Research And Implementation Of Short-Term Load Forecasting Method Of Power System

Load forecasting is the prerequisite and basis of the decision-making and economic operation for power system. The accurate load prediction is great significance on the power system safe and national economic development.As the representative of the linear model, the time series has been used widely in Classic load forecasting system, however, due to short-term load data vulnerable to weather, holidays and other factors interfere, it's often show a non-linear feature, so the results can not meet the actual needs. Based on the analysis of various classical loads forecasting method advantages and disadvantages, this paper advanced a non-linear method based on Core Vector Regression (CVR) and determined the best function through comparing different kernel functions and parameters in the results of prediction. In this paper also improved Particle Swarm Optimization (PSO) algorithm to determine the core parameters. Experiments show that the PSO-CVR model has comparable performance with SVR but has much faster training speed and produces fewer support vectors on very large data sets.