Research On Power Short-term Load Forecasting Based On Neural Network

Power system short-term load forecasting is a very important content of power department daily work scheduling, the accuracy of short-term load forecasting has a direct influence on the security, reliability and economy of power system. Therefore, the study of short-term load forecasting is an important direction of research scholars at home and abroad.As a kind of intelligent algorithm, artificial neural network has been widely studied and applied in the prediction of short-term electric power load forecasting in recent years. In this paper, the network structure and learning rules etc. of the artificial neural network are introduced. Then, based on BP neural network and Elman neural network, mathematical model of short-term load forecasting was established.In the modeling process of BP neural network, the first step is data preprocessing, including the correction of the abnormal data by transverse comparison and longitudinal comparison, and the data were normalized, then the learning algorithm was optimized by Lenvenbery-Marquard algorithm. In the process of training, data was input repeatedly to improve the utilization rate of samples and the efficiency network training, and multiple network parameters were selected among reasonable range, the group with best training effect was applied in the model.Elman neural network is a typical dynamic network, which is added a layer to undertake internal state storage based on the basic structure of BP neural network, so that the system has the ability to adapt to the time-varying characteristics. In this paper, the model was optimized from the aspects of the network algorithm, excitation function and network structure during Elman neural network modeling process, in order to improve the network convergence speed and prediction accuracy.In the simulation process, more than two mathematical models used the same data samples for short-term load forecasting. By comparing the prediction error and prediction accuracy, it is can be seen that Elman neural network model is superior to the BP neural network model.