Particle Swarm Optimization for Parameter Optimization of Support Vector Machine Model

Support Vector Machine (SVM) is a type of learning machine which has been proved to be available in solving the problems of nonlinear regression. The decision of SVM parameters is essential. In this paper a new SVM model based on particle swarm optimization (PSO) for parameter optimization has been proposed. PSO algorithm has extensive capability of global optimization. Once the PSO finds the optimal parameters of SVM, the model can be optimized. This new model is applied in short-term load forecasting of electric power system. The history load dada are the training data and the forecasting error is taken as the optimization objective. The simulation results show that both precision and efficiency have been improved by using PSO-SVM model than by using the single SVM model. The new model provides an alternative for forecasting electricity load due to its practicality and efficiency.