BEFORE THE NEW MEXICO PUBLIC REGULATION COMMISSION IN THE MATTER OF SOUTHWESTERN PUBLIC SERVICE COMPANY S APPLICATION FOR REVISION OF ITS RETAIL RATES UNDER ADVICE NOTICE NO., SOUTHWESTERN PUBLIC SERVICE COMPANY, APPLICANT. ) ) ) ) ) ) ) ) ) ) CASE NO. -009-UT DIRECT TESTIMONY JANNELL E. MARKS on behalf SOUTHWESTERN PUBLIC SERVICE COMPANY
TABLE OF CONTENTS GLOSSARY OF ACRONYMS AND DEFINED TERMS... iii LIST OF ATTACHMENTS... v I. WITNESS IDENTIFICATION AND QUALIFICATIONS... II. ASSIGNMENT AND SUMMARY OF TESTIMONY AND RECOMMENDATIONS... III. LOAD RESEARCH... IV. HISTORICAL CUSTOMER AND MWH SALES TRENDS... V. CUSTOMER AND MWH SALES FORECAST... VI. OVERVIEW OF SALES AND CUSTOMER FORECASTING METHODOLOGY... VII. FORECAST DEVELOPMENT... A. DATA PREPARATION... B. STATISTICALLY MODELED FORECASTS... C. CALENDAR-MONTH SALES DERIVATION... VIII. RATE SHEET FORECAST... 9 IX. CONCLUSION... 0 VERIFICATION... ii
GLOSSARY OF ACRONYMS AND DEFINED TERMS Acronym/Defined Term Base load Meaning Sales load component that is not associated with weather Base Period Calendar year 0 C&I Census class Commission DSM DW Global Insight IDR KW KWh MW MWh NCE Non-Census class NSPM Commercial and Industrial Customer class in which all customers have IDRs New Mexico Regulation Commission Demand-side Management Durbin-Watson IHS Global Insight, Inc. Interval Demand Recorders Kilowatt Kilowatt-hour Megawatt Megawatt-hour New Century Energies, Inc. Customer class in which not all customers have IDRs Northern States Power Company, a Minnesota corporation iii
Acronym/Defined Term NSPW Operating Companies PSCo R-squared SPS Meaning Northern States Power Company, a Wisconsin corporation NSPM, NSPW, PSCo, and SPS Public Service Company Colorado, a Colorado corporation Coefficient determination, a New Mexico corporation Test Year Calendar year 0 Total weather load Xcel Energy XES Sales load component that is influenced by weather Xcel Energy Inc. Xcel Energy Services Inc. iv
LIST OF ATTACHMENTS Attachment JEM- JEM- JEM- JEM- JEM- Description Test Year Forecast Retail Customer Counts and Sales (MWh) by Major Class (Filename: JEM-.xlsx) Test Year Forecast Adjustment for Incremental Demand-Side Management Savings at Delivery - MWh (Filename: JEM-.xlsx) Test Year Forecast Adjustment for New Load Growth MWh (Filename: JEM-.xlsx) Test Year Forecast Model Information (Filename: JEM-.xlsx) Testimony workpapers (Filename: JEM-.xlsx) v
Direct Testimony Jannell E. Marks I. WITNESS IDENTIFICATION AND QUALIFICATIONS 9 0 Q. Please state your name and business address. A. My name is Jannell E. Marks. My business address is 00 Larimer Street, Denver, Colorado 00. Q. On whose behalf are you testifying in this proceeding? A. I am filing testimony on behalf, a New Mexico corporation ( SPS ) and wholly-owned electric utility subsidiary Xcel Energy Inc. ( Xcel Energy ). Xcel Energy is a registered holding company that owns several electric and natural gas utility operating companies. Q. By whom are you employed and in what position? A. I am employed by Xcel Energy Services Inc. ( XES ), the service company subsidiary Xcel Energy, as Director Sales, Energy and Demand Forecasting. Xcel Energy is the parent company four wholly-owned electric utility operating companies: Northern States Power Company, a Minnesota corporation ( NSPM ); Northern States Power Company, a Wisconsin corporation ( NSPW ); Public Service Company Colorado, a Colorado corporation ( PSCo ); and SPS (collectively, Operating Companies ). Xcel Energy s natural gas pipeline subsidiary is WestGas InterState, Inc. Xcel Energy also has two transmission-only operating companies, Xcel Energy Southwest Transmission Company, LLC and Xcel Energy Transmission Development Company, LLC, both which are regulated by the Federal Energy Regulatory Commission.
Direct Testimony Jannell E. Marks 9 0 Q. Please briefly outline your responsibilities as Director Sales, Energy and Demand Forceasting. A. I am responsible for the development forecasted sales data and economic conditions for the Operating Companies, and the presentation this information to Xcel Energy s senior management, other Xcel Energy departments, and various regulatory and reporting agencies. I also am responsible for Xcel Energy s Load Research function, which designs, maintains, monitors, and analyzes electric load research samples in the Xcel Energy Operating Companies service territories. Finally, I am responsible for developing and implementing forecasting, planning, and load analysis studies for regulatory proceedings. Q. Please describe your educational background. A. I graduated from Colorado State University with a Bachelor Science degree in Statistics. Q. Please describe your pressional experience. A. I began my employment with PSCo in 9 in the Economics and Forecasting Department. In 9, I became a Research Analyst, and, in 99, I was promoted to Senior Research Analyst. In that position, I was responsible for developing the customer and sales forecasts for PSCo and the economic, customer, sales, and
Direct Testimony Jannell E. Marks demand forecasts for Cheyenne Light, Fuel and Power Company. In 99, when PSCo merged with SPS to form New Century Energies, Inc. ( NCE ), I assumed the position Manager, Demand, Energy and Customer Forecasts. In that 9 position, I was responsible for developing demand, energy, and customer forecasts for NCE s operating companies, including SPS. I also directed the preparation statistical reporting for regulatory agencies and others regarding historical and forecasted reports. In August 000, following the merger NCE and Northern States Power Company that created Xcel Energy, I was named Manager, Energy Forecasting, with the added responsibility for NSPM and 0 NSPW. I assumed my current position in February 00, with the added responsibility for the Operating Companies load research function. Q. Have you attended or taken any special courses or seminars relating to public utilities? A. Yes. I have attended the Institute for Pressional Education s Economic Modeling and Forecasting class and Itron s Load Forecasting Workshops. I have also attended industry forecasting conferences and forecasting stware user group meetings and training classes sponsored by the Electric Power Research
Direct Testimony Jannell E. Marks Institute. I am a member Itron s Energy Forecasting Group and Edison Electric Institute s Load Forecasting Group. Q. Have you testified before any regulatory authorities? A. Yes. I have testified before the New Mexico Public Regulation Commission ( Commission ), the Public Utility Commission Texas, the Colorado Public Utilities Commission, the Minnesota Public Utilities Commission, the North Dakota Public Service Commission, and the Public Service Commission Wisconsin. I have submitted written testimony to the South Dakota Public 9 0 Utilities Commission. My testimony has addressed the issues load research, sales and demand forecasts, and weather normalization.
Direct Testimony Jannell E. Marks II. ASSIGNMENT AND SUMMARY OF TESTIMONY AND RECOMMENDATIONS 9 0 Q. What is your assignment in this proceeding? A. The purpose my testimony is to:. describe SPS s load research function and the load research information that is used for cost allocation and rate design in this proceeding;. describe the historical customer and megawatt-hour ( MWh ) sales trends for SPS s New Mexico retail service territory; and. present and support SPS s New Mexico retail electric MWh sales and customer forecast for calendar year 0 ( Test Year ). In addition, I sponsor or co-sponsor Schedules P-, P-, P-, and Q- SPS s Rate Filing Package. Q. Please provide a summary conclusions and recommendations in your testimony. A. SPS uses information from Interval Demand Recorders ( IDR ) to determine the coincident and non-coincident peak demand for the Census classes, which are the customer classes in which all customers have IDRs. For those customer classes in which not all customers have IDRs, which are typically referred to as IDRs are meters capable recording loads for each interval time.
Direct Testimony Jannell E. Marks the non-census classes, it is necessary to perform load research. Using 9 0 information from the IDR meters for the Census classes and information from the load research samples for the non-census classes, I have provided class coincident and system non-coincident load factors to SPS witnesses Ian C. Fetters and Richard M. Luth, who incorporate those load factors in the cost allocation and rate design they present. I recommend that the Commission approve those load factors for purposes designing rates and allocating costs among classes. The total number retail electric customers in SPS s New Mexico service territory averaged annual growth.% per year from 0 through 0, while retail sales growth has averaged.% per year over the same period time, after accounting for unusual weather. In the Test Year, the number retail electric customers is expected to increase at an annual rate.%, and retail sales are expected to grow.%, over the previous year. The higher rate retail sales growth during the Test Year as compared to the historical average is due to stronger projected economic growth during 0 and the Test Year and expected additions new large Commercial and Industrial ( C&I ) loads. retail jurisdiction. All sales and customer count information cited in my testimony refer to SPS s New Mexico
Direct Testimony Jannell E. Marks 9 0 SPS s forecast customers and sales are developed using industry standard multiple regression modeling techniques and include appropriate adjustments to account for Demand-side Management ( DSM ) and new load growth. SPS relies on a number quantitative and qualitative tests to ensure that its forecasting models and sales projections are statistically valid. Thus, SPS s Test Year customer and sales forecast is reasonable and should be used to set rates in this proceeding. Q. How is your testimony organized? A. Section III provides a description SPS s load research function. Section IV provides historical customer and MWh sales trends. Section V provides the customer and MWh sales forecast. An overview the sales and customer forecasting methodology is provided in Section VI, followed by a detailed explanation the forecast development in Section VII. Section VIII presents a description how the rate sheet forecast is developed.
Direct Testimony Jannell E. Marks III. LOAD RESEARCH 9 0 Q. What is the purpose load research? A. Load research is the systematic collection and analysis customers electrical energy and demand requirements by time--day, month, season, and year. This data, which includes load research samples, is collected and analyzed by customer classes, stratums customer classes, and other subsets customer classes. Load research enables utilities to better understand customers, their consumption patterns, their consumption responses to various factors, and the impact customers energy requirements on the electric utility s system. In addition, load research data is used to develop demand and energy allocators for cost allocation studies and is used in designing rates. Q. What are load research samples? A. It is costly and not feasible to install IDR meters for all customers in all customer classes. Therefore, it is necessary for SPS to develop load research samples to determine the coincident and non-coincident peaks for certain classes. Load research samples are subsets the entire population that SPS surveys in order to estimate the characteristics the entire population.
Direct Testimony Jannell E. Marks 9 0 9 SPS s load research samples are developed using a stratified random sampling method. This technique divides the class interest into smaller groups with like-characteristics. This method effectively reduces the overall variance the class, thereby reducing the sample size. The samples are designed to meet or exceed the 90/0 load research standard specified by the Public Utilities Regulatory Policy Act. Accuracy Level. If sample metering is required, the sampling method and procedures for collecting, processing, and analyzing the sample loads, taken together, shall be designed so as to provide reasonably accurate data consistent with available technology and equipment. An accuracy plus or minus 0 percent at the 90 percent confidence level shall be used as a target for the measurement group loads at the time system and customer group peaks. Data validation is performed regularly on the load research samples to ensure that the energy use the sample corresponds closely with the population energy use. Q. Does SPS use load research to determine the demand all its customer classes? A. No. It is not necessary to conduct load research for customer classes in which all customers have IDR meters because the IDR meters provide actual measurements Code Federal Regulations, Title, Chapter, Subchapter K, Part 90.0, Subpart B. 9
Direct Testimony Jannell E. Marks 9 0 9 demand. Most the customers with IDR meters are in the Large General Service-Transmission class, although some Primary General Service customers with on-site generation also have IDR meters. In addition, a few the customers with individual rate schedules have IDR meters installed. As noted earlier, I refer to the classes in which all customers have IDR meters as Census classes. SPS uses the output those IDR meters to determine the Census classes demands for purposes allocation, rate design, and billing. Q. For which customer classes has SPS developed load research samples? A. SPS develops load research samples for its non-census classes throughout its service territory in both New Mexico and Texas. SPS developed load research samples for the following New Mexico retail customer classes: Residential Service; Residential Space Heating Service; Small General Service; Secondary General Service; Irrigation Rate Service; Primary General Service; Small Municipal and School Service; and Large Municipal and School Service. 0
Direct Testimony Jannell E. Marks 9 0 Q. How does SPS go about performing the load research? A. It is cost-prohibitive to install an IDR meter for every customer, so instead SPS installs IDR meters on a random sample customers in each non-census class (developed as I previously described) and uses the electric usage data from those sample customers to extrapolate the demand data for the remainder the class. Q. What load research statistics did you provide for SPS s rate design and cost allocation studies? A. For each SPS customer class, I provided: () the load factors at the time the monthly system peak, which is the class coincident peak; and () the load factors at the time the monthly class peak, which is the class non-coincident peak. Q. Please define the terms monthly system peak, class coincident peak, monthly class peak, and class non-coincident peak. A. The monthly system peak is the 0-minute interval in each month in which SPS s system experiences the highest demand, and each class s demand during that 0-minute interval is the class coincident peak. The monthly class peak is the 0-minute interval in each month in which a class experiences its highest demand. Unless the monthly class peak occurs during the same 0-minute interval as the monthly system peak, the monthly class peak is a class non-coincident peak.
Direct Testimony Jannell E. Marks 9 0 Q. What is a load factor? A. A load factor is the ratio the average load in kilowatts ( kw ) supplied during a designated period to the peak or maximum load in kw occurring in that period. For example, assume a customer used 0,000 kilowatt-hours ( kwh ) during a 0-day period (0 hours) and had a maximum demand kw during this same period. The customer s average load would be.9 kw (0,000 kwh / 0 hours =.9 kw). Dividing that number by kw leads to 0. (.9 / = 0.). That is then multiplied by 00% to arrive at a load factor %. Q. How did you determine each class s system peak load factor? A. I derived each class s system peak load factor from load research samples or census data. Q. How did SPS use the load factors derived from your load research? A. I provided the class coincident and non-coincident load factors at peak for each month to Mr. Fetters, who used them to develop demand allocators. Mr. Fetters discusses SPS s demand allocators in further detail in his testimony.
Direct Testimony Jannell E. Marks 9 0 Q. How did SPS calculate the demand at the time the monthly system peak and the demand at the monthly class peak? A. As explained by Mr. Fetters, each class s demand at the time the system peak was calculated by applying the monthly system peak load factors derived from the load research to the monthly sales by customer class. Each class s demand at the time the non-coincident peak was calculated by applying the monthly class peak load factors derived from the load research to the monthly energy sales by customer class. Q. Did you make any adjustments to the class demands at the time the monthly system peaks? A. Yes. Because the hourly loads for the sample classes are estimates, the sum each hourly demand, adjusted to generation level, will almost never equal SPS s total system load. To account for this difference, the sample classes were adjusted each month so that the sum all hourly demand equals the hourly system load at the hour SPS s monthly system peak demand. Mr. Fetters describes this process in his direct testimony. Both monthly system peak demand by class and monthly non-coincident class peak demands were adjusted consistent with the proportional allocation process discussed above.
Direct Testimony Jannell E. Marks IV. HISTORICAL CUSTOMER AND MWH SALES TRENDS 9 0 Q. Please discuss SPS s New Mexico historical retail customer and MWh sales growth trends. A. The total number retail electric customers in SPS s New Mexico service territory averaged,0 per month in the -month period ending in December 0 ( Base Period ). Total retail customer counts increased an average, customers per year from 0 through 0, for an annual average growth rate.% per year. The largest class customers is Residential, which averaged 9, customers per month during the Base Period, and represents 9.% SPS s New Mexico retail customers. Residential customer counts averaged an annual growth rate 0.9%, or additions per year during the period from 0 through 0. SPS s New Mexico weather normalized retail electric sales have increased at an average annual rate.% during the period 0 through 0. Residential weather normalized sales have averaged growth.%, while total C&I weather normalized sales have increased at an average annual rate.9% over this same period time. While the average growth rate for total retail sales during this historical period is.% per year, the growth has varied in individual
Direct Testimony Jannell E. Marks years. In 0, total retail sales increased.%. This was followed by growth.% in 0 and.0% in 0. The average annual percent change in customers and MWh sales by customer class during the period 0 through 0 is presented in Table JEM-. TABLE JEM- Historical Customer and Sales Growth by Major Customer Class 0-0 Average Annual Percent Change -months ending December 0 Customer Class Number Customers Weather Normalized Retail Sales % Retail Sales Residential 0.9%.%.% C&I.9%.9%.9% Street Lighting -0.% -0.% 0.% Public Authority.% -.%.% Total Retail.%.% 00.0% 9 0 Q. Why did you choose the 0 through 0 period time to describe historical growth trends? A. The past three years are representative the longer term trend in customer growth, but don t exhibit the individual year volatility seen during a longer historical time period. For example, the average growth in total retail customers
Direct Testimony Jannell E. Marks 9 0 from 00 to 0 is.0%, which is very similar to the.% average growth for the 0 to 0 time period. However, during the 00 to 0 time period, annual growth rates varied between -0.% and.0%, while the annual growth rates during the 0 to 0 time period only varied between 0.9% and.%. Because average growth during the 0 to 0 time period is representative the longer term trend and consistent between years, I determined that the three year period time was appropriate to use to describe growth trends. Q. Please explain what you mean by weather normalized sales. A. In order to calculate sales growth from year to year not influenced by weather, SPS subtracts the estimated MWh impact deviation from normal weather from actual sales, or weather normalizes sales. If weather is hotter than normal in the summer, the normalization process results in weather normalized sales that are lower than actual sales. Conversely, if weather is warmer than normal in the winter, the normalization process results in weather normalized sales that are higher than actual sales. Although historical sales have been weather normalized to assess historical growth trends within this testimony, the actual Base Period sales reported in Schedule P- in this filing have not been weather normalized.
Direct Testimony Jannell E. Marks V. CUSTOMER AND MWH SALES FORECAST 9 0 Q. Please describe the customer categories used in SPS s retail customer and sales forecasts. A. The Residential, C&I, Street Lighting, and Public Authority classes comprise the New Mexico total retail customer and sales forecasts. Q. How are the Test Year customer and sales forecasts used in this proceeding? A. With the exception one type customer, the customer and sales forecasts are used to calculate the following information: (a) the monthly and annual electric supply requirements; (b) Test Year revenue under present rates; and (c) Test Year revenue under proposed rates. The one exception relates to the customer counts for the lighting customers. SPS witness Richard M. Luth utilizes all the MWh forecasts and the customer counts that I provide for his revenue calculations other than the customer counts for the lighting customers (Street Lighting and Area Lighting). For lighting customers, it is necessary for Mr. Luth to determine the individual number lights being billed. My counts, however, reflect the number customers with lights, who may have a single light or multiple lights at one customer location.
Direct Testimony Jannell E. Marks 9 0 Q. What is SPS s forecast New Mexico retail electric sales and customers for the Test Year? A. Attachment JEM- summarizes monthly Test Year retail MWh sales and number customers for each major customer class. Total retail customers are projected to average 0,9 per month for the Test Year. Test Year total retail sales are projected to be,9,9 MWh. Q. How does the Test Year electric customer growth compare to historical customer growth? A. As I stated earlier, New Mexico retail electric customer counts increased at an average annual rate.% from 0 through 0, or, customers per year. Customers are expected to increase at an annual rate.% or, customers in 0. During the Test Year, the total number electric customers is expected to increase by an additional.% or,0 customers compared to expected 0 average customer levels. Q. How does the Test Year retail electric sales forecast compare to 0 electric sales? A. Total electric retail sales are projected to increase by 0.% in 0, and by.% in the Test Year. After increasing.% in 0, Residential sales are
Direct Testimony Jannell E. Marks 9 0 projected to increase.% in 0 and to increase an additional.% in the Test Year. C&I sales gained.0% in 0 and are projected to increase.% in 0 and 9.% in the Test Year. Following a decline 0.% in 0, Street Light sales are expected to increase 0.% in 0 and 0.% in the Test Year. Finally, Public Authority sales declined.% in 0 and are predicted to increase.% in 0 and.% in the Test Year. I will explain the methodologies used to develop the customer and sales forecast in the following section my testimony. Table JEM- provides the New Mexico weather normalized retail MWh sales by customer class for 0, 0, and the Test Year, as well as the 0 growth rate, the Test Year growth rate, and the average annual growth rate for the 0 through 0 time period. Customer Class TABLE JEM- Retail Sales by Major Customer Class (MWh) 0 Weather Normalized Sales Projected 0 Sales 0 % Change from 0 Projected Test Year Sales Test Year % Change from 0 0-0 Average Annual % Change Residential,,900,,.%,,99.%.% Total C& I,,,,.%,09,0 9.%.9% Street Lighting,, 0.%, 0.% -0.% Public Authority,9,.%,.% -.% Total Retail,00,0,, 0.%,9,9.%.% 9
Direct Testimony Jannell E. Marks 9 0 Q. Why is the projected Test Year total retail sales growth rate higher than the 0-0 average historical growth rate? A. The higher total retail sales growth rate during the Test Year compared to the 0-0 historical average growth rate is primarily due to strong expected growth in the Residential and C&I sectors. This growth is driven by generally stronger projected economic growth during 0 and the Test Year and the expected addition new large C&I loads, primarily in the oil and gas industry. As I explain more fully later in my testimony, SPS relies on IHS Global Insight, Inc. ( Global Insight ), an economic forecasting firm, as its source for economic forecasts and analysis. Projected growth in the Residential sector is a function real personal income per household in SPS s -county service area in New Mexico. During the 0-0 time period, the service area real personal income per household increased at an average annual rate.%. Real personal income per household is projected to grow.9% in 0 and an additional.0% in the Test Year. Total employment, which is a key indicator for C&I sales, increased at an average annual rate.% during the 0-0 time period, and is projected to increase.% in 0 and another.% in the Test Year. 0
Direct Testimony Jannell E. Marks Although economic growth is projected to increase in the Test Year, the most significant driver the strong projected growth in the C&I sector is the expected addition new large C&I load resulting from oil and gas extraction and processing activities in the Permian Basin shale play in southeast New Mexico. In addition, new large C&I load is expected in other industries, including mining. Without these new loads, growth in the total C& I sector would be projected at.% in 0 and.0% in the Test Year, which is in line with the.9% average growth during the 0-0 time period.
Direct Testimony Jannell E. Marks VI. OVERVIEW OF SALES AND CUSTOMER FORECASTING METHODOLOGY 9 0 Q. Is the Test Year forecast the same forecast SPS used for its 0 financial budget? A. Yes. The 0 financial budget sales forecast was developed in February 0 and was based on actual customers and sales through December 0 and projected economic and new load growth as discussed previously. Q. Please describe in general terms the methods used to forecast retail sales and customers. A. Preparation the electric sales and customer forecast utilizes econometric forecasting techniques and statistical analyses. The primary forecasting technique used is regression modeling. Regression models are designed to identify and quantify the statistical relationship between historical sales or customers and a set independent predictor variables, such as historical economic and demographic indicators, historical electricity prices, or historical weather. Once this relationship is defined, a forecast is developed by simulating the relationship over the forecast period using projected levels the independent predictor variables. Regression modeling is a well-known and proven method forecasting, and is commonly accepted by forecasters throughout the utility industry. This
Direct Testimony Jannell E. Marks 9 0 method provides reliable, accurate projections, accommodates the use predictor variables, such as economic or demographic indicators and weather, and allows clear interpretation the model. SPS has been using regression models to forecast MWh sales and customers for fifteen years. Q. Were any adjustments made to the forecast model results for the Test Year? A. Yes, several adjustments were made to the forecast model results. First, the forecast models are based on billing-month sales. The billing-month sales are converted to calendar-month sales to align Test Year revenues with the relevant projected Test Year expenses, which have been estimated on a calendar-month basis. The process to convert billing-month sales to calendar-month sales is described later in my testimony, in Section VI, Forecast Development. Next, the Residential and C&I sales forecast model results were adjusted to reflect the expected incremental impact DSM programs. An annual forecast the impact new DSM programs (excluding Saver s Switch) is developed by XES s DSM Regulatory Strategy and Planning Department. The DSM forecast used in the development the sales forecast submitted in this proceeding includes the expected DSM savings in 0 assuming achievement the goal level filed in SPS s 0 Energy Efficiency and Load Management Plan
Direct Testimony Jannell E. Marks 9 0 Settlement in Case No. -00 UT, and the expected savings in 0 necessary to meet the Efficient Use Energy Act requirement % in 00. The projected DSM MWh impacts are converted by class from calendar-month energy to billing-month sales volumes. The resulting sales volumes are used to adjust the class level sales forecasts that result from the regression modeling process. Impacts from all program installations through 0 are assumed to be embedded in the historical data, so only the impact from new program installations are included in the DSM adjustment. SPS s Saver s Switch program results in short-term interruptions service designed to reduce system capacity requirements rather than permanent reductions in energy use, so it is not considered here. Finally, the C&I sales forecast model results were adjusted to include projected new load growth as identified by SPS account managers for load growth that would not be embedded in the historical sales data. These new load growth projections include expected additions for oil and gas production activity in In the Matter s Application for Approval its: (A) 0 Energy Efficiency and Load Management Plan and Associated Programs; (B) Request for Financial Incentives for 0-0; (C) Cost Recovery Tariff Rider; and (D) Request to Establish Lower Minimum Savings Requirements for 0 Under the Efficient Use Energy Act, Case No. -00-UT, Final Order Adopting Certification Stipulation (June, 0).
Direct Testimony Jannell E. Marks 9 0 southeast New Mexico. The new load growth projections were developed in early 0 and reflect the lower oil and gas prices seen since mid-0. Attachment JEM- contains the Test Year adjustment for incremental DSM MWh reductions. Attachment JEM- contains the Test Year adjustment for new load growth. Q. Is the forecast methodology used to develop the 0 Test Year sales and customer forecast the same methodology SPS has utilized in the past? A. Yes, the methodology used for the 0 Test Year forecast is the same methodology SPS has utilized in the past, including the 0 Test Year forecast in Case No. -000-UT. Q. Please compare the 0 Test Year forecast in Case No. -000-UT, to the actuals for 0. A. The 0 Test Year sales forecast in Case No. -000-UT was projected to total,9,00 MWh and average customers were projected to be,0. Weather-normalized 0 actual sales totaled,00,0 MWh and actual customer counts were,0. The majority the variance in customer counts occurred in the Residential class, where actual counts were 90 customers or.0% higher than forecast. The majority the variance in sales occurred in the
Direct Testimony Jannell E. Marks C&I class, where sales were, MWh or.% higher than forecast. The stronger than expected growth in the C&I sector primarily was due to growth in the energy sector. Although the regression model results, in -000-UT, were adjusted to account for growth in the energy sector, the growth was stronger than expected. Because the growth in the energy sector has been stronger than previously expected and new load is also expected in the mining sector, the forecast presented in this testimony has been adjusted by a larger amount, as reflected in the projected 9.% increase in C&I sales in the 0 Test Year.
Direct Testimony Jannell E. Marks VII. FORECAST DEVELOPMENT 9 0 A. Data Preparation Q. Please describe the data and data sources SPS relied on to develop the Test Year sales and customer forecasts. A. Historical billing-month sales, monthly number customers, and billing-month rate revenues by rate class were obtained from SPS billing system reports. Historical electricity prices were calculated by dividing the billing-month rate revenues by total sales volumes. The forecast electricity prices was based on a forecast the Producer Price Index for electric power. Q. What measure weather did SPS use for the forecasts? A. The measure weather used was heating degree days and cooling degree days, using a sixty-five degree temperature base. This information was obtained from the National Oceanic and Atmospheric Administration and was measured at the Roswell, New Mexico weather station. Heating degree days were calculated for each day by subtracting the average daily temperature from degrees Fahrenheit. Cooling degree days were calculated for each day by subtracting degrees Fahrenheit from the average daily temperature. For example, if the average daily temperature was degrees Fahrenheit, then minus, or 0
Direct Testimony Jannell E. Marks heating degree days, were calculated for that day. If the average daily 9 0 temperature was greater than or equal to degrees Fahrenheit, then that day recorded zero heating degree days. The same process applies to cooling weather, with average daily temperatures greater than degrees Fahrenheit recording cooling degree days, and average daily temperatures less than degrees Fahrenheit resulting in zero cooling degree days. The workpaper for Schedule P- provides the historical data and the calculations applied to develop the weather variables used for the forecasts. Q. Did the weather reflect the same billing days as the sales data? A. Yes. The heating degree days and cooling degree days were weighted by the number times a particular day was included in a particular billing month. These weighted heating degree days and cooling degree days were divided by the total billing cycle days to arrive at average heating degree days and cooling degree days for a billing month. Q. Why does SPS use the Roswell, New Mexico weather station to represent SPS s New Mexico service territory? A. SPS uses data from the Roswell, New Mexico weather station because Roswell is a major population and economic center in the SPS New Mexico service territory.
Direct Testimony Jannell E. Marks 9 0 Also, it is located close to the geographic center the SPS New Mexico service territory. Q. What weather assumption was used for the Test Year? A. Normal weather was used for the Test Year, where normal is defined as a thirtyyear rolling average historical values. Daily normal heating degree days or cooling degree days were calculated by averaging thirty years daily heating degree days or cooling degree days using data from 9 to 0. These daily normal heating degree days and cooling degree days were weighted by billing cycle information to derive normal billing-month heating degree days and cooling degree days in the same manner as were the historical actual heating degree days and cooling degree days. Q. What was your source economic and demographic data? A. Historical and forecasted economic and demographic variables for the six-county SPS service territory, the state New Mexico, and the nation were obtained from Global Insight. The variables used in the models were service territory non-farm employment, population, households, and real personal income; New Mexico real Gross State Product; and U.S. real Gross Domestic Product and the Industrial Production Index for oil and gas extraction. This information is used to determine 9
Direct Testimony Jannell E. Marks 9 0 the historical relationship between customers and sales measures, and economic and demographic measures. Q. Are you also responsible for developing SPS s peak demand forecast? A. Yes. SPS develops its total system retail peak demand forecast using an econometric model, with monthly historical system retail megawatt ( MW ) peak demand as the dependent variable, and system retail sales, weather concepts, a linear trend, seasonal binary, and month specific binary variables as explanatory variables. For the full-requirements wholesale peak demand forecasts at the delivery point, SPS uses historical monthly load factors to develop monthly peak demands based on the projected monthly sales. As previously explained, a load factor is the ratio average sales over a period time (or average load) to the peak demand during that period time. The formula to calculate the load factors is: Load Factor(%)=[Sales(MWh) / hours per month] / Peak Demand(MW) The peak demand forecasts at the delivery point are then grossed up to at the source estimates by applying loss factors. The projected load factors are based on historical load factors. The monthly load factors and loss factors used for the forecast period are assumed to be the same as the historical load factors and 0
Direct Testimony Jannell E. Marks 9 0 approved loss factors. The partial requirements wholesale peak demand forecast is determined by contractual agreement. B. Statistically Modeled Forecasts Q. Please describe the regression models and associated analyses used in SPS s statistical projections sales and customers. A. The formulae in the regression models and associated statistics used in SPS s projections sales and customers are provided in Attachment JEM-. Specifically, Attachment JEM- shows, by customer class or major rate class, the formulae in the regression models with their summary statistics and output and descriptions for each variable included in the model. Q. What techniques did SPS employ to evaluate the validity its quantitative forecasting models and sales projections? A. There are a number quantitative and qualitative validity tests that are applicable to multiple regression analysis. Several the more common tests SPS uses are as follows: First, the coefficient determination ( R-squared ) test statistic is a measure the quality the model s fit to the historical data. It represents the proportion the variation the historical sales around their mean value that can
Direct Testimony Jannell E. Marks 9 0 be attributed to the functional relationship between the historical sales and the explanatory variables included in the model. If the R-squared statistic is high, the set explanatory variables specified in the model are explaining a high degree the historical sales variability. All regression models used to develop the sales forecast except one demonstrate R-squared statistics larger than 90%, which is satisfactory under this standard. The exception is the regression model for sales to the Municipal and School Service class, which produces an R-squared statistic %. However, given the amount volatility in the historical sales in this class, the R-squared statistic % is also satisfactory. Second, the t-statistic each variable indicates the degree correlation between that variable s data series and the sales data series being modeled. The t-statistic is a measure the statistical significance each variable s individual contribution to the prediction model. Generally, the absolute value each t-statistic should be greater than.90 to be considered statistically significant at the 9% confidence level and greater than. to be considered statistically significant at the 90% confidence level. This criterion was applied in the development the regression models used to develop the sales forecast. All but two variables in the final regression models used to develop the sales forecast
Direct Testimony Jannell E. Marks 9 0 tested satisfactorily under the 9% confidence level standard. The other two variables tested satisfactorily at the 9% and 9% confidence level. Third, each model was inspected for the presence first-order autocorrelation, as measured by the Durbin-Watson ( DW ) test statistic. Autocorrelation refers to the correlation the model s error terms for different time periods. For example, under the presence first-order autocorrelation, an overestimate in one time period is likely to lead to an overestimate in the succeeding time period, and vice versa. Thus, when forecasting with a regression model, absence autocorrelation between the error terms is very important. The DW test statistic ranges between 0 and, and provides a measure to test for autocorrelation. In the absence first-order autocorrelation, the DW test statistic equals.0. The final regression models used to develop the sales forecast tested satisfactorily for the absence first-order autocorrelation, as measured by the DW test statistic. Fourth, graphical inspection each model s error terms (i.e., actual less predicted) was used to verify that the models were not misspecified and that statistical assumptions pertaining to constant variance among the residual terms and their random distribution with respect to the predictor variables were not
Direct Testimony Jannell E. Marks 9 0 violated. Analysis each model s residuals indicated that the residuals were homoscedastic (constant variance) and randomly distributed, indicating that the linear regression modeling technique was an appropriate selection for each customer class sales that were statistically modeled. Fifth, the statistically modeled sales forecasts for each customer class have been reviewed for reasonableness as compared to the respective monthly sales history for that class. Graphical inspection reveals that the patterns the forecast fit well with the respective historical patterns for each customer class. The annual total forecast sales have been compared to their respective historical trends for consistency. Similar qualitative tests for reasonableness and consistency have been performed for the customer level projections. Q. Does SPS s forecasting methodology reflect current economic conditions? A. Yes. SPS s forecast relies upon the analysis relationships between sales and several explanatory variables, such as weather, price, and economic indicators. These relationships and their ultimate explanatory power have been tested and are viable, as described above. The forecast was prepared in February 0 and includes the most up-to-date economic information available at that time. In addition, the forecast is based on oil and gas prices as the beginning 0,
Direct Testimony Jannell E. Marks 9 0 which reflects the reduction in prices seen in mid-0. The Test Year customer and sales forecast is consistent with Global Insight s current economic outlook and current oil and gas prices. Q. Are the models used to prepare forecasts you discuss in your testimony provided in a fully functioning electronic format? A. No. The forecast models are not available in a manner which can be provided as such. SPS will re-run the models for input changes as reasonably required by staff and intervenors. C. Calendar-Month Sales Derivation Q. Please explain the difference between billing-month sales and calendarmonth sales. A. SPS reads electric meters each working day according to a meter-reading schedule based on billing cycles per billing month. Meters read early in the calendar month mostly reflect consumption that occurred during the previous calendar month. Meters read late in the calendar month mostly reflect consumption that occurred during the current calendar month. Consequently, the billing-month sales for the current calendar month reflect consumption that occurred in both the previous calendar month and the current calendar month.
Direct Testimony Jannell E. Marks 9 0 Thus, billing-month sales lag calendar-month sales. In order to determine the sales for a calendar month, SPS estimates unbilled sales, which is the electricity consumed in the current calendar month that is not billed to the customer until the succeeding calendar month. Q. What is the purpose estimating calendar-month sales? A. Calendar-month sales are needed to align the Test Year revenues with the relevant projected Test Year expenses, which have been estimated on a calendar-month basis. Q. Does SPS reflect calendar-month revenue on its books for accounting and financial reporting purposes? A. Yes. Q. How were the estimated monthly calendar-month sales determined? A. For the Residential and Small General Service, Secondary General, Irrigation Service, Municipal, and School classes, SPS calculated the Test Year calendarmonth sales based on the projected billing-month sales. The Test Year calendarmonth sales were calculated in terms the sales load component that is not associated with weather ( base load ), and the sales load component that is influenced by weather ( total weather load ). The weather was measured in terms
Direct Testimony Jannell E. Marks normal heating degree days and cooling degree days, as described earlier. The base-load sales and the total weather load sales components were calculated for each class. The two components were then combined to provide the total 9 0 9 0 calendar-month volumes. The calendar-month base-load component was calculated as follows: Step The billing-month total weather load was calculated. This was accomplished by multiplying the billing-month sales weather normalization regression coefficients (defined in terms billing-month heating degree days, cooling degree days, and number customers), times billing-month normal heating degree days and cooling degree days, times the projected customers. Step The billing-month base-load was calculated by taking the difference between the projected total billing-month sales and the billingmonth total weather load (as calculated in Step ). Step The billing-month base-load sales per billing day was determined by dividing the billing-month base-load sales (from Step ) by the average number billing days per billing month. Step The calendar-month base-load sales were then calculated by multiplying the billing-month base-load sales per billing day (from Step ) times the number days in the calendar month. The calendar-month total weather load component was calculated the same way as the billing-month total weather load was calculated (as described in Step above). However, the calculation was performed by substituting the calendar-month sales weather normalization regression coefficient (defined in
Direct Testimony Jannell E. Marks 9 0 terms calendar-month heating degree days, cooling degree days, and number customers) and the calendar-month normal heating degree days and cooling degree days. The calendar-month total sales were calculated by combining the calendar-month base load and calendar-month total weather load components. For the Area Lighting, Primary General, and Large General Service Transmission classes, SPS calculated the Test Year calendar-month sales simply based on the projected billing-month sales in the same manner as detailed above. However, for these classes, there are no total weather load sales. The Test Year calendar-month total sales for these classes were calculated only in terms their projected billing-month sales. The Street Lighting class is billed on a calendar-month basis in the succeeding month. Therefore, for this class, the calendar-month sales equal the billing-month sales in the succeeding month.
Direct Testimony Jannell E. Marks VIII. RATE SHEET FORECAST Q. What is a rate sheet level forecast? A. A rate sheet level forecast is a forecast for a particular rate sheet, or tariff, within a customer class. Q. How is the rate sheet level forecast derived from the customer class level forecast? A. For all classes, after the class level sales and customer forecasts are completed, the rate sheet level forecasts are developed. Monthly rate sheet sales allocation factors are developed based on historical rate sheet level sales data. The 9 0 allocation factors are then applied to the class level sales forecasts to derive the rate sheet level sales forecasts. Rate sheet customer allocation factors are developed based on year-end customer counts from the most recent year. The rate sheet customer allocation factors are then applied to the class level customer count forecasts to derive the rate sheet level customer count forecasts. I provided this rate sheet level detail to SPS witness Ian C. Fetters for use in the Class Cost Service Study and Richard M. Luth for Rate Design. 9
Direct Testimony Jannell E. Marks IX. CONCLUSION 9 0 Q. Were Attachments JEM- through JEM- prepared by you or under your direct supervision and control? A. Yes. Attachment JEM- provides the workpapers supporting my testimony. Q. Were the Rate Filing Package schedules that you sponsor or co-sponsor prepared by you or under your direct supervision and control? A. Yes. Q. Do you incorporate the Rate Filing Package schedules that are sponsored or co-sponsored by you into your testimony? A. Yes. Q. Does this conclude your pre-filed direct testimony? A. Yes. 0
Attachment JEM- Page Test Year Forecast Retail Customer Counts and Sales (MWh) by Major Class Customer Counts Commercial Street Public Residential & Industrial Lighting Authority Total Retail Jan- 9,, 9,, Feb- 9,0,9 9,,0 Mar- 9,0, 9,,9 Apr- 9,9,9 9,0 9,00 May- 9,,9 9, 9, Jun- 9,,9 9, 9,0 Jul- 9,0,00 9, 9, Aug- 9,0,0 9, 9, Sep- 9,,09 9,9 9,90 Oct- 9,, 9,0 9,0 Nov- 9,9, 9, 9, Dec- 9,9, 9, 9,99 Jan- 9,0,9 9, 0,09 Feb- 9,09,0 9, 0, Mar- 9,, 9, 0, Apr- 9,, 9,0 0,0 May- 9,,0 9, 0, Jun- 9,0, 9, 0, Jul- 9,, 9, 0, Aug- 9,0,9 9, 0, Sep- 9,,0 9, 0,99 Oct- 9,99, 9,9, Nov- 9,,9 9,, Dec- 9,9, 9,, 0 Monthly Average 9,,9 9, 9,9 0 Monthly Average 9,0,9 9, 0,9
Attachment JEM- Page Test Year Forecast Retail Customer Counts and Sales (MWh) by Major Class Sales (MWh) Residential Commercial & Industrial Street Lighting Public Authority Total Retail Jan-,0 0,090,0 0,9,0 Feb- 9,09,,0, 9, Mar- 9,9,,0 0,9, Apr-,0,,0 0,,0 May-,99,9,,9, Jun- 9,09,09,09,9,0 Jul-,0 90,0,,0, Aug-, 9,,,0,0 Sep- 9,,,0,09,9 Oct-,9,,0,,9 Nov-,,,09 9,9,0 Dec- 0,9 9,, 9,90 0,0 Jan- 0,,, 0, 0,0 Feb- 0,,,0 9,, Mar-, 9,9,9 0,009 9, Apr- 9,,,0 0,999 9,0 May-,0,9,,0 0, Jun- 99,,,0,,0 Jul- 0, 0,,, 0, Aug-,9,9,, 9, Sep- 9,9,,0,0,9 Oct-,,,,, Nov- 0,9,9,0 0,00, Dec-,,9, 0,00, 0 Total,,,,,,,, 0 Total,,99,09,0,,,9,9 0 Monthly Average 9,0,9,,,9 0 Monthly Average 9,,,,0,0