Probability Approach to Electric Utility Load Forecasting

A sound theory of forecasting may be defined in two steps. The first step is to separate the whole into components that are more easily analyzed and reasoned than the whole. Previously existing techniques supplemented by new applications of regression analysis[1] are available to solve this step. The second step is to forecast the components and to reunite them to obtain a forecast of the whole. A digital computer program that has been developed to solve this second step will be discussed. Most significant in this new load forecasting program is the addition of another dimension-the probability dimension. The user of this program expresses to the program his uncertainty of the component forecasts, and it generates the resulting uncertainty of the final whole forecast. The required probability theory and methods of expressing uncertainty are stressed.