Probabilistic methods in forecasting hourly loads

EPRI has investigated the use of independent hourly statistical models for short-term load and probabilistic forecasting. Model outcomes compared favorably with observed and predicted loads. The models performed satisfactorily in generating short-run forecasts of system loads in tightly controlled experiments against a wide range of altemative models and methodologies. In addition to load forecasting, the models also forecast the probability that the predicted load would exceed a prespecified level at each hour of the day. Changes in operational practices and personnel turnover are creating a situation in which new forecasting methods are needed. This report demonstrated that statistical models can be developed to reliably predict hourly loads and form probabilities of extreme events.