Short Term Load Forecasting with Least Square Support Vector Regression and PSO

Load forecasting has been an important topic in power system research. Short term forecasting is important for power dispatch, especially in the modern electricity market. In this paper, an approach based on least square support vector regression (LSSVR) is proposed to short term load forecasting. An effective forecasting model can only be built under optimal parameters. The algorithm of particle swarm optimization is applied to search optimal parameters of the above forecasting model. The experimental results based on above model for a sample load series are shown that the model proposed in this paper outperforms the BP neural network approaches and the simple LSSVR methods on the mean absolute percent error criterion.