Research On Short-Term Load Forecasting Model Based On Particle Swarm Optimization-Localized Support Vector Regression

Short-term load forecasting (STLF) can affect social economy benefits. It is the basis for safe and stable operation of power systems. With the continuous development of electricity market, high precision of STLF is required.In this paper, STLF's background, research status and significance were presented firstly. In order to understand the characteristics and disciplines of the load changes, this paper analyzed the short-term load characteristics and its influence factors. Because of the strong nonlinear relationship between short-term load and its influence factors, STLF model based on support vector regression (SVR) of was proposed.Although SVR has been successfully applied in many domains, it lacks the trend between training datas, and this will lead to decrease forecasting accuracy. Consequently, Localized Support Vector Regression (LSVR), which can capture this trend and inprove the performance of the model, was proposed in this paper. In addition, the selection of key parameters depends on the experience of the researches in the process of building LSVR model, so the forecasting model was not very accurate. The model based on PSO-LSVR was proposed to improve the accuracy of forecasting, which used particle swarm optimization (PSO) to find the better parameters.Finally, this paper used the datas from HangZhou's electric power department and built models based on SVR,LSVR and PSO-LSVR respectively. The performance comparison between the simulation of LSVR-based and the SVR-based forecasting model proved that LSVR is better than SVR. And the PSO-LSVR-based model is the best one, because its average relative error was smaller than the others.Simulation results showed that STLF based on PSO-LSVR has good predictive results. And the proposed method has research-worthy and social meaning.