Power Load Forecasting Based On Improved Elman Neural Network

In order to ensure the safety and reliability of power system operation, it is necessary for power dispatching departments to predict the future power load. Artificial neural network has the advantages of self adaptability, self-learning and fault tolerance.Therefore, it has been applied to the field of power load forecasting. Among the artificial neural networks, Elman neural network is more widely used.Firtly, this paper introduces the basic concept and research significance of load forecasting, disicusses the current research situation of load forecasting and Elman neural network at home and abroad. It also analyzes the error index of load forecasting and the reasons why the errors happed. All of those build up the foundation of follow-up work the improvement. Secondly, the basic principle, netwark structure, prediction on process, learning process and learning algorithm of Elman neural network are mainly studied. Againse Elman neural network is lacking of dynamic information proccessing capability, Sigmoid function affects the speed of convergence,the traditional algorithm in the learning process will appear oscillation problems, this article tries to improve Elman neural network in the aspects of network structure, incentive function and learning algorithm. Taking the historical load data of a city power grid in Anhui as an example, a simulation carried out by using MATLAR software is made.The results shwo that the improved Elman neural network model for load forecasting has chieved better effect.It improves the training speed and the loed forecast value is more close to the actual value. Besides, the absolute error and relative errors are reduced. Finally, in order to improve the accuracy of load forecastin further,combining wavelet decomposition which has good characteristics of partial time-frequency with the improved Elman neural network are applied to short-term load forecasting. It is still taking the historical load data of Anhui province power grid as an example,which is also simulated by using MATLAB software. The results show that the prediction accuracy has been further improved.