Short-term Load Forecasting And Amending Model Research Based On SVM Optimized By Intelligent Algorithm

Short-term load forecasting research has important academic value and practical application value. This article presents a relatively new model, which named least squares support vector machine (LSSVM) model, to do short-load forecasting for Baoding power grid. We make research on training algorithm of LSSVM, parameter selection, the construction of kernel function, load characteristics, as well as error amending model, based on the historical load data of Baoding and the conclusion of the existing theoretical results. RBF kernel function is chosen the kernel function of LSSVM, and dynamic inertia weight particle swarm algorithm (WPSO) is adopted to optimize the parameters of LSSVM, which are kernel parameter and regularization parameter. At last GARCH error amending model is applied to analyze and forecast the error, so to amending the forecasting results. The models brought forward are used to short-load forecasting of Baoding city, and the excellent forecasting effect of which are shown in the data results.