TRANSMISSION BUSINESS LOAD FORECAST AND METHODOLOGY

Transcription

1 Filed: September, 00 EB Tab Schedule Page of 0 TRANSMISSION BUSINESS LOAD FORECAST AND METHODOLOGY.0 INTRODUCTION 0 This exhibit discusses Hydro One Networks transmission system load forecast and the related methodology. The key load forecast supporting the transmission rate case is the hourly demand load forecast by customer delivery point. This forecast is used to prepare the charge determinant forecast for the following rate categories: network pool, line connection pool, and transformation connection pool. Hydro One Networks uses a number of methods, such as econometric models, end-use models, customer forecast surveys and hourly load shape analysis to produce the forecasts required for its transmission business. All forecasts presented in this exhibit are weather-normal, i.e. abnormal weather effects are removed from the base year for load forecasting purposes so that the forecast assumes typical weather conditions based on the average of the last 0 years. The weather correction methodology used by Hydro One Networks is a proven technique which was reviewed and approved by the OEB in its decision for RP /EB All forecasts produced are internally consistent. This means that the forecasts for all customer delivery points add up to the total for the entire customer base served by Hydro One Networks transmission system. Also, the forecasts presented in this exhibit are consistent with the economic assumptions which are used in the business planning process and that are described in, Tab Schedule. Hydro One Networks load forecasting team has significant expertise in preparing provincial electricity demand forecasts as well as hourly load shape analysis. The

2 Updated: February, 00 EB Tab Schedule Page of 0 performance of Hydro One Networks transmission system load forecast, since Hydro One Network s separation from the former Ontario Hydro, has been fairly consistent as shown in Table below. 0 0 Table Comparison of Average Monthly Transmission Peak Demand Forecast with Actual (Variance of forecast as % of actual on weather corrected basis) Forecast made Forecast for Forecast Forecast In Year current year for nd Year for rd Year -0.% -.% -.0% % 0.% 0.% 00-0.% -0.% 0.% 00 0.% 0.% -0.% 00 0.% 0.0% 0.% % 0.% n/a 00 0.% n/a n/a Mean -0.0% -0.0% -0.0% One standard deviation (+/-).%.%.0% Between and 00, the average variance of the transmission peak demand forecast compared to the weather corrected actual peak is within one standard deviation of the forecast. One standard deviation means there is a one in three chance that the actual will be outside the plus or minus range (alternatively, there is a two in three chance that the actual will fall within the plus or minus range). The use of the one standard deviation as a measure of forecasting accuracy is an accepted standard in the utility industry. 0 Section below provides more detailed discussion in respect of the various economic considerations that Hydro One Networks staff take into consideration when applying the methodology for deriving the load forecasts.

3 Filed: September, 00 EB Tab Schedule Page of 0 0 Hydro One Networks forecasting methodology comprises a combination of elements that include consensus input, mechanical adjustments to models commonly used in the forecasting business to include changes in economic forecasts, energy prices, population and household trends, industrial development and production, residential and commercial building activities, and efficiency improvement standards. Economic inputs were based on analyses prepared by major economic establishments in the country such as Global Insight, Conference Board of Canada, Centre for Spatial Economics, University of Toronto, Canada Mortgage and Housing Corporation, Clayton Research; efficiency standard assumptions used in the end-use models were based on discussion with Ontario Ministry of Energy staff; specific customer development was based on forecast survey results from major customers. Inputs from these entities form the economic database (referred to henceforth as the economic forecast) that is used to establish Hydro One Networks load forecast. 0 Section below provides a detailed description of the methodology used by Hydro One to develop its load forecasts. Section presents the charge determinant forecast for 00 and 00 using the methodology described in this exhibit. Detailed equations and definitions are provided in Appendices to..0 DISCUSSION OF KEY ASSUMPTIONS THAT INFLUENCE HYDRO ONE NETWORKS LOAD FORECASTS In this section we discuss some of the key assumptions that must be taken into account in the process of developing load forecasts and in the application of forecasting methodologies. The elements of the forecasting process used by Hydro One Networks are for the most part based on the knowledge of how the major economic drivers that affect the usage of electricity demand are likely to evolve over the forecast period, which in this case is for the years 00 and 00. Consequently for the purpose of this

4 Filed: September, 00 EB Tab Schedule Page of 0 application the focus is on the short term and the load forecast will reflect those impacts that are likely to have a major effect in this respect. The major assumptions used in the analysis are summarized in Figure below. Figure Key Assumptions Used in the Forecast Key Drivers Transmission System Forecast - Canadian output forecast - Provinical output forecast Econometric Approach End-use Approach - Population forecast - Short-term model - Forecast by sector and by - Housing forecast - Long-term model end-use - Commercial floorspace forecast - Forecast by customer class - Industrial production forecast and by sector Key Drivers Forecast Net of C&DM and By-Pass Impacts - C&DM forecast - Reduction of load due to C&DM savings - By-pass forecast - Reduction of load due to embedded generation, transformation connection and line connection by-pass Key Drivers Customer Forecast - Provinical output forecast - Population forecast Forecast by customer/ region - Housing forecast - Econometric analysis - Commercial floorspace forecast - Analysis by customer - Industrial production forecast - Customer load forecast survey Key Drivers Customer Delivery Point Forecast - Hourly load data by delivery point - Prepare load shape analysis for each delivery point - Hourly weather data - Ensure delivery point forecasts sum to customer forecast - Ensure customer forecasts sum to regional forecast - Ensure regional forecasts sum to transmission system forecast 0 Key information used in the analysis includes Canadian and provincial GDP, provincial demographic, industrial production and commercial floorspace forecasts and regional analysis included in the economic forecast. Also taken into consideration are the provincial C&DM plans and by-pass risks, which have a direct impact on Networks system energy demands. The load forecast in support of this application was prepared and released in May 00 using economic and forecast information that was available in April 00. The timing of

5 Filed: September, 00 EB Tab Schedule Page of 0 the load forecast is driven by the needs of business planning and rate submission processes.. Canadian GDP Forecast 0 After the September, 00 crisis in the U.S., the Canadian economy recovered strongly (. percent) in 00. Economic growth slowed to.0 percent in 00, due primarily to the impact of SARS, Mad Cow Disease, August Blackout, the strong Canadian dollar and uncertainty of U.S. growth resulting from the fight of terrorism and the Iraqi war. Oil and other commodity prices and the Canadian dollar continued to be strong since 00; real gross domestic product (GDP) in Canada grew. percent in 00 and The general consensus among economists in 00 was that high energy prices are fuelling an economic boom for oil producing provinces, but at the same time they are creating considerable economic challenges for Ontario and Quebec, which have a large manufacturing sector. The economic outlook in Canada in 00 was also dominated by the strong Canadian dollar. This is especially true in Ontario where 0 percent of Ontario s exports are destined for the United States. Since early 00, the Canadian dollar went up by more than 0 percent relative to the U.S. dollar, and therefore the issue, from a load forecasting perspective, is what sort of impact it would have on resource and manufacturing industries, given that the U.S. is the most important trading partner of Canada. The question is no longer whether the Canadian economy will be affected by the rising Canadian dollar, but rather how long and how strong the adverse impact will be on the economy. Since 00, economists have observed signs of slowing exports, increased imports, slower industrial growth and increased job cuts in the manufacturing and

6 Updated: February, 00 EB Tab Schedule Page of 0 resource industries. While the Canadian economy was not expected to slow down in the next few years, economic growth prospects are likely to be restrained for non-oil producing provinces. Within the manufacturing sector, exporters to the U.S. are expected to be the hardest hit. However, supported by relatively low interest rates, strong energy and commodity production, healthy domestic demand, relatively strong construction activities and robust U.S. economic growth, the Canadian economy is expected to grow by about.0% in 00. Based on the consensus forecast, the Canadian GDP is expected to grow by. percent and. percent in 00 and 00 respectively. 0. Provincial GDP Forecast 0 The provincial GDP forecast is a key driver for the load forecast. During the 0s, the Ontario economy grew faster than the national economy by an average of 0. percent per year. During the economic slowdown in 00 and 00, the Ontario GDP lagged behind the Canadian GDP by about 0. percent per year because of the SARS outbreak and the Blackout in 00 and rising Canadian dollar for which the provincial manufacturing sector was the hardest hit. Industries that were negatively affected in recent years include the pulp and paper, chemical and auto-related industries. The provincial economy grew.% in 00 and is expected to grow.% in 00. Based on the consensus forecast, the Ontario GDP is expected to grow at. percent in 00 and.0% in 00. Because of the strong Canadian dollar, Ontario is expected to lag behind the national growth rates, due to its manufacturing base.. Provincial Population Forecast The Ontario population grew. percent in 00,. percent in 00,. percent in 00 and.% in 00. Population growth for the province is forecast to continue to outperform the nation in the forecast period. The economic forecast indicates that the

7 Filed: September, 00 EB Tab Schedule Page of 0 Ontario population is expected to grow at. percent in 00 and.0 percent in 00. Steady population growth contributes positively to the load forecast.. Provincial Housing Forecast 0 Helped by the 0-year low interest rates, demand for housing was strong in the last few years. Housing starts statistics showed growth of,000 houses in 00,,000 houses in 00,,000 houses in 00 and,000 houses in 00. Demand for housing is expected to slow in the next years as interest rates begin to rise. The consensus forecast indicates that housing starts are expected to be about,000 units in 00, 0,000 units in 00 and,000 units in 00.. Commercial Floor Space Forecast 0 With the help of low interest rates, commercial construction activities remained relatively strong in Ontario in the last five years. After adding an average of about million square feet of commercial floor space annually over the -00 period, commercial construction slowed to million square feet of floor space in 00, followed by million square feet in 00. The economic forecast shows growth of about million square feet of floor space per year for 00, 00 and 00. The forecast for commercial floor space additions is an important contributor to the commercial sector load forecast.. Industrial Production Forecast After a record growth of. percent in 00, the manufacturing sector in Ontario is forecast to slow down considerably in the next few years due primarily to the impact of higher Canadian exchange rate. Since early 00, the Canadian dollar has appreciated more than 0% percent relative to the U.S. dollar. The economic forecast expects slow

8 Filed: September, 00 EB Tab Schedule Page of 0 industrial production growth in 00 (%) and 00 (.%) and slight recovery in 00 (.%) due to the high Canadian exchange rate. The industrial production forecast is an important contributor to the industrial sector load forecast but it is also prone to economic cycles.. Conservation and Demand Management Forecast 0 Hydro One Networks supports the Ontario Government s conservation and demand management (C&DM) target to achieve a percent peak load reduction (0 MW) by 00. Beyond 00, Hydro One Networks assumes that C&DM programs will continue in the province, achieving another 00 MW of peak load savings in 00. Table summarizes the C&DM impacts assumed in Hydro One Networks transmission system load forecast for 00, 00 and Table Load Impact of C&DM on Ontario Demand (MW) Load Impact on Load Impact on Year Maximum Peak Demand -month Average Peak Demand 00 00,0,0 00,0, C&DM programs will be planned and delivered through a number of agencies including the OPA, IESO and LDCs. Programs that have been implemented in the past two years or in the process of being initiated include the following initiatives: 0 improved building codes for new housing and more stringent efficiency standards for appliances;

9 Updated: February, 00 EB Tab Schedule Page of 0 conservation programs to encourage more efficient use of lighting and appliances; demand response programs to reduce air conditioning and water heating load in the summer months; use of smart metering and TOU rates to encourage consumers to shift consumption patterns to off-peak period; and programs to increase supply or reduce demand such as using back-up generation or requesting large industrial customers to reduce consumption on a temporary basis.. By-Pass Forecast 0 Hydro One Networks collects its transmission revenue through three types of OEB approved transmission charges (networks, line connection and transformation connection) for customers using its transmission system. When Hydro One Networks customers get power from their own embedded generation (instead of buying power through the IESO controlled transmission system) or building their own transformation station or line connections to their distribution system, Hydro One Networks loses transmission revenue. 0 The following summarizes the by-pass forecast assumptions used in the Transmission system load forecast for 00 and 00: 0 MW in 00, 0 MW in 00 and another MW in 00 of embedded generation were assumed in the load forecast based on available information of embedded generation projects during this period. Since it would take time for customers to plan, implement and get regulatory approval for building their own transformer station or line connection and Hydro One was not aware of any proposed new connections, no reduction in load was assumed in the test years 00 and 00.

10 Filed: September, 00 EB Tab Schedule Page 0 of 0.0 LOAD FORECASTING METHODOLOGY Hydro One Networks transmission system load forecast is developed using both econometric and end-use approaches. The forecast base-year is corrected for abnormal weather conditions and the forecast growth rates are applied to the normalized base-year value. Thus the forecast is weather-normal in the sense that it predicts the future load under normal weather conditions.. Weather Correction Analysis 0 This section discusses the weather correction methodology used by Hydro One Networks. Weather correction analysis removes the abnormal or extreme weather effects from the load data to yield average conditions that reflect the more normal or expected weather that is used in the forecast. It is essential that abnormal and extreme weather related impacts are removed before establishing the base-case load data, on which basis the load forecast will be developed. The volatility of abnormal or extreme weather conditions would likely adversely impact on the ability to provide a consistent and meaningful forecast for load growth. Hourly load data and hourly weather data of various weather stations across the province are used in the analysis. 0 Hydro One Networks weather correction methodology was developed jointly by forecasting and meteorology staff of the former Ontario Hydro. This weather correction method was used to forecast the total system load since and for forecasting local electric utility load since. The weather correction methodology used by Hydro One Networks is a proven technique that has performed well in the past years. Weather correction is a statistical process designed to remove the impact of abnormal or extreme weather conditions from historical load data. Normal weather data is defined to be data that is based on the average weather conditions experienced over the last 0

11 Filed: September, 00 EB Tab Schedule Page of 0 years. A weather-normal load forecast is a forecast of load assuming normal weather conditions with a weather-corrected base year. 0 Hydro One Networks weather correction methodology uses four years of daily load and weather data to establish a sound statistical relationship between weather and load at the applicable transformer station or delivery point used to supply customer demand. Weather variables used in the analysis include temperature, wind speed, cloud cover and humidity. The estimated weather effects are then aggregated up to the required time interval. Past experience shows that weather correction should best be done on a daily basis, rather than weekly, monthly or annual basis. Daily weather-correction is preferred because the timing of extreme temperatures combined with wind speed and humidity can have a substantial impact on load that would otherwise not be captured by averages over longer period of time. In particular, when abnormal weather conditions continue for several days, the cumulative impact is much greater than any single day s impact. 0 The loads that are most impacted by changes in weather conditions are electric space heating and cooling in residential and commercial buildings. Across Ontario, the penetration rate of such loads varies widely, which means the weather sensitivity of load supplied from one transformer station or delivery point may differ quite significantly from that of load supplied from another transformer station or delivery point, even in the same climate zone. The climate in Ontario varies considerably from the Niagara Peninsula to Thunder Bay, so it is important to use data from the appropriate weather stations that are in close proximity to the transformer station or the customer delivery point when correcting for weather effects.

12 Filed: September, 00 EB Tab Schedule Page of 0. Hydro One Transmission Forecasting Methodology 0 Hydro One Networks uses econometric (top-down) and end-use (bottom-up) models to forecast the transmission system load. For the top-down approach, both monthly and annual econometric models are used. For the bottom-up approach, end-use models are used to analyse the transmission system load by sector (i.e. residential. commercial and industrial customers). Key information used in the analysis includes economic, demographic, industrial production and commercial floorspace forecast provided in the economic forecast. The purpose of using both the econometric and end-use forecast models is to arrive at a balanced forecast that represents a consistent set when looked at from macro (econometric) and micro (end-use) perspectives. 0 Monthly Econometric Model The monthly econometric model uses a multivariate time series approach to develop the monthly forecast for total transmission system load. The model links monthly energy consumption to Ontario GDP and residential building permits, taking into account the August 00 blackout and C&DM programs. The transmission system load used in the model is weather-normal, i.e. corrected for abnormal weather effects. Appendix provides the detailed regression equations and definitions. Annual Econometric Model The annual econometric models cover five sectors of the economy- residential, commercial, industrial, agriculture, and transportation. Appendix provides the detailed regression equations and definitions. 0 The residential sector is modelled as a two-equation system for saturation and usage of electric equipment. Explanatory variables used include energy prices, personal disposable income per household and weather conditions as measured by heating degree days.

13 Filed: September, 00 EB Tab Schedule Page of 0 The commercial sector links energy usage to electricity price, commercial GDP and weather conditions as measured by heating and cooling degree days. The industrial model consists of an equation for total energy and a -equation model to determine share of electricity usage. Total energy is modelled as a function of energy price and industrial GDP. Energy shares are linked to relative energy prices. Dummy variables are used to capture unusual changes in energy growth in the 0 s and early 0 s and to measure the impact of technical change on energy shares. 0 The agricultural sector is modelled in relation to electricity price and income, while accounting for cyclical and trend changes. The transportation sector, which consists mainly of pipeline and road transport, is modelled by an equation relating electricity usage to income, electricity price, and a dummy variable to capture a change in load pattern since. 0 End-Use Model The end-use models cover the residential, commercial, industrial, agricultural and transportation sectors. Appendix provides details of the methodology used in the end-use analysis. In the residential sector, the end-uses analysed include space heating, water heating, air conditioning, and base load. The forecast of each end-use is based on the number of households having that end-use and unit energy consumption of the equipment.

14 Filed: September, 00 EB Tab Schedule Page of 0 The commercial model analyses energy use by building type. Key drivers used in the analysis are the commercial sector floor space and the intensity of end-use demand per unit of floor space. The industrial forecast is based on analysis for each major industrial segment, energy intensity and expected economic growth. The agricultural and transportation sector models are based on base year electricity consumption and the expected growth rates for each sector and segment. 0. Methodology for Customer Forecast This section discusses the load forecasting methodology used for deriving the customer forecast. Both econometric and customer analysis based on survey results from the customers, when available, are used in the forecast. This is supplemented by the economic data provided in the economic forecast. 0 In the spring of 00, Hydro One Networks conducted a customer load forecast survey with customers having more than MW of load. The survey also covered the station service load requirements of generating stations when they are not producing electricity. In addition to questions relating to the total load of the customer, information at each of the delivery point was also collected. The customer survey results are used in preparing the customer forecast. In addition to the information contained in the customer survey, a number of forecasting techniques are used to prepare the load forecast by customer. For large utility customers, each customer is modeled individually using the econometric approach. The drivers used in these models include provincial economic variables such as Ontario GDP, population,

15 Filed: September, 00 EB Tab Schedule Page of 0 number of household, energy prices, as well as local demographic and economic variables such as population and related industrial and commercial loads. The load impact of weather conditions is also taken into account. The best subset of the drivers is selected on the basis of regression criteria. For industrial customers, several information sources are used to prepare the forecast. They include: 0 historical load profile of the customer, knowledge of the customer through industry monitoring, forecast provided by customer through the survey, company information through Hydro One Networks account executives, industry and company forecasts from industry associations and government agencies, and production and industry forecasts provided in the economic forecast.. Methodology for Customer Delivery Point Forecast 0 This section discusses the forecasting methodology for the customer delivery point forecast. Electricity Power Research Institute (EPRI) s Hourly Electric Load Model (HELM) is used to normalize the hourly load for each of the transmission customer delivery points, taking out abnormal weather effects and load patterns. The customer forecast is used to drive the customer delivery point forecast. Key information used in the analysis includes hourly load and weather data. The forecasts for all customer delivery points add up to the regional and the total transmission system forecast. The most updated customer totalization table is used to retrieve hourly peak electricity demand data for each of the customer delivery points connected to the transmission system. The totalization table reflects the latest records from Hydro One Networks and

16 Filed: September, 00 EB Tab Schedule Page of 0 IESO. For each customer delivery point, at least one full year of hourly data is retrieved and checked for data quality. Hourly weather data is also retrieved to prepare weather sensitivity analysis. Weather data analyzed include temperature, wind speed, cloud cover and humidity. Data for five weather stations across Ontario are used in the analysis. They include Toronto, Windsor, Ottawa, North Bay and Thunder Bay. Each delivery point is linked to the closest weather station. 0 In preparing the database for the load shape analysis, missing values are estimated by load on a similar day and hour during the same month. For weather-sensitive load, weather conditions are also taken into account in estimating the missing values. To perform the latter task, an hourly regression model (relating load to weather conditions) for each delivery point with missing values is developed. 0 EPRI s HELM is used to prepare the hourly weather response analysis by each delivery point. The model takes into account differences in load depending upon time of use (that is weekdays, weekends and holidays) and weather conditions. Load of industrial customers is assumed to be insensitive to weather and as such are forecast in relation to load on a similar day and hour during the historical period. The customer forecast is used to drive the customer delivery point forecast. The resulting customer delivery point forecast is therefore consistent with the customer load forecast and the total transmission forecast as discussed above..0 LOAD FORECAST FOR 00 AND 00 The charge determinant forecast is derived from the Ontario peak demand forecast, which in turn was based on econometric, end-use, and customer forecasts discussed above. Table on page of this exhibit presents the forecast before and after deducting the load impacts attributed to embedded generation and CDM. The charge determinant

17 Filed: September, 00 EB Tab Schedule Page of 0 forecast is based on the current rates as approved by the OEB in its decision for RP Before deducting for the load impacts of embedded generation and CDM, Hydro One Networks is forecast to deliver an average of, MW in 00 (-month average peak), rising to,0 MW in 00 and,0 MW in 00. The forecast for the three charge determinants as presented in Table is summarized below: Excerpt from Table Year Ontario Demand Network Connection Charge Determinant Line Connection Transformation Connection (MW) (MW) (MW) (MW) 00,, 0,, 00,0,0 0,, 00,0,,0,0 0 After deducting for the load impacts of embedded generation and CDM, Hydro One Networks is forecast to deliver an average of, MW in 00 (-month average peak), lowering to, MW in 00 and rising to, MW in 00. The forecast for the three charge determinants as presented in Table is summarized below:

18 Filed: September, 00 EB Tab Schedule Page of 0 Year Ontario Demand Excerpt from Table Network Connection Charge Determinant Line Connection Transformation Connection (MW) (MW) (MW) (MW) 00,, 0,,0 00, 0,,,0 00, 0,,0, The forecast is weather-normal and the actual load could be below or above the forecast depending on the weather conditions and/or a different economic growth pattern. Table on Page 0 of this Exhibit presents the upper and lower bands of one standard deviation for the charge determinant forecast. Based on historical data, there is a two in three chance that the actual load in 00 and 00 will fall within the upper and lower bands. The bands are derived using Monte Carlo simulation technique relating variations in load to variations in Ontario GDP and weather.

19 Filed: September, 00 EB Tab Schedule Page of Table Load Forecast Before and After Embedded Generation and CDM (-Month Average Peak in MW) Charge Determinants Ontario Network Line Transformation Year Demand Connection Connection Connection Load Forecast before Deducting Impacts of Embedded Generation and CDM 00,, 0,, 00,0,0 0,, 00,0,,0,0 Load Impact of Embedded Generation Load Impact of CDM ,0,0,00 00,,0,0 Load Forecast after Deducting Embedded Generation and CDM 00,, 0,,0 00, 0,,,0 00, 0,,0, Note. All figures are weather-normal.

20 Updated: February, 00 EB Tab Schedule Page 0 of Table One Standard Deviation Uncertainty Bands for Hydro One Networks Charge Determinants (Using Current Rates) (MW) Year Lower Band Forecast Upper Band Network Connection 00 0,,, 00 0, 0,, 00 0, 0,,0 Line Connection 00, 0, 0, 00,, 0, 00,,0 0, Transformation Connection 00,0,0, 00,,0,0 00,,,