Improve short-term electricity consumption forecasting using a GA-based weighted fractional grey model

Abstract

This study proposed a weighted fractional grey model (WFGM) based on a genetic algorithm for forecasting annual electricity consumption. WFGM has two parameters that can be used to adjust the order of the summation based on different data sequences and reflect the new information priority. The key issue with the WFGM model is determining two optimum fractional-order values to improve the accuracy of electricity consumption forecasts. The Genetic Algorithm (GA) is used to select the best values for the weighted fractional-order accumulation, which is one of the most important aspects determining the grey model's prediction accuracy. The additional linear parameters of grey models are estimated using the least squares estimation method. Finally, two real data sets of electricity consumption from Malaysia and Thailand are presented to validate the proposed model. Numerical results show that the new proposed prediction model is very efficient and has the best prediction accuracy compared to the models of GM(1,1) and FGM(1,1).