Power System Short-term Load Forecast Research Based On BP Neural Network

Short term load forecasting(STLF) is an important task of running and dispatching power system,it is closely linked with ensure operating safely economically and achieve sciectific management in the power systerm.with the development of power systerm, STLF has become increasingly important and the requirements for high accuracy are increasing.Short term load changes periodically,it also has Similarity,affects by any other Special conditions and shows strong the property of the nonlinear. Because of the strong nonlinear reflection ability of Artifical neural network, we make use of improved BP neural network to forecast the short term load, and then to simulate the load with Matlab software tool.We begin with a brief introduction of present short term load forecast technology both domestic and overseas and the basic theory of artifical neural network, then it studies the BP neural network, such as basic algorithm, the advantages and disadvantages and improved methods .Sencondly it has made a depth research into ANN modelilng problem and the disfigurement of BP neural network, on which studies the inputting data processing, layers number identification,hidden nodes determination,the initial weights , the learning rate and other such things during the establishment of model and provides the theoretical derivation of the BP algorithm on the network with two hiden layers. Finally,whith a improved algorithm, establishes a BP network model whith two hidden layers to frocast the load and conducts calculation analysis, which has been proved the validity for the short-term power load forecasting by Simulation experiments on the real data and contrastive experiments, it meets the accuracy requirements of load forcasting and has a good fitness.Finally, the practical application shows the feasibility and veracity of this model.