Abstract: In this paper,a method based on support vector machine and genetic algorithm was proposed for the power system load forecasting．In this method,a next hour load forecast is developed by using structure risk minimization instead of traditional empiric risk minimization to mine more information from the original data.The genetic algorithm is used to optimize the SVM parameters to improve the performance of forecasting and the training speed．Historical load,atmospheric data and the calendar factors are the model inputs．Forecasting results show that this model is effective and feasible,as well as the better robustness and forecast accuracy than the BP neural method．

Key words: Power system load forecasting,Short-term load forecasting,Support vector machine,Genetic algorithm optimization