page 31 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 provide. As this is the first Forward Capacity Auction that has taken place the full three years ahead of the time period for which it is contracting resources, there is no experience with the auction or the response rate of enrolled resources. In performing reliability assessments, ISO-NE does not include real-time emergency generation resources, which account for about 20 of the 44 MW of non-EE demand resources. Real-time emergency generation is not included because it is called only under Action 12 of ISO-NE Operating Procedure 4, and OP-4 actions are only implemented in the case of a real time capacity deficiency.23 Consistent with ISO-NE’s treatment, E3 separated real-time emergency generators from the other demand response resources for the remainder of the analysis. E3 DR Forecast From 2010 forward, E3 considers it most appropriate to use the results from the FCM (excluding energy efficiency) now that they are available. This results in an available demand response potential in Vermont of about 24 MW in 2010. When performing reliability assessments, ISONE derates demand response resources in two steps. First, they are scaled down by a factor of 1.143 to accurately model their system benefits, and then they are derated by an additional 11% to reflect the historic failure to respond. Applying this methodology to the 24 MW listed above results in 18.4 MW of demand response resource availability for the VELCO system in 2010. E3 further derated the DR resources to reflect their geographical location in the state, as load reductions have different levels of effectiveness depending on where they are located. Both the load forecast and the peak load thresholds used to determine the need for the project reflect the total Vermont plus CVEC load levels. VELCO has stated that load reductions in the Southern and Northern regions would provide little or no load relief for the Coolidge Connector project. Therefore, to the 23 Direct Testimony of Stephen Rourke, FERC Docket ER08-633-000, page 9 at 4. Energy and Environmental Economics, Inc. Page 32 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 extent that the share of load reductions from the Southern and Northern load regions is greater than the share of total load in those regions, the DR reductions need to be adjusted downward. E3 assumed that 50% of DR resources were located in the Southern and Northern regions, as compared to 21% of the total Vermont load. Accordingly, E3 adjusted the DR resources counted for load reduction to 63% of the total. This results in a total DR resource availability of 11.6 MW for the purposes of E3’s analysis. Energy and Environmental Economics, Inc. Page 33 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Peaker Units in Vermont In their prefiled testimony, LaForest and Diebold stated that for the generation-dispatch assumptions for all studies, …Highgate Converter is unavailable when considered as a long-term outage and modeled with approximately 200 MW of import otherwise; Ryegate is dispatched at 20 MW; McNeil is dispatched at 51 MW; the Coventry landfill generator is modeled at 5 MW while the other thermal units are off and assumed held in reserve for loss of McNeil.24 The ability of “other thermal units” to provide peaking capacity was studied by E3 in order to determine the potential for these units to provide capacity beyond their ability to back-up the McNeil generator. While these peaking resources may be highly unreliable when considered individually, the aggregate ability of these units to provide peaking capacity in combination with other generation, import, and demand side resources should be fully considered in the determination of the need for additional transmission capacity. Generation Units The specific units modeled for reliability purposes and their associated capacities are listed in Table 8. These FCM listed capacities were obtained from generator bids submitted into FCM as detailed in ISONE’s March 3, 2008 “Forward Capacity Auction Results Filing” with FERC.25 The assumed capacities used in this analysis are from Table 2 in the Draft Responses to the DPS Supplemental Informal Discovery Requests for VELCO filed on May 28, 2008. The assumed capacity levels are generally lower than the FCM bid capacity levels as they reflect the contribution of the generator in mitigating the N-1-1 transmission constraints on the VELCO system. In addition the DPS 24 Southern Loop Project, Pre-filed Testimony of Dean L. LaForest and Christopher Diebold, Section 7.3, page 44. 25 Filed in Docket ER-08-000. Available online at http://www.isone.com/regulatory/ferc/filings/2008/mar/er08-633-000_03-0308_fca_results_filing.pdf Energy and Environmental Economics, Inc. Page 34 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 discovery requests did not contain capacity assumptions for the small diesel units of St. Albans, Barton, and Enosburg. The assumed capacity numbers for the Enosburg and Barton units were de-rated to 20% of their capacity based on VELCO reported transmission constraints.26 Table 10: Vermont Generators Excluded Generators FCM Capacity Assumed Capacity27 McNeil 52 40 Ascutney GT 9 10 St Albans Diesel 1.2 1.2 Berlin A&B GT 37.7 30 Gorge GT 6.6 5 Essex Diesels 2 5 Vergennes Diesels 4 0 Burlington GT 18.6 15 Enosburg Diesel 0.7 0.14 Barton Diesels 0.6 0.12 Rutland 5 GT 10.1 5 Florence Cogen A 3 3 Florence Cogen B 2.9 2 Swanton Peaker 1 20 12.5 Swanton Peaker 2 20 12.5 136.4 101.46 Total excluding McNeil italics – proposed Reliability Metrics The capacity available from the system of generators was modeled based on three statistics for each generator; the mean time to repair (MTR), the mean time to failure (MTF), and the Equivalent Demand Forced Outage Rate (EFORd). The MTF and MTR are approximations of the number of 26 Northern Zone resources contribute 20%, based on A.DPS:PET.1-118 27 The assumed capacities are based on Table 2 from the Draft Responses to the DPS Supplemental Informal Discovery Requests for VELCO, May 28, 2008. Energy and Environmental Economics, Inc. Page 35 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 hours a plant will be in an available or un-available state respectively. The implied outage frequency from these two rates is simply the ratio of the MTR to the sum of the MTR and MTF. Implied Outage Frequency = MTR MTR MTF The MTFs and MTRs for several of the generators were calculated directly from historic performance data for the generator. In the cases where no direct performance data was available generators were assigned MTFs and MTRs identical to other generators with available data and of similar fuel source and size. In the case where neither historical data nor a clear mapping to another generator was available generators were assigned the average MTR and MTF from A.DPS:PET.1-60.27 The EFORd reliability metric was also used to estimate each generator’s individual reliability performance. The EFORd metric is a more appropriate metric for estimating outage rates of peak capacity generators than the traditional Equivalent Forced Outage Rate (EFOR) as it accounts specifically for a generator’s expected availability during demand periods. The EFORd metric was adopted as the reliability metric for ISO New England capacity planning procedures in the Standard Market Design for New England (SMD) and re-confirmed in a 2005 FERC Order.28, 29 As with the MTR and MTF, the EFORd value was calculated directly from generator historical performance data where possible. In the absence of individual EFORd values the ISO New England EFORd class averages were used.30 27 This average was calculated excluding unit K 28 See New England Power Pool, 100 FERC ¶ 61,287 at P 12, 96-98 (2002), order on rehearing, 103 FERC ¶ 61,304 at P 77 & n.29 (2003). 29 Docket Nos. ER05-715-000,ORDER ACCEPTING, AS MODIFIED, PROPOSED INSTALLED CAPACITY REQUIREMENTS FOR THE 2005/2006 POWER YEAR , May 2005, Analysis 32 30 ISO New England EFORd Class Averages from NERC Brochure, Feb 2008. http://www.iso-ne.com/genrtion_resrcs/gads/index.html Energy and Environmental Economics, Inc. Page 36 of 58 Coolidge Connector Need and NTA Analysis Modeling Assumptions and Results Exhibit BKH-2 The available capacity from the system of generators was modeled in two ways. One approach involved a simulation of generator states using the MTF and MTR statistics for each generator. This approach returned an empirical distribution, through repeated simulation, for the available capacity from the system of generators in each hour in a 4-month period. The other approach used the generator EFORd rates to calculate the implied distribution of available capacity from the system of generators. The distribution calculated in this analysis represents the distribution of capacity in any given peak demand hour. Generator Simulation The MTRs and MTFs were used to simulate four months of hourly availability data for each generator, which was then combined with the capacity for that generator to determine the available capacity in each hour. The available hourly capacities for each of the generators listed in Table 8 were then combined to develop an hourly available capacity for the entire system of generators. The purpose of this approach was to estimate available capacity from these generators over the course of a 3 month summer system peak period when they could be used to alleviate capacity constraints on the Vermont grid. A more detailed description of this modeling approach can be found in Appendix B. The analysis looked at several scenarios for the generator MTRs and MTFs used for modeling purposes. In each scenario the MTR was held constant at the directly reported and mapped rates, while the MTF was varied between each scenario. The first “un-adjusted” scenario used the directly reported and mapped MTFs to simulate generator availability. The second scenario shifted the MTF in order to calibrate the plant implied outage frequency to the assumed plant EFORd value. Thus, given the EFORd rate and holding the MTR rate constant, the following equality was solved for the MTF: MTR EFORd MTR MTF Energy and Environmental Economics, Inc. Page 37 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 The EFORd scenario is an approximation of the reliability of the generators when they are used primarily for capacity demand purposes. The final “mid-point” scenario used the average MTF rates between the first two scenarios. Due to confidentiality agreements the specific plant performance data cannot be directly reported; however, Table 10 below lists the pertinent statistics for the EFORd and MTR values as well as the MTF and implied outage frequency values for each scenario. Table 11: Generator Statistics All Scenarios Maximum Average Minimum Un-Adjusted Mid-Point EFORd Matching EFORd MTR (hrs) MTF (hrs) Implied Outage Frequency MTF (hrs) Implied Outage Frequency MTF (hrs) Implied Outage Frequency 25.79% 12.33% 3.47% 83 51 7 366 126 7 73.08% 33.68% 4.69% 515 282 33 36.60% 17.00% 3.99% 931 438 59 25.79% 12.33% 3.47% The available capacity for the system of generators was simulated over 2,928 hours (24 hrs x 122 days). This process was repeated 1,000 times in order to approximate the distribution of available capacity in each hour. This resulted in 1,000 approximations for the available capacity in each individual hour over the course of three months. From the 1,000 approximations the 1st, 5th and 10th percentiles of available capacity for each hour were calculated. These represent the minimum capacity levels that can be expected to be available in each hour with 99%, 95%, and 90% confidence respectively. The percentiles for each of the 2,928 hours are shown in Figures 12-,14 with each figure representing a different scenario from Table 9. Energy and Environmental Economics, Inc. Page 38 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 12: Base Scenario Capacity Simulation - Un-Adjusted Scenario 120 1st Percentile 5th Percentile 10th Percentile MW 100 80 60 40 1 241 481 721 961 1201 1441 Hour 1681 1921 2161 2401 2641 2881 Figure 13: Mid-Point Scenario Capacity Scenario - Mid Point 120 MW 100 80 60 1st Percentile 5th Percentile 10th Percentile 40 1 241 481 721 961 1201 1441 Hour 1681 Energy and Environmental Economics, Inc. 1921 2161 2401 2641 2881 Page 39 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 14: EFORd Adjusted Scenario Capacity Simulation - EFORd Scenario 120 MW 100 80 1st Percentile 60 5th Percentile 10th Percentile 40 1 241 481 721 961 1201 1441 Hour 1681 1921 2161 2401 2641 2881 The results of these simulations in Figure 12 indicate that the generators listed in Table 8 can provide over 40 MW of capacity with 99% reliability over the course of a capacity constrained summer. Alternatively when the generators are used for capacity purposes as opposed to base generation, as illustrated by the EFORd case, they can provide over 70 MW of capacity with 99% reliability. Capacity Distribution Approach The first step in calculating the various capacity probabilities for the system of generators was to define all possible configurations of the generators in the system. Using the assumption that all generators were either on at their full capacity or forced out with zero capacity, the total available capacity was calculated for each configuration. Finally the probability of each configuration was calculated based on the individual EFORd values for the generators under the assumption that generator forced outages are independent of each other. The details of the assumed generator configurations and their associated probabilities can be found in Appendix C. The results of this analysis provided the available capacity distribution shown in Figure 15. Energy and Environmental Economics, Inc. Page 40 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 15: Dependable Capacity based on Probabilistic analysis 9. 5% 10 .0 % 9. 0% 8. 5% 8. 0% 7. 5% 7. 0% 6. 5% 6. 0% 5. 5% 5. 0% 4. 5% 4. 0% 3. 5% 3. 0% 2. 5% 2. 0% 1. 5% 1. 0% 0. 5% 110 100 90 80 70 60 50 40 30 20 10 0 0. 0% Capacity (MW) Total Capacity Based on EFORd Probability of Less than or Equal to this Capacity The probabilities associated with each capacity represent the probability that the system capacity will not exceed that particular level in any given demand hour during the year. These results show the overall capacity at all levels of reliability under the full system of generators including McNeil. The probability that the system will provide less than 80 MW of capacity during demand hours is approximately 1%. Alternatively the probability that the system will provide less than 60 MW of capacity during demand hours is approximately less than 0.5%. To redefine these results from a typical reliability perspective selected available capacity probabilities are displayed in Table 10. Table 12: EFORd Probabilistic Capacity Probability of Capacity Availability Capacity (MW) 99.99% 99.97% 99.00% 98.00% 97.00% 96.00% 95.00% 94.00% 93.00% 92.00% 91.00% 90.00% 41.32 46.96 80.84 86.46 91.34 93.96 96.46 98.46 100.26 101.46 103.46 106.34 Energy and Environmental Economics, Inc. Page 41 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 These capacity probabilities can be tied back to the generator simulation scenario in which the MTF hours were set to match each generators EFORd rate to the implied outage frequency rate. The mean capacity levels for the 1st, 5th and 10th percentile simulation graphs should match closely the capacity levels listed above for the 99%, 95%, and 90% probability respectively. Incremental Peaking Unit Capacity The generations analyses presented above evaluate the combined dependable capacity provided by the peaker generators and McNeil. However, VELCO has already included the capacity from McNeil in their determination of the peak load thresholds. Therefore, from an NTA perspective, the incremental dependable capacity provided by the peaker generators is the totals from above less 40MW. This adjustment is applied to each of the three generator simulation scenarios as well as the EFORd distribution capacity results. The incremental dependable capacity values are summarized in Table 11. The numbers reported for the generator simulation results are the minimum capacity percentile amongst all hours in the 4-month simulated period. Table 13: Dependable Capacity of Peaker Generators by Reliability Level (MW) Unadjusted recorded Mid-point simulation EFORd-based simulation EFORd probabilities 99 percent 95 percent 90 percent 3 32 41 13 45 54 28 51 59 41 56 66 The mid-point scenario for the generator simulation provides 45MW of dependable capacity above the McNeil generator. E3 considers the 95 percent reliability level to be reasonable for the need determination, because that level of aggregate reliability is comparable to what one could expect from a “reliable” unit that a transmission company would include in its reliability analyses. If the system of generators was used primarily in a peak capacity role then the performance would be closer to the results for the EFORd simulation scenario or similarly the EFORd distribution results, which both provide over 50 MW of incremental capacity above McNeil at a 95% reliability level. Energy and Environmental Economics, Inc. Page 42 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Combined Heat and Power (CHP) E3 also reviewed the CHP data provided in Attachment A.DPS:PET.218c. Based on the review of this data, E3 agrees with La Capra’s finding that there is no economically feasible CHP in the study area projected to be available to reduce summer load in 2011. This result is driven largely by two factors: (1) the relationship between projected electricity and natural gas prices and (2) the assumption that all CHP units will be sized only for hot water and heating requirements, with no units configured to accommodate summer cooling loads. However, if electricity prices increase substantially in the future (due, for example, to the expiration of long-term procurement contracts in 2012) the CHP projects could provide cost-effective summer peak reductions achieved through the addition of cooling equipment to the CHP unit. Economic Feasibility Economic feasibility for CHP installations is determined by a calculation of the ratio of benefits to costs. In the calculation performed by La Capra (included in Attachment A.DPS:PET.2-18c), the costs are defined as participant capital and operating costs relating to the CHP unit, and the benefits are based on ratepayer savings. The ratio is the NPV of benefits divided by the NPV of costs at a discount rate of 10%, and a ratio greater than 1.0 indicates an economically feasible outcome. The NPVs are based on 10-year projections commencing in 2007. No incentives or tax credits are assumed to be available to participants. It should be noted that using 2007 as the first year is not accurate for CHP projects going into service in later years, however the annual power and gas price data provided does not vary in ways that would change the results if this methodology were changed. Power and Gas Price Projections Attachment A.DPS:PET.2-18c provides the power and natural gas price estimates used in La Capra’s benefit-cost ratio test. These assumptions are described in the table below. Energy and Environmental Economics, Inc. Page 43 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Table 14: Power and Gas Price assumptions Row (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) New England ICAP forecast, $/kw-yr NE Hub All-hours Price, $/MWh (from AESC MA Definition1) VT Price All-hours ($/MWh) (from AESC) Forecasted Propane Prices, 2005$/mmbtu Forecasted Propane Prices, nominal$/mmbtu Exh.ES-2 Commerical/Industrial All ($/mmbtu) Exh.ES-2 Commerical/Industrial All (2005$/mmbtu) Ancillary Service Price Forecast ($/MWh) Wholesale Distillate Fuel 2005$/mmbtu Wholesale Distillate Fuel 2005$/mmbtu 2007 21.09 78.84 76.54 15.08 15.92 11.99 11.41 2.00 9.25 $ 9.77 $ $ $ $ $ 2008 20.80 66.24 64.76 14.90 16.13 10.10 9.38 1.69 9.07 $ 9.82 $ $ $ $ $ 2009 19.56 53.28 53.34 15.10 16.75 9.26 8.39 1.39 9.27 $ 10.29 $ $ $ $ $ 2010 31.03 46.33 47.79 15.57 17.70 8.63 7.63 1.25 9.74 $ 11.08 $ $ $ $ $ 2011 49.22 48.16 51.20 15.27 17.80 8.93 7.70 1.34 9.44 $ 11.01 $ $ $ $ $ 2012 78.07 50.07 54.86 14.98 17.90 9.40 7.91 1.43 9.15 $ 10.94 $ $ $ $ $ 2013 80.25 50.53 56.54 14.69 17.99 9.89 8.12 1.48 8.86 $ 10.85 $ $ $ $ $ 2014 82.49 51.00 58.27 14.40 18.07 10.88 8.71 1.52 8.57 $ 10.76 $ $ $ $ $ 2015 84.79 51.47 60.05 14.09 18.13 10.61 8.29 1.57 8.26 $ 10.63 $ $ $ $ $ 2016 87.15 51.95 61.89 14.14 18.64 10.90 8.31 1.62 8.31 $ 10.96 $ $ $ $ $ Row (6) provides the nominal natural gas price forecast corresponding with the nominal AESC power prices in rows (2) and (3). Over the 10year term from 2007 through 2016, nominal power prices are projected to decrease approximately 34%, with a 2016 nominal price of $51.95 per MWh, while gas prices are projected to decrease approximately 9%, with a nominal 2016 price of $10.90 per MMBtu. These projections equate to a marginal heat rate of 6,575 Btu/kWh in 2007, declining to 4,766 Btu/kWh in 2016. It is not clear what market fundamentals are driving the dramatic decline in power prices projected through 2016 while natural gas prices remain relatively constant in nominal terms. This high cost of natural gas fuel relative to the low value of electricity savings is a significant factor contributing to benefit-cost ratios of less than one. CHP Unit Configuration La Capra’s analysis assumes that all units are sized to accommodate the thermal load from hot water and heating requirements only, despite the fact that in some cases summer cooling loads are expected to be sizeable. The La Capra report concludes that out of a potential 54.2 MW of CHP capacity in the study area, only 4.1 MW would be available in the summer, all at a benefit-cost ratio of less than 1. The low coincidence with summer peak results from the assumption that CHP units will only serve heating and domestic hot water loads, a very small proportion of which are assumed to be operating during the summer peak. However, there are several potential additional ways each CHP project could be configured, which each result in different amounts of generation available during the summer peak. Energy and Environmental Economics, Inc. Page 44 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 The CHP units could be sized to cover the peak summer cooling, summer heating, and DHW loads. This configuration would require additional cooling equipment and a potentially substantial resizing of the CHP unit, both of which would increase the capital cost. The CHP unit could also be sized to cover the peak summer cooling, summer heating and DHW loads, up to a maximum equal to the winter loads. This would increase capital costs due to a reconfiguration of the CHP unit and the addition of cooling equipment and would result in maximum available summer capacity in the range of 54.2 MW, assuming all of the projects were implemented. The CHP unit could be sized to cover the facility’s electric load, with waste heat above facility needs not recovered. E3 did not consider this case because the facilities’ electric loads are quite large and thus this configuration would result in significant waste heat. In order to serve additional cooling loads, the CHP would need to be configured with equipment that can produce air conditioning. This equipment can comprise a combination of a desiccant dehumidifier and an absorption chiller, and may require additional or modified HVAC cooling coils, piping and controls equipment. Installation of this equipment is anticipated to cost approximately 7% more per kW, and approximately 50% more in annual O&M costs. CHP Cost Effectiveness under Alternate Electricity Cost Scenarios While the economics of the CHP projects improve substantially when the CHP units are also sized to serve summer thermal cooling loads, the improvement is not significant enough to achieve economic feasibility under La Capra’s electricity price forecast. With wholesale power costs of approximately $75 in 2011, CHP health applications become costeffective, yielding 4.4 MW of summer capacity. With power prices of about $110 per MWh in 2011, about 32 MW of summer CHP capacity would be cost-effective. Energy and Environmental Economics, Inc. Page 45 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Table 15: Cost Effective CHP Options in Vermont Summary Benefit/Cost Ratio - current Estimated Number of Sites Summer Peak kW Served Installed Capacity kW Total Summer Potential MW Power Prices for C:B >1 ($/MWh) Potential MW if C:B>1 College School 0.78 7 87 87 0.62 $ Grocery 0.82 122 106 106 12.97 125.00 12.97 Health 0.67 29 13 13 0.37 $ 0.91 13 328 328 4.40 75.00 4.40 Lodging $ 0.85 28 137 137 3.80 100.00 3.80 Office $ 0.85 137 180 180 24.58 110.00 24.58 Restaurant Retail 0.66 146 9 9 1.31 $ Warehouse 0.62 393 15 15 5.89 200.00 5.89 0.62 16 14 14 0.23 From the information in Table 13, one can develop a supply curve of cost-effective CHP, based on increasing levels of electricity power prices. The supply curve is shown in Table 16 below, but has not been adjusted for estimates of adoption levels. Table 16: Cost-Effective CHP Supply Curve Power Price ($/MWh) $ 75 $ 100 $ 110 $ 125 $ 200 Incremental Cost Effective CHP Summer MW 4.4 3.8 24.6 13.0 5.9 Increase in Cumulative Cost Power Price Effective CHP (%) Summer MW 46% 4.4 95% 8.2 115% 32.8 144% 45.8 291% 51.6 Care should be exercised in using the values in this table because the values are based on societal cost test benefit cost ratios of 1.0. The table does not evaluate payback periods and likely participation rates for the customers. Retail electricity prices will likely increase significantly in the future with the expiration of the current long-term contracts with Entergy and Hydro Quebec starting in 2012. Those retail price increases could increase customer interest in options such CHP, which could encourage adoption of CHP. However, we would generally expect CHP adoption levels to be far lower than EE adoption levels because of the typically higher complexity of CHP projects. Because the amount of CHP potential is relatively small under power price levels that are as much as twice what La Capra has assumed, and because there is a high level of uncertainty regarding the level of adoptions of cost-effective CHP by customers, E3 has not included CHP reductions in its need date analysis below. Energy and Environmental Economics, Inc. Page 46 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Transmission Need Date Figure 16 presents the need analysis performed by E3. The downward sloping line (vertical line marker) is the peak load threshold from VELCO’s pre-filed testimony.31 The top line is La Capra’s forecast,32 and the next line is E3’s forecast net of the status quo EE activities (triangle). Subtracting out peak load reductions for DR located in the Central and Northwest regions in Vermont yields the next line (solid circle). This (solid circle) peak load line indicates a need date of 2011, consistent with the planned in-service date of the Coolidge Connector. Factoring in the capacity that could be dependably provided by the small generators not included in the development of VELCO’s peak load threshold results in either the shaded or empty circle lines. The difference in the two lines is the level of dependability that one requires from the generators. The shaded yellow line requires 99% reliability from the generators, resulting in a need date of 2011. The empty circle line requires only 95% reliability from the generation, resulting in a 2013 need date. Because of the uncertainty surrounding the performance of the generators and the small margin in 2013, E3 believes that it is prudent to assume 2012 rather than a 2013 need date for the 95% generation scenario. Finally, the figure shows the potential impact of incremental EE activities beyond the status quo levels. That incremental EE pushes the need date out two years to 2014, but is less certain in deliverability than the status quo EE, DR, or peaker generation capacity. 31 From Reliability Needs Assessment table on page 25 of RSH-2, "target state-wide loads". 32 La Capra 90/10 w/ SQ EE is from Reliability Needs Assessment table on page 25 of RSH-2, "Total VT." Energy and Environmental Economics, Inc. Page 47 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Figure 16: Project Need Date 1,300 1,200 1,150 1,100 1,050 1,000 950 900 850 2015 2013 2011 2009 800 2007 Summer Peak MW (VT + CVEC) 1,250 E3 90th %tile Forecast w/ Losses & SQ EE E3 Forecast (w/ Status quo EE and DR) E3 Forecast (w/ SQ EE, DR, 99th %tile Gen) E3 Forecast (w/ SQ EE, DR, 95th %tile Gen) Peak Load Threshold La Capra 90/10 w/ SQ EE E3 Forecast (w/ SQ EE, DR, 95th %tile Gen, incr EE) Table 17: Vermont + CVEC Peak Loads and Peak Load Threshold (MW) Summer Peak Load Adjustments Year 2008 2009 2010 2011 2012 2013 2014 2015 2016 Demand Response 99th %tile 95th %tile (DR) Gen Gen -11.6 -11.6 -11.6 -11.6 -11.6 -11.6 -11.6 -13 -13 -13 -13 -13 -13 -13 -45 -45 -45 -45 -45 -45 -45 Incr EE -23.3 -32.0 -40.5 -45.9 -51.0 -56.1 -60.9 Adjusted Summer Peak Loads E3 90th %tile Forecast w/ Losses & SQ EE 1,138 1,150 1,162 1,175 1,189 1,204 1,220 1,235 1,251 E3 forecast less DR 1,151 1,164 1,178 1,193 1,208 1,224 1,239 Less DR, VELCO PreLess DR Less DR 95th %tile filed Peak and 99th and 95th Gen, Incr Load %tile Gen %tile Gen EE Threshold 1,180 1,170 1,138 1,106 1,082 1,155 1,151 1,119 1,087 1,145 1,165 1,133 1,092 1,135 1,180 1,148 1,102 1,115 1,195 1,163 1,112 1,095 1,211 1,179 1,123 1,085 1,226 1,194 1,133 1,075 It is important to note that the load threshold listed in the La Capra report (and employed in the analysis above) may be too high. The threshold load depends on a variety of assumptions about generator availability outside the state, acceptable voltage levels, power factor caps, and a “droop” in the Essex STATCOM and Granite synchronous connectors.33 Under some combinations of the treatment of these variables, the threshold load level at which the system is at risk can fall as low as 900 33 These factors are discussed further in the Draft Responses to the DPS Supplemental Informal Discovery Requests for VELCO, May 28, 2008. Energy and Environmental Economics, Inc. Page 48 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 MW, well below the current load levels. The Department’s experts have determined that 945 MW is the appropriate peak load threshold in 2011. This is 210 MW lower than what VELCO presented in their pre-filed testimony, and would indicate that the transmission project should be constructed according to the current proposed schedule. Energy and Environmental Economics, Inc. Page 49 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Cost of Delay or Deferral The pre-filed testimony of Ralph Roam discusses the projected cost of building the Coolidge Connector and the impact that specific delays in the completion of the project would have on the cost. Mr. Roam presents a base case and two alternative delay cases. The first delay case assumes that the project starts on time but takes longer than expected, incurring additional costs through extended hours needed to complete the project, cost increases due to inflation, and increased capital interests due to the longer construction time. The second delay case assumes that construction proceeds normally until the beginning of 2009, at which point it is indefinitely suspended. Eventually, it is resumed and completed within the normal time frame for the remaining portion. E3 posits a third deferral case, in which all project delays happen before the project is begun, and therefore there is no increase in hours or the interest paid on capital. The only cost increases are the result of inflation in the cost of transmission materials. Base Case In Mr. Roam’s Base Case, the construction term is approximately three years and the project has a commercial operation date (COD) of 4/1/2011. The total Base Case capital cost is $264.8 million, divided into categories as described in Table 15 below: Table 18: Base Case Capital Costs Material Labor Equipment Indirects Capital Interests Contingency Total Delay Cases Base Case COD 4/1/2011 $84,211,979 $42,030,987 $17,579,760 $54,047,180 $22,807,861 $44,135,553 $264,813,320 The two cases examined by Mr. Roam describe scenarios where the construction term is of longer duration than the Base Case. In the Increased Construction Term case, the project is commenced on schedule, the construction term is longer than expected, and the project achieves commercial operation later than expected. In the Construction Energy and Environmental Economics, Inc. Page 50 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Halted case, the project is commenced on schedule, halted, and then resumed after a period of months or years. Both cases result in increased costs for Material, Labor, and Equipment due to general inflation for these components. Both cases result in increased costs for Indirects (costs associated with engineering and design, operations, impact studies, etc.) due to inflation as well as the assumption that more hours related to Indirect activities will be charged to the project if the construction period is extended. Capital Interest costs also increase in both cases because of increased costs and the longer term of the construction period loan. The Contingency provision is not expected to change. The projected costs for Mr. Roam’s Increased Construction Term case are described in Table 16 below: Table 19: Increased Construction Term Case Material Labor Equipment Indirects Capital Interests Contingency Total Base Case Ave Annual Ave Annual COD 4/1/2011 COD 1/1/2012 % Increase COD 1/1/2013 % Increase $84,211,979 $89,477,886 8.42% $97,525,054 8.75% $42,030,987 $44,800,437 8.88% $48,862,704 8.99% $17,579,760 $18,739,419 8.89% $20,346,675 8.71% $54,047,180 $66,092,151 30.77% $82,610,490 27.44% $22,807,861 $27,665,873 29.36% $35,147,752 28.03% $44,135,553 $44,135,553 0.00% $44,135,553 0.00% $264,813,320 $290,911,319 13.35% $328,628,228 13.13% Because E3’s Deferred Construction case does not assume that the construction term has increased, the results from this Increased Construction Term case are not utilized in E3’s Deferred Construction case analysis. The projected costs for Mr. Roam’s Construction Halted case are described in Table 20 below: Table 20: Construction Halted Case Material Labor Equipment Indirects Capital Interests Contingency Total Base Case Ave Annual Ave Annual Ave Annual % Ave Annual % COD 4/1/2011 COD 1/1/2014 % Increase COD 1/1/2015 % Increase COD 1/1/2016 Increase Increase COD 1/1/2017 $84,211,979 $106,384,456 8.87% $116,141,896 8.95% $126,892,557 9.02% $138,742,159 9.07% $42,030,987 $53,331,052 9.04% $58,248,111 9.09% $63,661,166 9.13% $69,622,683 9.17% $17,579,760 $22,110,380 8.70% $24,047,338 8.71% $26,176,291 8.74% $28,518,147 8.78% $54,047,180 $72,385,314 11.21% $73,795,532 8.66% $75,245,934 7.21% $76,737,664 6.29% $22,807,861 $34,916,370 16.75% $34,916,370 12.03% $43,287,704 14.44% $47,953,737 13.80% $44,135,553 $44,135,553 0.00% $44,135,553 0.00% $44,135,553 0.00% $44,135,553 0.00% $264,813,320 $333,263,125 8.72% $351,284,800 7.83% $379,399,205 7.86% $405,709,943 7.70% Energy and Environmental Economics, Inc. Page 51 of 58 Coolidge Connector Need and NTA Analysis E3’s Deferred Construction Case Exhibit BKH-2 The Deferred Construction case examines the impacts of delaying commencement of construction until approximately three years before the line is needed. It is similar to the base case except (a) capital cost estimates incorporate inflation rates expected in future years, (b) capital interests are increased to reflect capital cost increases due to inflation, and (c) the project achieves commercial operational at a later date. The construction term remains the same as in the base case. Table 18 shows the results of the Deferred Construction Case. The Material, Labor, and Equipment costs in the Deferred Construction case are identical to those in the Construction Halted case. This is because cost increases in these categories are the result of inflation only. The Indirects category in the Construction Halted case includes cost increases due to additional hours related to Indirects activities, as well as inflation. Mr. Roam’s testimony does not supply enough information to determine how to adjust the figure to remove the additional hours. The Indirects category has therefore been inflated using the minimum inflation rate of the Labor category and the Indirects category. Capital Interests line has been increased pro rata with increased construction costs because the construction term is the same as in the Base Case. The Contingency line has not been adjusted. Table 21: Deferred Construction Case Base Case Ave Annual Ave Annual Ave Annual % Ave Annual % COD 4/1/2011 COD 1/1/2014 % Increase COD 1/1/2015 % Increase COD 1/1/2016 Increase Increase COD 1/1/2017 Material $84,211,979 $106,384,456 8.87% $116,141,896 8.95% $126,892,557 9.02% $138,742,159 9.07% Labor $42,030,987 $53,331,052 9.04% $58,248,111 9.09% $63,661,166 9.13% $69,622,683 9.17% Equipment $17,579,760 $22,110,380 8.70% $24,047,338 8.71% $26,176,291 8.74% $28,518,147 8.78% Indirects $54,047,180 $68,577,808 9.04% $73,795,532 8.66% $75,245,934 7.21% $76,737,664 6.29% Capital Interests $22,807,861 $27,758,920 7.41% $29,816,217 7.41% $31,676,907 7.16% $33,716,817 7.03% Contingency $44,135,553 $44,135,553 0.00% $44,135,553 0.00% $44,135,553 0.00% $44,135,553 0.00% Total $264,813,320 $322,298,170 7.41% $346,184,647 7.41% $367,788,408 7.16% $391,473,023 7.03% PV 4/1/2011 $264,813,320 $263,698,176 $263,309,131 $260,054,867 $257,322,416 Total PV Savings/(Cost) $1,115,144 $1,504,189 $4,758,453 $7,490,904 Annual Average PV Savings $405,507 $401,117 $1,001,780 $1,302,766 Though Mr. Roam found that his delay scenarios resulted in high cost overruns and cost increases, E3’s deferral case shows a net present value benefit from delaying commencement of the project. The average annual project cost increases resulting from E3’s deferred construction case are less than or equal to 7.41%. Because VELCO’s after-tax WACC is 7.57%, this result means that the present values of the Deferred Energy and Environmental Economics, Inc. Page 52 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Construction case capital costs are less than those in the Base Case. In present value terms, delaying construction to commercial operation dates in January of 2014, 2015, 2016, or 2017 are all less costly in present value terms than the Base Case. For example, the three year deferral case has an average annual savings of $405 thousand per year. When expressed in revenue requirement levels, the present value savings would be approximately 40% higher,34 or approximately $567 thousand per year.35 34 Revenue requirements are the total amount of money that the utility collects from customers to pay for all operating and capital costs, including a fair return on investment. The present value of revenue requirements (PVRR) is higher than the present value of the project costs because of additional items that are typically not including in project cost estimates such as property and income taxes and O&M. The 40% adder is a typical PVRR adjustment for transmission projects. For example, the San Diego Gas and Electric Company’s Sunrise Powerlink transmission project has a 42% adder for a project with a 40 year expected life. Typical PVRR adders range from 30% to 60%. 35 The testimony of Brian Horii uses a simplified approach to estimating the deferral value that arrives at a deferral value of $550 thousand per year for this value. Energy and Environmental Economics, Inc. Page 53 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Conclusion Overall, E3 finds that the VELCO and La Capra analyses were thorough and fairly complete, especially given the information available to them at the time the reports were prepared. Incorporating updated information on peak loads and participation in the ISO-NE Forward Capacity Market, as well as allowing small generators to contribute to resolving the peak capacity problem, can push the need date for the Coolidge Connector to between 2012 and 2014 under the threshold values presented by La Capra in their report. E3 disagrees with Mr. Roam’s testimony that delays in the project will necessarily result in substantial cost increases in present value terms. While this is true for the delay cases that Mr. Roam showed where the project was started and then suspended or delayed, this would not be the case if the start date for the project were delayed and the project proceeded on a normal schedule once begun. Due to the high rate of inflation in the cost of transmission elements, the monetary value to deferral is relatively small. However, deferral of the line has the advantage of allowing Vermont to gather information and gain confidence in the new levels of EE activity and the evolving DR market. In addition, deferral could allow time for new generation to be built in response to the expiration of the long-term contracts, and reveal customer responses to increasing electricity prices. The discussion above is predicated on the peak load threshold levels in VELCO’s pre-filed testimony. If the Board determines that the Department’s lower peak load threshold is the appropriate value to use for planning purposes, then deferral would not be possible and the transmission project should proceed on its current schedule. Energy and Environmental Economics, Inc. Page 54 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Appendix A: 1-in-10 versus 90th percentile peak load forecast This section explains the two methods of making a 1-in-10 peak demand forecast. The first method assumes a 1-in-10 weather event (e.g., very hot day). It captures the extreme weather’s effect on peak demand. The second method derives a 1-in-10 peak demand forecast for a future year as the 90-percentile of that year’s forecast distribution. Suppose the summer peak demand data is generated by a linear regression: yt = xt + t where yt = peak demand in year t; xt = row vector of explanatory variables (e.g., the intercept, time trend, and average WTHI); = column vector of coefficients to be estimated; and t = normally distributed error with zero mean and finite variance 2. Let b be the ordinary least squares (OLS) estimate of . Conditional on a forecast of the explanatory variables xF in a future year, the peak demand forecast is xF b. Suppose xF = x’ captures a 1-in-10 weather event. The forecast based on the 1-in-10 weather event is simply x’ b. Suppose (m, v2) are the mean and variance of a peak demand forecast based on uncertain drivers. The 90-percentile of the forecast’s distribution is (m + 1.28 v), the 1-in-10 forecast that an actual peak may exceed with a 10% probability. To derive (m, v2), assume xF = x”, which captures the median weather event. The mean forecast is m = x”b, whose conditional variance is: var(m | xF = x”) = xF xFT + s2 Energy and Environmental Economics, Inc. Page 55 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 where = covariance matrix of b and s2 = OLS estimate of 2. This conditional variance understates the dispersion of m when xF is uncertain with covariance matrix An estimate of the unconditional variance of m is:36 v2 = (xF xFT + s2) + bT b + trace (). Hence, v2 the sum of (a) (xF xFT + s2), the forecast’s conditional variance, (b) bT b, the explanatory variables’ dispersion magnified by the estimated coefficients, and (c) trace (), the explanatory variables’ dispersion magnified by the coefficient estimates’ dispersion. Feldstein, M. (1971) “The Error of Forecast in Econometric Models when the Forecast-Period Exogenous Variables are Stochastic.” Econometrica 39(1): 5560. 36 Energy and Environmental Economics, Inc. Page 56 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Appendix B: Peaking Generation Analysis The approach for simulation of hourly available capacity from a system of generators follows from Mazumdar et al.37 The process for each simulation of 3 month hourly system capacity starts with individual simulations for each generator in the system. This is accomplished under the assumption that the state of each generator in any given hour is independent of all other generators in the system. For each generator the only parameters necessary are the Mean Time to Failure (MTR) and Mean Time to Repair (MTR). The initial state of a generator is simulated with the following probabilities: P[up] MTF MTF MTR P[ Down] MTR MTF MTR Once the initial state has been determined the amount of time in hours that the generator is in that initial state is assumed to be exponentially distributed with mean equal to MTF if the generator is in the up state, and mean equal to MTR if the generator is in the down state.38 Therefore an exponential random variable with the appropriate mean is simulated and that number of hours is assigned the current state. Once the hours for that state have been simulated, the generator is then assumed to switch to the opposite state and the number of hours that it remains in the second state is again simulated as an exponential random variable with mean equal to MTF if the generator is in the up state and MTR is the generator is in the down state. This process is repeated until it accounts for all hours in the three month period. Mazumdar M., Coit D. and Shih FR (1999), “A Highly Efficient Monte Carlo Method for Assessment of System Reliability Based on a Markov Model”, American Journal of Mathematical and Management Sciences 19(1-2): 116-133. 37 38 Details of the exponential distribution can be found for example in: Johnson N., Kotz S., and Balakrishnan N., Continuous Univariate Distributions, Volume 1 (1994), Wiley Series in Probability Energy and Environmental Economics, Inc. Page 57 of 58 Coolidge Connector Need and NTA Analysis Exhibit BKH-2 Appendix C: Generator System Capacity Distribution We assume for a general system of i = 1,…,n generators and with the capacity of generator i defined as ci then we can calculate the possible system configurations and their associate probability as follows. Assuming each generator is either on at full capacity or off at zero capacity then there are a total of j = 1, … , 2n possible on/off configurations for all n generators. If we further define the state of generator i in configuration j as si,j where si,j is 0 is generator i is out in configuration j and 1 if generator i is available in configuration j. Our available system capacity SC j in configuration j is SC j i 1 c j si , j n Having the capacity available in each configuration j we must also calculate the probability of that particular configuration occurring. The configuration probabilities are calculated assuming independence of the generator states in each configuration that is whether one generator is available or out has no effect on the states of the other generators. We define the probability that a generator i is unavailable during any given hour of the year as pi, where pi is assumed to be that generators EFORd. Conversely the probability that the generator is available in any given hour is 1- pi. Having these probabilities the probability of any given generator configuration SP j is the product of the probabilities of each generators state in that configuration. SPj (1 pi ) pi i on i off This leads to a system of j generator configurations with associated configuration capacities SC j and configuration probabilities SP j. These configurations represent the possible states of the generator system and their associated probability in any given demand hour during the year. Energy and Environmental Economics, Inc. Page 58 of 58