Time series based support vector machine (SVM) model is provided for short term price forecasting. The Structure Risk Minimization (RSM) principle is embedded into the SVM, so on the basis of learning by fewer samples the presented model can conduct fast and accurate forecasting. It has better generalization. In this method, except considering main influential factors such as previous competitive load, system rotary reservation, competitive generating capacity etc, the past price data which are time series style or not hive been included as attributes in input parameters. A day ahead MCP forecasting model is established. The results show that the proposed model has better forecasting accuracy in practical application.
