LOAD FORECASTING FOR ELECTRIC UTILITIES (ECONOMETRICS, PLANNING, TIME SERIES, END-USE, MODELS)

The prime directive of any regulated electric utility is to provide adequate and reliable electricity supplies to the consuming public at a reasonable cost. The ability of a utility to minimize the cost of electricity depends directly on the ability of the load forecast to predict the level of energy sales and peak demand over time.;The results showed that "Does the forecast make sense?", data availability, and historical performance of the model were the most important selection/evaluation criteria for all three client groups, namely, utility analysts, utility senior managers, and regulators. Those differences which did exist were primarily between utility respondents and commissions with the commissions rating explainability and acceptability to the commission low compared to the utilities. The major forecast use was for generation planning followed by state and federal filings, rate design, rate cases, and market program evaluation.;Analysis of historical accuracy of utility forecasting was performed by forecast horizon, forecast vintage, time devoted to forecasting, sector, technique, and type of forecast (energy or peak). The results showed that end-use models have performed particularly well in the residential sector while customer surveys have worked well for short-term forecasts in the industrial sector. Econometric techniques have a somewhat disappointing record and in most cases actually do worse than trend/judgment techniques.;The performance of five time series techniques was compared using historical utility sales data. The techniques tested were Holt's Exponential Smoothing, Univariate Adaptive Estimation Procedure, Linear Regression, a combination technique, and Multiple Regression with state real per capita income, state population, and national real electricity price as the independent variables.;The research contained in "Load Forecasting for Electric Utilities" is designed to answer three principal questions: (1) What has been the historical accuracy of electric utility forecasts? (2) How important is historical accuracy in selecting or evaluating an electric utility forecast? and (3) How well do advanced time series techniques perform versus the utility models?;The combination technique proved to be the best overall technique across all measurement methods, forecast vintages, and horizons. The Univariate Adaptive Estimation Procedure also performed well in all situations. Actual utility forecasts did extremely well versus the time series methods for the two-year horizon, but their performance deteriorated with longer horizons.