Load Forecasting Based On Elman Neural Network Optimized By Beetle Antennae Search Optimization

Power system load forecasting is the basic content of power system operation management and real-time control. The accuracy of prediction results is of great significance to the security, optimization and economic operation of power system. Elman neural network has a lot of research in data prediction, in which the method of calculating weight and threshold is gradient descent method. In this paper, beetle antennae search optimization is used to find the weight and threshold of Elman neural network, and apply it to the set network, so as to construct the final training model. The model constructed by this method can overcome the problems of poor stability and easy to fall into local optimization of standard Elman neural network. According to the load data of each hour of the power company from 2008 to 2009, as well as the daily temperature, holiday type and other data, the load of 12:00 and 6 pm every day in January 2010 is predicted. It is verified that the accuracy of this model is higher than that of Elman neural network. The experimental results proved that the algorithm is competitive.