Research On New Power Price Forecasting Methods Based On Structural Risk Minimum

Along with the electric market reforming, the way of forming electric power price has experienced a revolution-from form scheduled to today competing. Power price changes along with demand, and also influences demand in reverse. It has been gradually revealing its importance as a lever of adjusting power producing and consuming.In the power market, price is the final footing tool. As the benefit of generating,supplying and consuming part varies as power price changed, getting price information previously is of great importance to every participant of power market. As long as power price being precisely forecasted , power generators could make realistic production plan and make more benefit; the macroscopic readjustment department could control and readjust economic more efficient ; the planning department could make up dispatching plan and also construction plan of generators and transmission lines, etc.Before forecasting power price, we must analysis its forming process and attribute. As to widely used Market Clearing Price (MCP), market price is define as the marginal price. MCP take on the features of fluctuate and periodicity. There are four important factors that can affect power price: historical price,load demand,generator bidding strategy and time.To get a satisfied result of price forecasting , one must get a efficient tool. Then we tend to Support Vector Machine (SVM). In this paper , the structural risk minimization property of SVM is researched, which is the most important property of SVM and on which SVM theory is build. Then the paper introduces SVM theory and models. This paper also design models consist of data mining theory and SVM to forecast power price, the result of which is satisfying. After this, the paper associated Genetic Algorithm (GA) theory and data analysis method , and finally constructed self-conduct SVM. Result of calculating affirmed the efficiency of this method. At last , the paper proposed several possible points to carry on. the research of SVM as also as power price forecasting.