Study of core vector regression and particle swarm optimization for rapid electric load forecasting

Load forecasting is very essential to the operations of electric companies. This paper presents a rapid electric load forecasting algorithm based on Particle Swarm Optimization (PSO) and Core Vector Regression (CVR), called PSO-CVR algorithm. PSO is applied to determine the parameters of CVR, then CVR manages the issues of forecasting and training. In order to compare the results among different size of data sets, 4 training sets of different size are created based on a standard data set for global electric load forecasting competition. Experiment results indicate that the PSO-CVR algorithm is comparable with Support Vector Regression (SVR) and can achieve faster training and forecasting speed.