17. Settlement Baseline for Utility Programs
17.1. Current Baseline for Settlement and Studies
Some demand response programs pay a customer for reducing its usage during demand response events. Since a customer's meter only measures how much energy a customer actually used, the reduction in energy use must be estimated by estimating how much energy the customer would have used if a demand response event had not been declared. The estimate of energy use in the absence of a demand response event is referred to as the "baseline."170 An important issue in this proceeding is determining whether the existing baseline methodology for utility programs should be changed, and, if so, to what alternative baseline method.
The utilities currently use a "3-in-10 unadjusted baseline" method in most demand response programs. This method takes the average of a customer's energy use during the three highest days out of the last 10 business days. The 10 business days exclude any event days and holidays. The baseline is "unadjusted" in that the calculated average is not adjusted up or down based on the usage the morning of the event. In these applications, the utilities offer various proposals for changing the settlement baseline in the 2009-2011 period.
17.2. Utilities' Baseline Proposals
For customers with loads greater than 200 kilowatts enrolled in most of its demand response programs, PG&E proposes to replace the 3-in-10 unadjusted baseline with a "10-in-10 adjusted baseline," under which a customer's baseline would be calculated as the average of that customer's energy use during the 10 previous non-event business days, adjusted up or down based on the customer's usage in the four hours immediately before the event, a so-called "default morning-of adjustment."171 PG&E proposes allowing customers to opt out of the adjustment, in which case the baseline would be the average of the 10 previous non-event business days. PG&E proposes that customers with loads under 200 kilowatts should not be offered the morning-of adjustment, stating that it would not be feasible to develop, administer, and implement for mass market customers during 2009 though 2011.172 PG&E proposes this new 10-day baseline approach be phased-in starting in 2009, applied first to its PeakChoice program, Optional Binding Mandatory Curtailment Program, Capacity Bidding Program, and Business Energy Coalition Program. PG&E would maintain the current 3-in-10 baseline for ABEC and Demand Bidding Program.173
In contrast, SCE proposes to retain the current 3-in-10 day baseline as a default baseline for its Energy Option program, but provide customers with the ability to choose a 10-in-10 day baseline with a day-of adjustment.174 SCE does not specify how it would calculate its day-of adjustment used with its 10-in-10 baseline, or whether SCE intends to extend this baseline option to other programs. SDG&E does not propose changes to its existing baseline methodology.
17.3. Party Positions on Appropriate Settlement Baselines
CDRC proposes to use an individual baseline for aggregated programs. In this method, the hourly loads for each of an aggregator's customers are used separately to identify that customer's highest 3-in-10 days (or 5-in-10, or 10-in-10, depending on the methodology). The average loads over those three days (or five or ten) are calculated, and then the individual customer baseline loads are summed up to produce the total aggregator baseline load for each event-type day. The resulting sum of individual baselines is then compared to the actual sum of the usage of those same customers. Nonetheless, CDRC argues that no party has argued that individual baselines are not preferable for customers.
In contrast, utilities use an "aggregated baseline" for all customers enrolled in a demand response program through a third-party aggregator. This allows the utilities to compensate the performance of the aggregated resource as a whole. Under the "aggregated baseline" method, the hourly load for all of an aggregator's nominated customers are summed, and the resulting aggregator loads are used to identify the three days in the past 10 (or 5-in-10, or 10-in-10) in which total usage of all customers enrolled through that aggregator was highest. The average loads over those three days (or five or ten) are calculated. The resulting aggregator baselines are then compared to the actual aggregator load for each of the event days.
In its opening brief, SCE asserts that the use of individual baselines for aggregated programs would overstate load reduction and, in some scenarios, could result in a payment for a load reduction that did not occur. SCE claims, "Unbundling the estimation of baseline through the individualized approach recommended by the CDRC distorts the impact of the aggregated resource by `cherry picking' individual customers for the usage on different peak days that maximizes their individual contribution, not the coincident contribution of the aggregated resource. This unbundling effect would relieve the aggregator of its responsibility to manage the aggregated resource for the coincident impact and unfairly reward passive resources."175 PG&E agrees with SCE, and argues that aggregators have access to PG&E's meter data and can use it to verify each customer's individual performance. Both PG&E and SCE argue that aggregators have the ability to provide individual customer baselines to assist these customers, if appropriate.176 PG&E and SCE state that aggregators are compensated to manage the customers they enrolled, and that utilities do not know how the aggregators compensate their individual customers.177 PG&E does acknowledge an ongoing study by Christensen Associates that examines the issue of individual baseline versus aggregated baseline, and suggests that any decision in favor of an individual baseline should be deferred until the outcome of the study.178 This study, which was filed with the Commission in April 2009, has not been entered into the proceeding record, and therefore is not considered here.
TURN and CDRC disagree with some or all of the utilities' baseline proposals in these applications.179 TURN objects to SCE's proposal to retain the current 3-in-10 day baseline as a default, and urges the Commission to require ongoing monitoring of baseline accuracy. TURN notes that the 2008 study by Christensen Associates finds that the unadjusted 3-in-10 baseline tends to overstate actual loads on event days for many SCE customers, thereby promoting free ridership and resulting in higher payments than appropriate to some customers.180 TURN proposes that the Commission implement the finding of the Christensen study181 that recommends customers with high load-variability should be guided toward demand response programs that do not require baseline calculations, such as Critical Peak Pricing.
CDRC proposes a 5-in-10 baseline with an asymmetrical (upward only) day-of adjustment capped at 20%. PG&E argues that the 5-in-10 baseline is not recommended by any studies. CDRC acknowledges that research does not indicate that one methodology is superior, but proposes a 5-in-10 baseline as a "compromise" between the 3-in-10 baseline currently used and the 10-in-10 baseline proposed by PG&E. In a reply brief, PG&E acknowledges that the 5-in-10 (adjusted) method has its own merits under certain situations, but the KEMA report cited by CDRC recommends overall that the adjusted 10-day model with a two-way adjustment be used in most situations. Both SCE and PG&E assert that the consideration of this should await the results of PG&E's baseline pilot, which tested a temperature-sensitive baseline using a "morning of" adjustment to the current "3-of-10" baseline methodology, capped at 20%, similar (though not identical) to the baseline adjustment CDRC proposed in this proceeding. In comments on the proposed decision, CDRC reiterates its recommendation, asserting that a 10-in10 day baseline may significantly understate usage for certain customers.
PG&E proposes to adjust the 10-in-10 baseline based on customer usage during the four hours prior to the beginning of an event. PG&E argues that the hours immediate preceding the event should be used (and not hours further removed from the event start time) in order to avoid gaming. Gaming of a baseline that allows adjustment of a calculated daily average for usage before an event can occur if a customer deliberately increases its load in the morning before the event to inflate the baseline. PG&E believes that using a relatively large period of time (e.g., four hour) to calculate the adjustment discourages gaming by making gaming more costly to participants, whose electric bill would increase during the whole adjustment period.
CDRC proposes to use a five-hour window prior to the event start time, and to use only the first three hours of that window to determine the adjustment. CDRC asserts that some customers are penalized under PG&E's suggested four-hour adjustment because they begin curtailment actions before the beginning of an event, for example ramping down manufacturing or lighting to ensure load drop at the very beginning of an event.182 PG&E argues that this window period used to calculate the adjustment is too far from the event window, and will lead to a less accurate adjustment.183 PG&E does not oppose CDRC's proposal in concept, but believes there should be a shorter gap between the adjustment calculation window and the event window, such as using the first three hours of a four-hour window prior to the event.184 In its reply brief, CDRC believes that PG&E's recommendation to measure the morning-of adjustment period using the first three of the four hours prior to the event is reasonable.185 In comments on the proposed decision, CDRC further recommends that if a 10-in-10 day baseline is used (rather than CRDC's preferred 5-in-10 methodology), the maximum adjustment should be raised from plus or minus 20% to plus or minus 35%, to offset the possibility that the 10-in-10 day baseline may understate certain customers' usage.
17.4. Baseline Recommendations
A properly designed baseline calculation methodology is important for the success of any demand response program as it provides the benchmark by which performance is measured. A methodology that systematically over-estimates "business as usual" loads will over-value the contribution of a demand response resource and pay a customer for demand reductions that did not actually occur. Conversely, a baseline methodology that under-estimates the "business as usual" loads will under-value the demand response reduction provided by a customer and not provide the appropriate compensation.
Two studies186 have examined the performance of various baselines in recent years. The studies uniformly suggest there are better baselines than the current three-day unadjusted baseline for the large commercial and industrial customers. The studies also conclude that a day-of adjustment based on usage data from the morning before an event can significantly reduce the bias and improve the accuracy of this type of baseline. The regression methodology used by SDG&E is generally accepted to be reasonably accurate, but has the disadvantage of being complex and costly to calculate, and difficult for participants to understand.
Based on the record of this proceeding, including various studies, there is no one single baseline that will provide accurate settlement calculations for all customers. The KEMA 2003187 study suggested that a good baseline for settlement should be simple to calculate, unbiased, predictable to customers prior to an event, and minimize the possibility of gaming. Both KEMA 2003 and Quantum 2006 studies recommend a 10-day baseline with a day-of adjustment. This approach calculates an average for each hour, using the last 10 weekdays prior to an event, excluding any event days and holidays prior to the event. The day-of adjustment is a ratio of (a) the average load of certain hours before the event to (b) the average load of the same hours from the last 10 weekdays, excluding event days and holidays. KEMA suggests that this method performs well for both weather-sensitive and non-weather sensitive customers in its sample.
Based on the record presented in this proceeding, we adopt a 10-in-10 baseline with a day-of adjustment, and require that an individual baseline be used for customers enrolled in a utility demand response program directly through a utility and for customers enrolled in these same programs by an aggregator.188 The adjustment will be symmetrical (upward or downward, as indicated by usage in the window time period), is capped at 20%, and will be based on the first three of the four hours prior to the event. Utilities shall offer customers the opportunity to opt in to the adjustment. This change in the baseline should be applied to Capacity Bidding Program, Demand Bidding Program, Optional Binding Mandatory Curtailment Program, PG&E Peak Choice, and SCE's Energy Options program.
The adopted approach will provide customers with a relatively simple and understandable baseline that minimizes bias and the possibility of gaming by participants. It is reasonable for customers to have their baselines calculated in the same way, whether they enroll in a program through an aggregator or through a utility. Similarly, it is reasonable for customers of SCE, SDG&E, and PG&E to be subject to the same baseline. This will make the baseline methodology more consistent and transparent to customers.
The Commission agrees with TURN's recommendation that in the long term, utilities should attempt to steer customers with highly variable loads away from demand response programs that require baselines, and towards programs that do not require baseline calculation such as Critical Peak Pricing. To facilitate this, we direct the utilities to work with parties for an agreement on the definition of highly variable load customers, and to prepare and file a report in R.07-01-041 or a successor proceeding by September 1, 2010, on the definition of highly variable load customers along with an estimate of the number of highly variable load customers that are currently in its baseline demand response programs, and the number of megawatts contributed to the programs by those customers. This report will also include information on the proportion of customers choosing the morning-of adjustment option that reach or exceed the maximum adjustment of 20%, and how often that maximum adjustment is reached. This information will assist us in reviewing the effectiveness of the 20% adjustment. The report should propose a plan for steering highly variable load customers towards demand response programs that do not require baseline calculation.
170 In contrast, a baseline is not necessary for dynamic pricing rates, which generally charge a customer variable prices for the energy that the customer actually uses.
171 Exhibit 201, pp. 2-29.
172 PG&E 2009-2011 Demand Response Programs and Budgets Amended Prepared Testimony, September 19, 2008, p. 2-30.
173 Demand Bidding Program will be phased out at the end of 2009.
174 SCE Volume I, Amended Testimony in Support of SCE's Amended Application for Approval of demand response Programs, Goals, and Budgets for 2009-2011, September 19, 2008, p. 21.
175 Opening Brief of Southern California Edison Company in A.08-06-001 et al., January 28, 2009, p. 33.
176 SCE Reply Brief, p. 9. PG&E Reply Brief, p. 30.
177 SCE Reply Brief, p. 10. PG&E Reply Brief, p. 31.
178 PG&E Opening Brief, p. 21.
179 Exhibit 705 and Exhibit 420, p. 10.
180 Christensen Associates Energy Consulting, LLC, S. Braithwait, M. Welsh, D. Hansen, and D. Armstrong, "California Day-Ahead demand response Program Baseline Load Analysis and PY-2006 Impact Evaluation - Final Report," June 6, 2008, pp. 8-12.
181 Exhibit 420, p. 10.
182 CDRC Opening Brief, p. 39.
183 PG&E Opening Brief, p. 22.
184 PG&E Opening Brief, p. 22.
185 CDRC Reply Brief, p. 13.
186 Exhibit 210 - Protocols Development for Demand - Response calculations: Finings and Recommendations, Prepared for the CEC by KEMA-Energy. CEC 400-02-017F; Exhibit 211 - Evaluation of 2005 Statewide Large Nonresidential Day-ahead and Reliability Demand Response Programs. Prepared for Working Group 2 Measurement and Evaluation Committee, by Quantum Consulting Inc. and Summit Blue Consulting, LLC 2006.
187 Exhibit 705, Attachment 15.
188 This requirement does not apply to demand response contracts between a utility and an aggregator approved by this Commission that specify a baseline as part of a contract.