Short-term Power Load Forecasting Using Support Vector Machine based on Differential Evolution

Abstract

Support Vector Machine (SVM) as a machine learning method is based on solid theory foundation of statistical learning theory, and it has a excellent fitting ability which focuses on small samples, in this paper, as an efficient stochastic global optimization technique Differential Evolution (DE) is used to optimize parameters of SVM, and the SVM fitting model based on DE is illustrated in detail, which is applied to short-term load forecasting in electric power systems. Finally, the comparative simulation results show the parameters of SVM can be well optimized using the DE, the established fitting model has a better fitting ability and forecasting effect for short-term load forecasting in electric power systems than its counterparts.