(U 338-E) 2018 General Rate Case A Workpapers. 3 rd ERRATA. T&D- Grid Modernization SCE-02 Volume 10

Transcription

1 (U 338-E) 2018 General Rate Case A Workpapers 3 rd ERRATA T&D- Grid Modernization SCE-02 Volume 10 September 2016

2 124a Customer Interruption Cost Analysis Results Year 1 Customer Interruption Cost Analysis Results Southern California Edison (SCE) retained Nexant to estimate the costs that customers incur as a result of sustained and momentary interruptions 1 on its system. Using three years of historical power interruption data from SCE, this report provides lower and upper bound estimates of the average cost per customer minute interrupted (CMI) and the average cost per momentary interruption for each customer. Table 1-1 shows the results of the customer interruption cost per CMI analysis (in 2016$). The lower bound of the cost per CMI estimate is based on the following nationwide study: Lawrence Berkeley National Laboratory (LBNL) Report No. LBNL-6941E, Updated Value of Service Reliability Estimates for Electric Utility Customers in the United States. 2 The upper bound is based on a comparative analysis of the LBNL report and the 2012 systemwide customer interruption cost survey by Pacific Gas & Electric Company (PG&E). 3 This comparative analysis found that the nationwide LBNL study underestimates interruption costs in California by about 40%. The lower bound of the average cost per CMI for SCE s service territory across the three years analyzed is $ ($907,530,958954,553,754 average total estimated cost / 546,328,512550,198,137 average total CMI). This is 40% lower than the upper bound of $ cost per CMI. The Customer Average Interruption Duration Index (CAIDI) represents the average duration a customer can expect for a given interruption. In other words, a customer in SCE s territory has, on average, an expected duration of minutes for any interruption (546,328,512550,198,137 total CMI / 4,665,0564,657,020 total number of customers affected 4 ). Total # of Interruptions Table 1-1: Interruption Costs per Customer Minute Interrupted Avg. # of Customers Affected Total CMI CAIDI (Minutes) Total Estimated Costs (2016$) Cost per CMI (2016$) Lower Upper , ,781, $877,824,659$890,853,398 $1.72$1.74 $2.88$ , ,851, $946,475,040$960,908,574 $1.68$1.71 $2.83$ ,07327, ,352,948577,961, $898,293,174$1,011,899,290 $1.59$1.75 $2.66$2.94 Average 25, ,328,512550,198, $907,530,958$954,553,754 $1.66$1.73 $2.79$2.91 Table 1-2 shows the results of the customer cost analysis for momentary interruptions. The lower bound of the average cost per momentary interruption per customer across the three years analyzed is $68 67 ($445,698,777451,460,362 average total estimated cost / 6,635,703 total number of customers affected). This is 40% lower than the upper bound of around 1 SCE defines a momentary interruption as an interruption lasting five minutes or less. 2 Sullivan, M.J., J. Schellenberg, and M. Blundell (2015). Updated Value of Service Reliability Estimates for Electric Utility Customers in the United States. Lawrence Berkeley National Laboratory Report No. LBNL-6941E. 3 Sullivan, M.J., J. Schellenberg and others (2012). Pacific Gas & Electric Company s 2012 Value of Service Study. Filed in PG&E s General Rate Case. 4 Total number of customers affected is not shown in Table 1-1, but can be approximated by multiplying the total number of interruptions by the average number of customers affected. 124a

3 125a Customer Interruption Cost Analysis Results Year $ Although there were fewer momentary interruptions compared to sustained interruptions, each momentary interruption affected more customers on average. The Momentary Average Interruption Frequency Index (MAIFI) represents the average number of momentary interruptions a customer experiences during a year. Overall, any SCE customer can expect to experience a little over one momentary interruption per year (6,635,703 total number of customer interruptions / 5,005,487 total system customers). Total # of Momentary Interruptions Table 1-2: Customer Interruption Costs per Momentary Interruption Total # of Customer Interruptions Avg. # of Customers Affected MAIFI Total Estimated Costs (2016$) Cost per Momentary Interruption per Customer (2016$) Lower Upper ,902 5,972, $410,614,837$416,005,282 $68.8$69.7 $115.4$ ,460 6,802, $456,181,510$462,049,113 $67.1$67.9 $112.5$ ,206 7,132, $470,299,984$476,326,690 $65.9$66.8 $110.6$112.0 Average 7,523 6,635, $445,698,777$451,460,362 $67.2$68.0 $112.7$ Application of PG&E Study to Estimate Upper s In 2012, Nexant conducted a systemwide customer interruption cost survey among PG&E s residential and business customers. To assess the extent to which the LBNL meta-analysis underestimates customer interruption costs in California, Nexant conducted a comparative analysis of the 2015 LBNL study and the 2012 PG&E study. This comparative analysis found that the nationwide LBNL study underestimates interruption costs in California by about 40%. While the lower bound of the cost per CMI estimate is based on the nationwide 2015 LBNL meta-analysis, the upper bound adjusts for the 40% underestimate that was identified in the comparative analysis. The methodology is described in more detail in Section 2.3. This underestimate may be due to the higher costs of living and of doing business in California. While the nationwide 2015 LBNL meta-analysis adjusts for the varying distribution of industries, customer usage, demographics, and some other factors across states, it does not include variables that adjust for general economic drivers such as varying costs and productivity across states. Therefore, the results from the 2012 PG&E are used as a proxy for the extent to which the LBNL study underestimates customer interruption costs in California. Given that the LBNL study may not underestimate costs as much for SCE, this estimate is characterized as an upper bound. 125a

4 126a 2 Nexant analyzed historical power interruptions and customer level data to estimate the total cost of interruptions for SCE customers, cost per CMI, and cost per momentary interruption. The remainder of this section provides a brief overview of the methodology used to quantify interruption costs and summarizes the analysis that was done using that methodology in this specific application. 2.1 Estimating Customer Interruption Costs The preferred method for estimating customer interruption costs is a survey that describes several hypothetical interruption scenarios and asks customers to detail the costs that they would experience under those conditions, as described in the Electric Power Research Institute s Outage Cost Estimation Guidebook. 5 Various parties have proposed alternative approaches for estimating customer interruption costs. The strengths and weaknesses of each approach are described in a literature review for the National Association of Regulatory Utility Commissioners. 6 As discussed in that review, customer surveys are the preferred method for estimating customer interruption costs because they directly measure the costs that customers experience under a variety of interruption scenarios without relying on the relatively weak assumptions that alternative methods use. The primary drawback of surveys is that they require collecting detailed information from large, representative samples of residential, commercial, and industrial (C&I) customers. As a result, only a few of the largest utilities in the U.S. have conducted customer interruption cost surveys. To address this barrier, the Department of Energy (DOE), LBNL, and Nexant have been working together for over a decade to make reasonable interruption cost estimates readily available for utilities that have not conducted their own surveys. The first step in developing a national estimate of interruption costs was to combine results from all of the interruption cost surveys that were conducted using the methods outlined in the Outage Cost Estimation Guidebook. This aggregate statistical study, called a meta-analysis, was first done in 2003 (with results from 24 surveys) and then updated in 2009 and 2015 (with results from 34 surveys, including the original 24) Analysis of Historical Interruptions In this analysis, Nexant applied the econometric models from the 2015 meta-analysis of nationwide customer interruption costs. These models are the same equations that serve as the basis for the Interruption Cost Estimate (ICE) Calculator, which is a publicly-available, online tool that estimates customer reliability benefits associated with user-specified reliability improvements that arise from smart grid and other types of investments. 8 In this case, Nexant 5 Sullivan, M.J., and D. Keane (1995). Outage Cost Estimation Guidebook. Report no. TR Palo Alto, CA: EPRI. 6 Sullivan, M.J., and J. Schellenberg (2011). Evaluating Smart Grid Reliability Benefits For Illinois. National Association of Regulatory Utility Commissioners Report. 7 Sullivan, M.J., J. Schellenberg, and M. Blundell (2015). Updated Value of Service Reliability Estimates for Electric Utility Customers in the United States. Lawrence Berkeley National Laboratory Report No. LBNL-6941E. 8 Available here: Developed by DOE, LBNL and Nexant. 126a

5 127a did not use the ICE Calculator itself because SCE was able provide granular customer information and historical power interruption data to conduct a more robust, disaggregated analysis that fully leverages the econometric models from the 2015 meta-analysis. Nexant analyzed both sustained and momentary interruptions specifically for the past three years ( ), and linked this historical interruption information with SCE s customer database to customize the interruption cost estimates to the specific circuits that were affected by each interruption. The following outlines the main steps of the analysis: 1. Generate key model inputs using SCE s customer database; 2. Estimate interruption costs for all customers on a given circuit; 3. Combine individual customers interruption costs with the historical interruption data at the circuit level; 4. Interpolate to generate interruption costs for every historical interruption and scale each cost by the number of customers affected; and 5. Aggregate the individual costs to generate total interruption costs for all sustained and momentary interruptions for each year, Generate key model inputs Using customer level characteristics data, the econometric models produce interruption cost estimates for various scenarios for each circuit in SCE s territory. The models can estimate costs for any interruption lasting up to 960 minutes. Key inputs include customer class (residential, small C&I and medium and large C&I), usage (annual kwh 9 ), and industry type (for C&I customers only), since average interruption costs vary significantly across these factors. These inputs vary with the distribution of customers at each circuit, which is captured in the customer characteristics data. The result is estimates for interruption costs of various durations, seasons, and times of day, specific to the customer mix on each circuit. This information is combined with the database of historical interruptions to estimate total interruption costs, total CMI, and cost per CMI, and cost per momentary interruption Estimate interruption costs for individual customers and combine with historical interruption data The econometric models use the inputs generated from the customer characteristics data to estimate interruption costs for all customers on a given circuit for a fixed set of interruption scenarios. 10 The resulting dataset contains estimates for interruption costs of various durations, 11 seasons, and times of day that are specific to the customer mix on each circuit. 9 The interruption cost is assumed to be zero for customers with zero annual kwh. 10 An interruption scenario is defined by a duration, time of day the interruption occurs, and the season in which the interruption occurs. 11 Due to limitations in the underlying survey data from the meta-analysis, the econometric models cannot reliably estimate costs for interruptions longer than 16 hours (960 minutes). Therefore, interruptions longer than 960 minutes were capped at 960 minutes. 127a

6 128a Nexant received data on every interruption that occurred in SCE s service territory for the years 2013 through This database contained key attributes of each interruption, including date and time of the occurrence, interruption duration, the circuit it occurred on, the total number of customers affected, and an indicator for momentary interruptions. This information was then combined with estimates of individual customer interruption costs from the econometric models at the circuit level. The result is a cost estimate for every historical interruption on a given circuit, but for the set of interruption durations used as inputs in the econometric models Interpolate and aggregate to generate total annual costs In order to generate cost estimates for every historical interruption given the actual durations, the previous dataset is used to interpolate the missing cost estimates for all durations ranging from zero to 960 minutes. This step produces interruption cost estimates for all historical interruptions that reflect their true interruption durations. The cost estimates are then scaled by the ratio of the number of customers affected by the interruption to the total number of customers served by that circuit. This step assumes that the customer mix affected by the interruption is the same as the distribution of customers on that circuit as a whole. After the costs are scaled, each interruption is flagged as either a sustained or momentary interruption, and the costs are summed to generate aggregate interruption costs per year by interruption type. This step also aggregates the data to generate the other interruption metrics shown in Tables Table 1-1 andtable 1-2 (e.g., total number of interruptions, total CMI, and overall MAIFI). 2.3 Application of PG&E Study to Estimate Upper s The 2015 meta-analysis is based on data that mostly includes utilities from areas of the country that are known to have lower costs of living and of doing business than in California. Therefore, the customer interruption costs based on this meta-analysis are characterized as a lower bound. To assess the extent to which the LBNL meta-analysis underestimates customer interruption costs in California, Nexant conducted a comparative analysis of the 2015 meta-analysis and the 2012 PG&E systemwide customer interruption cost study. As in the analysis described in Section 2.1, Nexant estimated interruption costs for PG&E customers using the 2015 metaanalysis. Nexant then compared those results to the average customer interruption costs from the 2012 PG&E survey for a 4-hour interruption on a summer weekday afternoon. 13 Table 2-1 summarizes the results of this comparative analysis by customer class. The estimated interruption costs from the 2015 meta-analysis are nearly 45% lower than the PG&E survey results for residential customers, around 39% lower for small/medium businesses and nearly 57% lower for large businesses. Overall, the population-weighted average cost for this interruption scenario is $743 per PG&E customer based on the survey. The meta-analysis produces a population-weighted average cost estimate that is 40.4% lower, at $443. The upper 12 To avoid double counting, Nexant removed the July 15, 2015 Chestnut outage record from the analysis, given that those customer interruption minutes were distributed to other records in the dataset. 13 This power interruption scenario is used in the comparative analysis because it is the scenario that is most common across both studies. 128a

7 129a bound estimates of cost per CMI and the cost per momentary interruption in Tables 1-1 and 1-2 were adjusted for this underestimate. Given that the LBNL study may not underestimate costs as much for SCE as it does for PG&E in this comparative analysis, these estimates are characterized as an upper bound. Table 2-1: Comparison of 2012 PG&E Survey Results and Estimated Customer Interruption Costs from 2015 Meta-Analysis (4-hour Interruption on a Summer Weekday Afternoon) Customer Class PG&E Population Average Interruption Cost from PG&E Survey Estimated Interruption Cost from 2015 Metaanalysis % Underestimated Residential 4,196,099 $14.70 $ % Small/Medium Business 360,850 $8,700 $5, % Large Business 1,225 $152,914 $65, % Overall 4,558,174 $743 $ % 129a