Short Term Electricity Load Forecasting Based On Kernel Principal Component Analysis And GA-BP Neural Network

Short-term load forecasting of power system is a major foundation for power attemper department to set down generating electricity plan and arrange attemper plan,power supply plan and bargaining plan under market environment,it is mainly applied in forecasting the power load of the next few hours or one day to several days.High efficiency and precision load forecast system is the guarantee of that electric power market supply and demand balance and power system operates reliably,and the key issues of power load forecasting is the technical problems and mathematical modeling problems of the forecasting,The main work of this paper is to study and improve the BP neural network model,and to improve the prediction accuracy and efficiency.This paper firstly studied the all kinds of existing short-term and their advantages and disadvantages.In the existing methods,BP neural networks have been applied in power load forecasting because of their own adaptive self-learning,high fault tolerance,and a series of advantages,and it has achieved the ideal results.But BP neural networkhas some of its own shortcomings,the problem of slow convergence and easy to fall into local minima.In order to overcome these problems,this paper propose a kernel principal component analysis' s genetic BP neural network model.The model use the global search performance of genetic algorithm to optimize the weight threshold of the BP neural network,can effectively overcome the problem of local convergence of the BP algorithm.In order to ensure the accuracy of short-term load forecasting,we need to consider many correlative factors,there are nonlinear correlation and redundant information among the factors,and these factors can be used as input variables of neural inputs,we know to many input variables increase the training burden and reduce the prediction efficiency.So,KPCA was applied for dimension reduction optimization on input numbers of training samples so as to replace original great deal of input information by less input,and this method can delete some redundant information retain most information,so the convergence speed and the prediction efficiency is improved.In the paper,the model is simulated by MATLAB software,this models are divided into BP neural network model?GA-BP neural network model?KPCA-GABP three models,the three models are used to simulate the power system load in a certain area of HuNan.Finally,the simulation results show that the KPCA-GABP prediction model is superior to the other two models in prediction accuracy and efficiency.Then in order toshow that the kernel principal component analysis method can extract the nonlinear information better than the principal component analysis method,so as to improve the accuracy of short-term load forecasting,we proposed a PCA-GABP prediction model and compared with the KPCA-GABP model.Finally,the simulation results show that the KPCA-GABP model is better than the PCA-GABP model in prediction accuracy.