Procedure for creating custom MLR-based STLF models by using GA optimization

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