Short-term Load Forecasting Based On Bayesian Theory Of Support Vector Machine

Power system short-term load forecasting is an important reference for power system security and economic operation, With the steady progress in China’s reform of marketization of power industry, Electric power company must promptly grasp the information of load change, People begin to pay more and more attention to importance of short-term load forecasting, but also for short-term load forecasting accuracy is put forward higher requirements.Compared with traditional methods, support vector machine has nonlinear fitting, strong generalization ability, training the advantages of fast convergence rate, and limited to a small sample problem with special processing capacity. Using advantages of support vector machine and based on the nonlinear characteristics of in the short-term load forecasting, Proposed a method for the short-term load forecasting that based on support vector machine.In order to overcome the uncertain data information effect on short-term load, on the basis of research on support vector machine method, and then proposed a kind of based on bayesian theory of support vector machine, a new method for short-term load forecasting.The main research of this paper is as follows:1. Making a detailed analysis on short-term load characteristics at first, and then build short-term load forecasting model based on the SVM, and through the example that compared to the neural network method which has been widely applied to the load forecasting, Analysis of prediction results, verified its good prediction performance.2. Because of the influence of uncertain factors including the weather, random holidays and so on, In order to effectively overcome the uncertain data information effecting on short-term load, Adding the theory of bayesian, Proposed to the method of combining by bayesian theory and support vector machine, Improving its study training mechanism and build a forecasting model based on bayesian theory of support vector machine.3. In order to make the prediction accuracy higher, the computational complexity lower, keep the model stability, Using the bayesian evidence framework of inference rules make the optimization choice for related parameters in the model and nuclear parameter.4. Making the instance of short-term load forecast based on the bayesian theory of support vector machine forecasting model, and comparing with predicted results of the SVM prediction model analysis. The simulation experiment indicates that based on bayesian theory of the support vector machine forecasting method is not only improved the prediction precision, but also effectively overcome the influence of the uncertain data information for short-term load change, the stability of prediction has been greatly improve overall.