A new approximation method for generating day-ahead load scenarios

Unit commitment decisions made in the day-ahead market and resource adequacy assessment processes are based on forecasts of load, which depends strongly on weather. Two major sources of uncertainty in the load forecast are the errors in the day-ahead weather forecast and the variability in temporal patterns of electricity demand that is not explained by weather. We develop a stochastic model for hourly load on a given day, within a segment of similar days, based on a weather forecast available on the previous day. Identification of similar days in the past is based on weather forecasts and temporal load patterns. Trends and error distributions for the load forecasts are approximated by optimizing within a new class of functions specified by a finite number of parameters. Preliminary numerical results are presented based on data corresponding to a U.S. independent system operator