Load Granularity Refinements. Pricing Study Results and Implementation Costs and Benefits Discussion

Transcription

1 Pricing Study Results and Implementation Costs and Benefits Discussion January 14, 2015

2 Table of Contents I. Executive Summary... 3 II. Introduction... 3 III. Stakeholder Process... 4 IV. Pricing Study Results... 5 A. Average nodal LMPs ( )... 5 B. Difference of nodal and DLAP LMPs C. Regression Analysis D. The Greater Fresno area V. Implementation Costs VI. Benefits A. Potential Benefits B. Quantifying benefits VII. Second level of aggregation VIII. Next Steps The Load Granularity Issue paper is a separate document which can be located at: CAISO/M&ID/KW 2 January 14, 2015

3 I. Executive Summary The California Independent System Operator (ISO) filed for a permanent waiver from the Federal Energy Regulatory Commission s (FERC s) request to further disaggregate Default Load Aggregation Points (DLAPs) by Release 2. In the Order denying the waiver, FERC granted the ISO a one year extension to either comply with, or seek further relief from, the request to disaggregate the Load Aggregation Points (LAPs) by June 3, The ISO conducted a fully nodal pricing study to evaluate the impacts of load granularity pricing refinements in the CAISO market. Day-ahead nodal locational marginal prices (LMPs) were analyzed in three dimensions: average nodal LMPs, differences of nodal and DLAP LMPs, and a regression of nodal LMPs on DLAP LMPs. The results indicate that nodal price dispersion across the system and variation from the DLAP LMPs are minimal. The price dispersion and variation that does exist is not consistent or contiguous to be used to efficiently create more granular load zones, with the exception of the Fresno area. However, the congestion driven pricing in the Fresno area is inconsistent and will likely dissipate as already approved transmission enhancements become operational in the future. The ISO has collected implementation cost estimates from eight stakeholders and the ISO as requested by FERC in the recent order denying the ISO s request for permanent waiver. Cost estimates total $21.7 million in one-time implementation costs with $2.5 million annually for slight disaggregation, and $147 million in one-time implementation costs with $12.3 million annually for fully nodal disaggregation. More granular load pricing does have the potential to provide benefits. Those benefits are expected to be negligible given the pricing study results. Furthermore, there are other factors that will impact the magnitude of benefits such as existing market products and processes that can be used to capitalize on the same benefits, and the probability of changes to the current retail rate structure. Quantifying benefits in a meaningful manner is challenging and would require several contentious assumptions to be made. Therefore given the expected negligible benefits, the ISO does not intend to conduct extensive quantitative studies but rather will speak to the benefits qualitatively. II. Introduction FERC s original September 21 st, 2006 Order on MRTU found the DLAP approach reasonable and a simplified method for introducing LMP pricing, while minimizing its impact on load. However, FERC also directed the ISO to increase the number of LAP zones in order to provide more accurate price signals and assist participants in hedging of congestion charges after three years of experience with the new market. In 2010, the ISO initiated a stakeholder process to evaluate LAP disaggregation. The ISO conducted a pricing study, which found that with the exception of one small area prices in sub-lap regions did not differ significantly from the DLAPs. The Market Surveillance Committee (MSC) did a spatial pricing study CAISO/M&ID/KW 3 January 14, 2015

4 showing that short of nodal pricing, there was no efficient way to group nodes in large zones. Stakeholders indicated that they would face significant implementation hurdles and that many potential benefits were already available to them through other ISO processes, or would be soon. FERC accepted the ISO s request to delay disaggregation of LAPs: We find that more pricing information and additional experience with the MRTU design changes, such as proxy demand response and convergence bidding will allow CAISO to develop a proposal to further disaggregate the default LAPs, as directed. 1 FERC extended the deadline to disaggregate LAPs until October 1, In 2013, the ISO initiated a stakeholder process to again evaluate with stakeholders disaggregation of the existing LAPs. The ISO performed a simple pricing analysis comparing prices in the existing sub-laps to the DLAPs, which indicated that there were no major price disparities. Based on this information, and stakeholder input indicating that they did not see enough price disparity and they would incur substantial costs from disaggregating the DLAPs, the ISO filed a motion for a permanent waiver of FERC s directive to disaggregate the LAPs. On June 3, 2014, FERC issued an order that denied the ISO s request for a permanent waiver to complying with FERC s previous orders to disaggregate existing LAPs. 2 The order extended for one year from the date of the order the time for the ISO to comply, or seek further relief from the disaggregation request. The ISO initiated another stakeholder process to evaluate LAP disaggregation in September 2014 with an issue paper and stakeholder meeting. The paper and meeting presented a proposal for a study of the pricing dispersion between the individual demand nodes and the current DLAPs used for bidding and pricing of load. An information request was subsequently sent out asking stakeholders to provide estimated implementation costs for nine cost categories and four levels of disaggregation. This paper presents the pricing study results, estimated implementation costs, and a discussion of potential benefits. III. Stakeholder Process The ISO will host a conference call with stakeholders on January 21, 2015 to allow stakeholders to ask questions or comment on the pricing study results and implementation costs and benefits discussion. Written comments from stakeholders should be sent to the ISO by January 30, The next phase of stakeholder engagement will begin in February with a draft straw proposal. If the proposal requires a tariff filing, the ISO will initiate that stakeholder process concurrently with presenting its recommendation to the Board such that the language can be filed before the June 3 rd FERC filing date. The end result of the stakeholder process will be presenting the ISO s recommendation to the Board of Governors in May 2015 on whether to further disaggregate the LAPs or that the existing DLAP structure is just and reasonable. 1 Cal. Indep. Sys. Operator Corp., 136 FERC 61,055, at P 6 (2011). 2 Cal. Indep. Sys. Operator Corp., 147 FERC 61,181 (2014). CAISO/M&ID/KW 4 January 14, 2015

5 The schedule for stakeholder engagement is listed below. Date January 21 January 30 February 19 March 3 March 13 Late March early April April - May May June 3 Milestone Stakeholder call to discuss Pricing Study Results Receive written comments from stakeholders Post Straw Proposal Hold Stakeholder Meeting Receive written comments from stakeholders If needed, post draft final proposal, hold stakeholder call, and receive stakeholder comments Prepare Board documents and FERC filing Present recommendation to Board File at FERC IV. Pricing Study Results This study reflects a full nodal spatial price dispersion analysis to evaluate the impacts of load granularity pricing refinements in the CAISO market. Specifically, we focus on whether there are significant price differences between nodal LMPs and DLAPs, and the cause of any large differences. The study analyzes day-ahead LMPs at all load nodes on the system. We used day-ahead hourly energy prices from 2011 through November 14, 2014 in the study 3. First we conducted a historical review of average nodal LMPs at all load nodes from Second, we analyzed the differences between nodal and DLAP LMPs. Last, we performed a regression analysis to ascertain the relationship between the nodal LMPs and the DLAP LMPs. In addition to evaluating the magnitude of price dispersion, this study also analyzes the consistency of any price dispersion over time and by location. A. Average nodal LMPs ( ) First, we reviewed historical nodal prices by taking the simple average nodal price at each load node over the four year period between 2011 and 2014 from the day-ahead market. We examined the average price at each load node to determine (1) the spatial variation across the system of average prices, and (2) the areas where groups of nodal LMPS are on average higher or lower than other nodal LMPs. Figures 1 through 6 below show that variation in average prices across the system is minimal and remains consistent year to year. With the exception of Fresno and surrounding counties in 2014, there are no contiguous regions where nodal prices are on average higher or lower relative to the other nodes on the system. 3 We excluded the first two years of the nodal markets, 2009 and 2010, from the study as two previous pricing studies conducted by the ISO and one by the Market Surveillance Committee found minimal price dispersion during those years. February 6 th and 7 th 2014 were also excluded from the study due to a significant spike in gas prices. CAISO/M&ID/KW 5 January 14, 2015

6 Figure 1 shows the ( ) average nodal prices across the system range from $52/MWh to $26/MWh with 90% of average prices between $44/MWh and $35/MWh. We also analyzed the average nodal prices by Load Aggregation Point and by year to identify any regions or time periods where a subset of nodes have consistently higher or lower average prices compared to the other load nodes. Figure 1 Range of average nodal LMPs ( ) $55 $50 Average nodal LMP $45 $40 $35 $30 $25 Average Nodal LMPs Figures 2 and 3 below examine the average nodal prices geographically. Figure 2 uses the same average prices in Figure 1 but shows them by LAP 4. The three major DLAPs have similar price variations. This indicates that the load nodes with higher or lower average LMPs from Figure 1 are dispersed throughout the three major DLAPs rather than being concentrated in one. 4 Valley Electric Association (VEA) is not included in this chart because there is only two years worth of data ( ), which made it appear as though VEA had consistently higher average prices than the other three DLAPs. CAISO/M&ID/KW 6 January 14, 2015

7 Figure 2 Range of average nodal LMPs by LAP ( ) $55 $50 Average nodal LMP $45 $40 $35 $30 $25 SCE SDGE PGAE Figure 3 plots the average prices in Figures 1 and 2 on a map of California to determine if there are regions within each LAP that have higher or lower average priced load nodes. The majority of load nodes have an average LMP in the $35-$40/MWh range (green). Load nodes with lower average prices (blue) and higher average prices (pink) are scattered throughout California. There is a cluster of nodes in the $40-$45/MWh range (yellow) concentrated in Fresno, Madera, Modesto, and Mariposa Counties. CAISO/M&ID/KW 7 January 14, 2015

8 Figure 3 Average nodal LMPs heat map ( ) Figure 4 compares the average nodal LMPs by year in order to assess if there are significant changes in the average price variation year to year. Average nodal prices in 2011 and 2012 are similar. The average nodal prices shift up in 2013 and then again in Greenhouse gas emission compliance costs and higher gas prices contributed to the upward shift in 2013 and higher gas prices contributed to the additional upward shift in Despite higher prices in 2013, the price variation (difference between the highest average and lowest average nodal LMP) remains the same compared to 2011 and 2012, approximately a $15/MWh variation. In 2014, the price variation increased to approximately a $22/MWh difference. CAISO/M&ID/KW 8 January 14, 2015

9 Figure 4 Range of average nodal LMPs by year ( ) $70 $65 $60 Average nodal LMP $55 $50 $45 $40 $35 $30 $25 $ Given the change in variation of average prices in 2014 compared to the other years, we plotted the average nodal prices for each year on a map of California to identify the location of those load nodes. The maps for 2011 (Figure 5) and 2014 (Figure 6) are shown below 5. Average LMPs in 2011 were primarily in the $30-$35/MWh range (pink) with a few higher priced nodes (orange) scattered throughout California. The average LMP map for 2014 shows significant changes in both the average LMPs overall (note the change in price categories defined in the legend) as well as higher average LMPs in Fresno and surrounding counties. Prices in the Greater Fresno area did not increase until 2014 due to congestion on lower voltage transmission lines. 5 The map for 2012 is similar to The map for 2013 is similar to that of 2014 without the significant change in average LMPs in Fresno and surrounding counties. CAISO/M&ID/KW 9 January 14, 2015

10 Figure 5 Average nodal LMPs heat map 2011 Figure 6 Average nodal LMPs heat map 2014 CAISO/M&ID/KW 10 January 14, 2015

11 Assessing the average nodal prices from has shown that the variation in average prices across the system continues to remain minimal and is consistent year to year. With the exception of the Greater Fresno area in 2014, there is no contiguous group of nodes that are on average higher or lower relative to the other nodes on the system, making disaggregation challenging short of fully nodal. Further discussion of the Greater Fresno area prices in 2014 is provided in a subsequent section. B. Difference of nodal and DLAP LMPs Currently, day-ahead load is bid in and settled at the DLAP LMP as opposed to the nodal LMP. Within a LAP, some load nodes have higher LMPs and some lower LMPs relative to the DLAP LMP. Because they are all within the same LAP, all load is charged the same DLAP LMP. Analyzing the difference between the nodal LMPs at load nodes and the DLAP LMP for which the node resides will indicate the extent to which lower priced nodes are subsidizing higher priced nodes. The following four figures show the frequency of nodal LMPs above and below the DLAP LMP. The figures reflect the difference of each hourly nodal LMP and DLAP LMP 6 from , with the exception of Valley Electric Association which has hourly LMPs from Nodes with hourly LMPs that have a difference less than $0/MWh (i.e. negative) are those subsidizing the nodes with hourly LMPs with a difference greater than $0/MWh (i.e. positive). Overall, most of the hourly nodal LMPs are within $2/MWh of the hourly DLAP LMP, representing less than 5% of a $40/MWh or greater DLAP LMP. Pacific Gas and Electric (Figure 7) and Southern California Edison (Figure 8) LAPs have similar distributions. Approximately 40% of nodal LMPS in PGAE and 44% in SCE are within $0.50/MWh of the DLAP LMP, and both have almost 80% of nodal LMPs within $2/MWh of the corresponding DLAP LMP. The distribution of price differences in PGAE and SCE are slightly skewed to the left. In other words, there are more LMPs in those LAPs that are lower than the DLAP LMP. We also analyzed the distribution of differences by year. The differences in LMPs in 2011 and 2012 are more centered on $0/MWh, indicating that nodal LMPs were closer to the DLAP LMP in those years compared to 2013 and We calculated the difference as the hourly nodal LMP minus the hourly DLAP LMP. CAISO/M&ID/KW 11 January 14, 2015

12 Figure 7 Difference of nodal LMP and DLAP LMP PGAE % 30% 20% 10% 0% < -$25 -$25 to -$10 -$10 to -$5 Percent of hourly nodal prices -$5 to -$2 -$2 TO -$1 -$1.5 TO -$0.50 -$.50 TO $0 $0 TO $0.50 $0.50 TO $1 $1 TO $2 $2 TO $5 $5 TO $10 $10 TO $25 > $25 Figure 8 Difference of nodal LMP and DLAP LMP SCE % 30% 20% 10% 0% < -$25 -$25 to -$10 -$10 to -$5 Percent of hourly nodal prices -$5 to -$2 -$2 TO -$1 -$1.5 TO -$0.50 -$.50 TO $0 $0 TO $0.50 $0.50 TO $1 $1 TO $2 $2 TO $5 $5 TO $10 $10 TO $25 > $25 The distribution of LMP differences in SDG&E (Figure 9) and VEA (Figure 10) LAPs are centered on $0/MWh. The SDG&E LAP has 75% of nodal LMPs within $0.50/MWh of the DLAP LMP and 92% are within $2/MWh of the DLAP LMP. Valley Electric has 97% of nodal LMPs within $0.50/MWh of the DLAP LMP and 99% are within $2/MWh of the DLAP LMP. The nodal LMPs are extremely close to the DLAP LMPs, therefore the degree of cross-subsidization is minimal. CAISO/M&ID/KW 12 January 14, 2015

13 Figure 9 Difference of nodal LMP and DLAP LMP SDGE % Percent of hourly nodal prices 40% 30% 20% 10% 0% < -$25 -$25 to -$10 -$10 to -$5 -$5 to -$2 -$2 TO -$1 -$1.5 TO -$0.50 -$.50 TO $0 $0 TO $0.50 $0.50 TO $1 $1 TO $2 $2 TO $5 $5 TO $10 $10 TO $25 > $25 Figure 10 Difference of nodal LMP and DLAP LMP VEA % 40% 30% 20% 10% 0% < -$25 -$25 to -$10 -$10 to -$5 -$5 to -$2 -$2 TO -$1 Percent of hourly nodal prices -$1.5 TO -$0.50 -$.50 TO $0 $0 TO $0.50 $0.50 TO $1 $1 TO $2 $2 TO $5 $5 TO $10 $10 TO $25 > $25 Any difference in a nodal LMP relative to the DLAP LMP is due to differences in the congestion components and/or loss components of the LMPs. The contribution each component has to the difference in each hour for each node was calculated. Then the average difference and average contributions were taken by the same pricing categories as Figures Figure 11 below is for the PGAE LAP; all other LAPs show a similar pattern. CAISO/M&ID/KW 13 January 14, 2015

14 When the average price difference is large (tail ends of the chart), the majority of difference is due to congestion. As the difference decreases, i.e. gets closer to $0/MWh, loss becomes the main contributing factor. In general, congestion causes the highest difference between nodal LMPs and DLAP LMPs but occurs less frequently than losses. Losses are calculated every hour, therefore have a high frequency of occurrence but the difference they create between the nodal LMPs and DLAP LMPs is minimal. This analysis was done to see how price differences may be minimized if load zones were based on loss factors or congestion conditions. The loss component of each node is based on the transmission voltage levels at that location. In the ISO footprint, there are high voltage lines right next to low voltage lines. Two nodes physically close to one another geographically can have significantly different loss components. Therefore creating contiguous zones based on similar loss components would be infeasible. Creating zones based on congestion patterns would be difficult as well, because those patterns are not consistent. The ISO system is constantly evolving with new transmission, new resources, resources retiring, as well as unforeseen outages and de-rates. Therefore congestion is not consistent and oftentimes unpredictable. Creating load zones based on something that is constantly changing would require continuous re-evaluation and possibly re-defining zones. Figure 11 Contributing factors to nodal and DLAP LMP differences PGAE Average price difference (nodal LMP - DLAP) lmp) $70 $60 $50 $40 $30 $20 $10 $0 Congestion Loss The average difference between the nodal LMP and DLAP LMP at each load node from is shown geographically in Figure 12. The magnitude of the average differences are scattered throughout the ISO footprint with the exception of the Greater Fresno area. Fresno and surrounding counties have nodal LMPs in that are, on average, higher than the DLAP LMP by more than $3/MWh. As previously noted, this is due to congestion going into the Fresno area. CAISO/M&ID/KW 14 January 14, 2015

15 Figure 12 Average difference of nodal LMP to DLAP LMP heat map ( ) C. Regression Analysis The ISO conducted a regression analysis similar to that of the Market Surveillance Committee in The current analysis regresses day-ahead nodal LMPs for load nodes on the DLAP LMPs using data from This analysis will show how the nodal LMPs move relative to the DLAP LMPs and how close the average nodal LMP is likely to be to the average DLAP LMP. The regression results are presented in the following four figures, one for each current load zone. The variable on the horizontal axis is the intercept term from the regression results divided by the average DLAP LMP 7. The variable on the vertical axis is the coefficient term of the regression results. The coefficient term indicates how well the two LMPs move together. A coefficient greater than 1 indicates the nodal LMP will have a larger movement relative to a movement in the DLAP LMP. A coefficient less than 1 indicates the nodal LMP will have a smaller movement relative to a movement in the DLAP LMP 8. 7 The intercept was normalized so the intercept and coefficient terms can be interpreted together in a clean manner. 8 For example, if the coefficient is.5 then a $1 increase/decrease in DLAP LMP will result in a $0.50 increase/decrease in nodal LMP. A coefficient equal to 1.5 means a $1 increase/decrease in DLAP LMP will result in a $1.50 increase/decrease in nodal LMP. A negative coefficient means an increase/decrease of DLAP LMP will result in a decrease/increase of nodal LMP. CAISO/M&ID/KW 15 January 14, 2015

16 If the average nodal LMP is equal to the average DLAP LMP, then the point will fall on the green reference line. Points above the reference line have an average nodal LMP greater than the average DLAP LMP; points below the reference line have an average nodal LMP less than the average DLAP LMP. If all the nodes fall on the reference line, then the linear regression line (orange dashed line) will also fall on the reference line. In all four LAPs, the majority of regression results are clustered on or near the reference line, and the linear regression line (orange dashed line) is close to the reference line as well. Furthermore, the cluster in each LAP is around (intercept, coefficient) equal to (0, 1). PGAE (Figure 13) and SCE (Figure 14) LAPs have a minimal amount of points away from the reference line and/or (0, 1). SDGE (Figure 15) regression results are mostly along the reference line with the exception of a few outliers. When the outliers are removed from the chart, the orange linear line follows the reference line more accurately. VEA (Figure 16) results all follow the green reference line. This analysis was also conducted by year. As would be expected, all results are similar to what is shown below with the exception of PGAE in 2014 due to the load nodes in Fresno and surrounding counties. Figure 13 Regression results PGAE CAISO/M&ID/KW 16 January 14, 2015

17 Figure 14 Regression results - SCE Figure 15 Regression results - SDGE CAISO/M&ID/KW 17 January 14, 2015

18 Figure 16 Regression results - VEA D. The Greater Fresno area The Fresno and surrounding counties have higher than average nodal LMPs in 2014 relative to the other load nodes on the system. Furthermore, the difference between the nodal LMPs in those counties and the PGAE DLAP LMP is higher than the difference between other nodal LMPs and the corresponding DLAP LMP. Therefore, that area is one that could potentially be aggregated into another DLAP if the pricing trend were to become consistent. However, a few transmission projects have already been identified and approved to address potential concerns in the area. The Transmission Plan 9 identified reliability-driven transmission projects to address potential overload and voltage concerns in the Greater Fresno area, such as the Gates-Gregg 230 kv Line. Given the transmission projects that have been identified and approved through the Transmission Planning process, the ISO does not believe the high congestion and nodal LMPs in the Fresno area will continue to exist. The congestion in that area should be mitigated with the expansion of the transmission system in the Greater Fresno area. Therefore, potentially creating a new DLAP based on congestion patterns that will likely become negligible is not conducive to the current ISO markets. 9 Please refer to the Board Approved Transmission Plan for more details on projects identified to address concerns in the Greater Fresno area. CAISO/M&ID/KW 18 January 14, 2015

19 V. Implementation Costs In the recent FERC order which granted the ISO one additional year to either disaggregate or seek further relief from disaggregating, FERC stated that the ISO needs to provide properly documented implementation cost estimates. On October 28 th, the ISO issued an information request asking stakeholders to provide implementation cost estimates. The cost estimates were for nine categories (shown in the table below) and four levels of disaggregation, slight disaggregation, load aggregation to minimize error (assume 23 LAPs), customized LSE specific LAPs, and fully nodal. Stakeholders were also asked to indicate which costs would be a one-time implementation cost and which would be on-going annual expenses. Eight stakeholders provided cost estimates. The ISO also developed cost estimates. Table 1 below summarizes the estimated implementation costs by the nine categories for the four levels of disaggregation. Total implementation costs for the stakeholders and the ISO are provided for each level of disaggregation broken out by one-time implementation and on-going annual costs. On the low end, i.e. slight disaggregation, it is estimated to cost seven 10 stakeholders and the ISO a combined total of $21.75 million in one-time costs and $2.5 million in annual expenses. If the ISO were to go fully nodal, it is estimated to cost stakeholders and the ISO a combined total $147.2 million in one-time costs and $12.6 million each year. Other Business Integration Costs, specifically for one-time costs, are a significant portion of the estimated costs. Costs included in this category are project management, contingency, and Allowance for Funds used during Construction (AFUDC). Besides the other cost category, the majority of costs, both one-time and annual, are for load forecasting and metering/telemetry. Actual implementation costs for all market participants are expected to be higher because the total estimates presented below only include those stakeholders that provided cost estimates. 10 Eight stakeholders and the ISO provided estimates for LAPS to minimize error and Fully Nodal. Seven stakeholders and the ISO provided estimates for Slight Disaggregation and Custom LSE specific LAPs. CAISO/M&ID/KW 19 January 14, 2015

20 Table 1 Estimated Implementation Costs 11 One time and capital costs Yearly Costs Load Forecasting $ 3,910, ,492 Metering and Telemetry $ 5,134, ,164 Price Forecasting $ 649, ,000 Bidding and Scheduling $ 732,500 85,164 Settlements and Billing $ 1,344, ,576 Demand Response $ 120, ,000 CRR Procurement/Settlement $ 269,000 91,740 Data Integration and Storage $ 933, ,000 Other Business Integration Costs $ 8,654, ,000 Total $ 21,748,748 2,468,136 Slight Disaggregation LAPs to minimize error Custom LSE Specific LAPs Fully Nodal $ 7,888,646 $ 22,059,646 $ 1,749,000 $ 1,953,000 $ 5,135,646 $ 1,320,000 $ 377,810 $ 3,377,000 $ 27,915,280 $ 71,776,028 One time and capital costs Yearly Costs $ 692,492 $ 2,349,164 $ 473,000 $ 230,746 $ 275,582 $ 300,000 $ 126,746 $ 307,000 $ 1,582,200 $ 6,336,930 $ 3,773,422 $ 12,140,422 $ 744,000 $ 987,000 $ 2,115,422 $ 550,000 $ 260,810 $ 1,392,000 $ 8,486,250 $ 30,449,326 One time and capital costs Yearly Costs $ 249,820 $ 932,328 $ 85,000 $ 171,328 $ 231,164 $ 50,000 $ 82,328 $ 132,000 $ 111,600 $ 2,045,568 $ 13,024,052 $ 47,940,052 $ 2,339,000 $ 3,486,000 $ 7,196,052 $ 2,640,000 $ 774,900 $ 7,483,000 $ 62,320,090 $ 147,203,146 One time and capital costs Yearly Costs $ $ 1,249,820 $ $ 3,420,328 $ $ 523,000 $ $ 1,028,910 $ $ 406,164 $ $ 575,000 $ $ 282,328 $ $ 1,506,000 $ $ 3,630,700 $ $ 12,622, Entities provided contingency estimates which are included in the Other Business Integration Costs category. For slight disaggregation, contingency estimates totaled $3.725 million for one-time costs and $446 thousand for on-going costs. For LAPs to minimize error, contingency estimates totaled $14.9 million for one-time costs and $1.58 million for on-going costs. For custom LSE specific LAPs, contingency estimates totaled $3.55 million for one-time costs and $111 thousand for on-going costs. For fully nodal, contingency estimates totaled $38.78 million for one-time costs and $3.43 million for on-going costs. CAISO/M&ID/KW 20 January 14, 2015

21 VI. Benefits Moving away from the current aggregation of load zones and average prices could provide benefits to the wholesale market, market participants, and other parties. The potential benefits expected from load disaggregation include the following, which are discussed in more detail below: Provide more accurate wholesale price signals to provide incentives for investment Improved congestion hedging More efficient day-ahead market outcomes by reducing the frequency of uneconomic adjustments Reduce the subsidization of high-price areas by low-price areas. Being able to quantify these benefits in any realistic manner that would allow a direct comparison to costs would be challenging. Furthermore, there are other factors that also influence the size of the benefits. Retail electricity rates in California are established by the California Public Utilities Commission, along with other Local Regulatory Authorities. These retail rates currently don t reflect any locational price differences between regions within the existing DLAPs. Other tools, such as convergence bidding or the ISO s Participating Demand Resource program, allow market participants to already realize some of the benefits from disaggregated load prices. A. Potential Benefits Accurate wholesale price signals to incent investment There is some heterogeneity of prices within the DLAPs, but wholesale load settles at an average of those prices. Customers in those areas with lower priced nodes see prices that are averaged with areas with higher priced nodes. Therefore, it is possible that more accurate geographic price signals may increase response of load to changes in prices of generation. This can in turn lead to increased investment and participation in demand response by load and investment in generation and transmission infrastructure where it is needed most. The potential benefits that could be realized as a result of more accurate price signals are directly related to 1) the extent of heterogeneity of prices, and 2) how those price signals are also transferred to the retail level. The pricing study results have shown that price dispersion within the existing DLAPs, though it does exist, is minimal. Therefore potential benefits gained from more accurate price signals would also be minimal. Furthermore, responses to more accurate price signals would be weakened due to the current regulatory structure. Currently, even if wholesale load was disaggregated on a nodal level, regulatory policy would prevent the disaggregated prices from flowing through to the retail level. Any impact of disaggregating load at the wholesale level would therefore be informational rather than incentivizing. There are existing ISO programs and processes that can already provide some of the benefits more granular load zones would also offer. There is opportunity for some load to currently settle at more CAISO/M&ID/KW 21 January 14, 2015

22 disaggregated levels. For example, proxy demand resources are able to settle at custom created load aggregation points ranging from a single node to the SLAP level. This already incents demand response in the ISO market to participate in higher priced areas. Transmission investment decisions are already based on nodal LMPs through the ISO s transmission planning process. Therefore more accurate price signals would have negligible impact on investment decisions in transmission infrastructure. Improved congestion hedging There are a few important benefits anticipated from the disaggregation of the default LAPs with respect to congestion hedging. The first was articulated by LECG in their February 23, 2005 Comments of the California ISO MRTU LMP Market Design 12. In particular, LECG brought up the circumstance of a vertically integrated load serving entity located within a high-priced area of a default LAP that also has generation sited to meet its load within that constrained region. In that circumstance, the LSE would be paid counter-flow charges for scheduling its generation to meet its load, although those schedules would actually have no impact on congestion. This scenario is important relative to congestion revenue rights (CRRs) because to the extent that such vertically integrated LSEs receive counter-flow payments for financial schedules that provide no counter-flow on the real transmission system, the number of CRRs that can be allocated to other LSEs is reduced, raising their costs. LECG concludes that these cost shifts, or cross-subsidization, can be avoided by defining load zones that are more homogeneous with respect to prices, and correspondingly, with respect to congestion. Secondly, disaggregating the larger DLAPs would allow for the release of more CRRs in the tier 1 of the annual allocation process with better alignment of LSEs CRR awards with the congestion exposure of the load they serve. When determining how many CRRs to release in each tier, the optimization software (the simultaneous feasibility test or SFT) applies fixed load distribution factors (LDFs) to the pricing nodes (PNodes) that make up the sink or load location of each nominated CRR. If the sink is a large DLAP, particularly if there is heterogeneity of prices over the DLAP, then it is more likely that a binding constraint that affects a single constituent PNode of the DLAP will limit the CRR awards in that tier. Using more granular load zones with more homogeneous prices within each load zone as the sinks of nominated CRRs will thus eliminate these large area DLAP constraints and thereby allow more CRRs to be allocated in tier 1. In recognition of this fact, the current CRR allocation process allows LSEs to nominate sub-lap sinks starting in tier 2 of the annual process and in the monthly allocation process to enable a larger quantity of CRRs to be awarded. This sub-lap provision is somewhat of a compromise, however, because although it allows more CRRs to be awarded, those CRRs will settle based on sub-lap prices that do not align accurately with default LAP-based congestion charges that will be applied in settlement of the load. Thus moving to more granular and more price-homogeneous load zones will both increase the amount of CRRs that can be released and eliminate the current misalignment of sub-lap CRRs with load settlement. 12 See pages of the report, which is appended to Harvey s testimony: CAISO/M&ID/KW 22 January 14, 2015

23 Related to aligning LSE s congestion exposure with CRR awards is the impact it would have on CRR shortfalls. CRRs are funded through revenues from IFM congestion charges. IFM Congestion charges are calculated as the IFM [marginal congestion component] amount for all scheduled Demand and Virtual Supply Awards minus the IFM [marginal congestion component] amount for all scheduled Supply and Virtual Supply Awards 13. When the marginal congestion component (MCC) is lower at the DLAP where demand is settled, it reduces the IFM Congestion charges. Therefore, the amount available in the CRR balancing account to fund CRRs is reduced, possibly contributing to CRR revenue shortfalls. Having load settle at a more granular level would better align the IFM congestion charges with CRR payments, thus potentially reducing CRR revenue shortfalls. The quantity of congestion hedging benefits gained with more granular load zones is likely to be minimal for a couple reasons. First, LSEs have the ability to be allocated more CRRs after the first tier of the annual allocation process by nominating CRRs sinked at Sub-LAPs. Therefore, LSEs already are, or have the ability to, realize some of the benefits more granular load zones would offer. Secondly, higher price differences, specifically nodal LMPs with higher MCCs relative to the DLAP LMPs, will have a greater impact on CRR revenue shortfalls. As the pricing study results indicate, the price differences have been minimal, therefore any benefit gained in terms of CRR revenue adequacy will also be minimal. More efficient day-ahead market outcomes More granular load zones could improve the solution of the integrated forward market (IFM) optimization. Currently, in the day-ahead market, the optimization may have to adjust load to solve a constraint. When load is adjusted, it is adjusted at the DLAP Level. All nodes within the DLAP move up and down in lockstep according to their LDFs until the constraint is solved. For example, the optimization may be forced to decrease load by 100MW at the DLAP level to get a 5MW change on a constraint within the DLAP. If the large DLAPs are disaggregated, the optimization may be able to adjust load at an individual node by a fraction of the amount to get the same 5MW change. This may allow the IFM optimization to reach a more precise solution within each individual load zone. Adjusting bid in load in the day-ahead market to solve a constraint occurs infrequently, therefore the expected benefits of a more efficient market outcome will likely be trivial. Furthermore, convergence bidding can also be used to achieve the same benefit. Market participants can submit economic virtual supply bids at nodal locations. The day-ahead market, if economical, will adjust the virtual supply bids prior to adjusting bid-in load at the DLAP level. Therefore if sufficient virtual supply bids are submitted, the market will not have to rely on adjusting bid-in load at the DLAP to solve a constraint. Reduce subsidization of high priced areas by low priced areas Moving away from averaging wholesale prices for load across large areas with heterogeneous nodal prices would reduce the subsidization of high-price areas by low-price areas. Reducing this cross- 13 CAISO Tariff section CAISO/M&ID/KW 23 January 14, 2015

24 subsidization would provide increased incentives for load to locate in low priced areas and for load in high priced areas to undertake actions to reduce the price of power in those areas, possibly by increasing transmission capacity to the high cost areas to remove congestion or to increase supply in the constrained areas. Again, the potential benefit will only be fully realized with changes to the current retail rate structure allow the wholesale price signals to flow through to end-use customers. As previously mentioned, the probability of this occurring is nominal. B. Quantifying benefits Quantifying each benefit in a meaningful and realistic manner would be challenging. The main difficulty is constructing a plausible counter-factual to the current system. To conduct any reasonable benefit study, several assumptions would have to be made that are difficult to defend. In addition, estimated benefits would not align with all the levels of disaggregation of the cost estimations. Therefore a direct comparison of implementation costs to benefits for most of the levels of disaggregation would not be feasible. Further, throughout the current stakeholder process, stakeholders have noted that most of the aforementioned benefits can be realized through other market products and processes. They have indicated that any benefit estimations should only measure the benefits that are incremental to the amount of each benefit that can be gained through these other products and processes. For example, an LSE can nominate CRRs sinked at sub-laps after Tier 1 of the annual allocation process to increase their CRR holdings. Market participants can also use convergence bidding to capture, and overtime minimize, differences in prices between the nodal LMPs and DLAP LMPs. Lastly, while some customers might see these locational prices, the retail rate structure for most of California, as determined by the CPUC, has uniform rates across the IOU service territories for different customer classes. Analyzing the full benefits of more granular prices also has to take into account the probability of the nodal prices being transferred to end use customers as well. Given the minimal price dispersion found in the pricing study, any estimated benefits would be minimal as well. The benefits would further be reduced if the ISO were to estimate benefits incrementally from other market products and processes, or take into account the probability of changes in the current retail rate structure. Therefore, at this time, the ISO does not believe the benefits of more granular pricing would exceed the cost of implementing. In addition, any quantified benefits would be based on several contentious assumptions that would be difficult to defend and leave the results open for debate. The ISO is open to receiving feedback on studies that could potentially quantify benefits of disaggregated load prices in a realistic and meaningful manner. VII. Second level of aggregation The FERC order indicates that the future ISO filing must include an analysis of a reasonable range of different alternative levels of disaggregation. FERC further provides, in a footnote, For example, the CAISO/M&ID/KW 24 January 14, 2015

25 study could include results using the current LAPs in comparison with aggregation using sub-lap groupings, sub-laps on their own, or a more granular level including individual nodes. The ISO previously identified three other levels of disaggregation, in addition to nodal, to potentially study: slight disaggregation, aggregation to minimize error, and custom LSE specific LAPs. The ISO also recognized that any actual level of disaggregation would need to be determined after the pricing study because other aggregations may become apparent through the results. The pricing study results indicate that, short of nodal, there is no other logical level of disaggregation to analyze. The magnitude of benefits that could be gained through disaggregation is directly related to the magnitude of price dispersion. Nodal LMPs would theoretically provide the greatest benefits. Since the nodal analysis shows minimal price dispersion, potential benefits will also be minimal, and therefore any less granular level of aggregation will have even fewer benefits. At this time, the ISO does not believe conducting analysis for a second level of aggregation will provide any additional information that could be utilized. VIII. Next Steps The ISO will discuss this Pricing Study Results Paper with stakeholders during a conference call on January 21, Stakeholders should submit written comments by January 30, 2015 to lgr@caiso.com. A draft straw proposal will be posted mid-february, which will be followed by a stakeholder meeting currently scheduled for March 3, CAISO/M&ID/KW 25 January 14, 2015