Procedure for creating custom multiple linear regression based short term load forecasting models by using genetic algorithm optimization

This paper presents a novel procedure for short-term load forecasting in distribution management systems. The is forecasted feeders that can be of primarily residential, commercial, industrial or combined type. Each feeder has various amounts distributed energy resources installed, which accounts multiple different patterns. Hence, the systems cannot use single model all forecasts. proposed addresses specificity each particular type, by creating customized models. It uses genetic algorithm to select best inputs linear regression chooses variables from dataset constructed using and temperature measurements. extended adding non-linear transformations mutual interaction effects measurements, as well calendar variables. extension enables modelling influences extracts non-linearity domain input models? performance assessed mean absolute percentage error. applied set measurements collected an US electric power utility, operates city Burbank, Cal., USA. obtained compared with previously na?ve benchmark, special comparison model, developed correlation analysis. method extendable suit types electricity consumers.