A Review of Load Forecasting Methodologies

In response to increasing criticisms of their load forecasts and forecasting methods, Iowa\u27s electric utilities sponsored an independent review of past and present load forecasting methodologies. The review was conducted by an Iowa research team and followed two approaches. One was to evaluate various energy and peak demand models used by United States\u27 electrical utilities, with emphasis on models developed during the period 1973 through 1979. The second approach involved construction of econometric energy demand models for an Iowa utility. Historical energy and peak demand models were classified by methodology (statistical, econometric-end use analysis) and demand class (residential, commercial, and industrial). Statistical and econometric models were examined for forecast and backcast accuracy and parameter stability over time. Econometric-end use simulation models were observed for parameter sensitivity and, when possible, accuracy. The energy demand models were constructed for the residential and commercial classes with the purpose of incorporating variables considered relevant by economic theory and available literature. These variables, and their various combinations, were tested for statistical significance and logical applicability to Iowa. The results of this study will provide a foundation on which to begin construction of a comprehensive set of load forecasting models for use by Iowa utilities and legislators