Wind Speed Prediction Based On Composite Characteristic Elman Neural Network

Wind speed time series prediction plays a very important role in the operation of wind farm. It is not only related to the input operation of the wind turbine, but also related to the safety, reliability and economic operation of power system. The accuracy of wind speed time series prediction directly affects the economy and stability of the wind power system, so the time series prediction is one of the key areas of domestic and foreign.In recent years, with the maturity of artificial neural network modeling, artificial neural network model is widely used as a kind of important prediction method in the prediction of wind speed time series. In the practical application, the BP neural network model is used to build the BP neural network model. Compared with the BP neural network, the Elman neural network model has a definite advantage in the complex wind speed time series forecasting. However, the Elman neural network has some disadvantages, such as easy to fall into local minimum, the prediction accuracy is poor.Because of the shortcomings and deficiencies existing in the traditional Elman neural network, in this paper we combine the hysteresis with Elman neural network method,using the smooth data constructe the composite properties of Elman neural network model. The specific approach is based on the traditional Elman neural network, the hysteresis factor is introduced into the hidden layer and the associated layer, using the hysteresis function constructed in this paper to replace the traditional Sigmoid function. Due to hysteresis function is composed of two closed curves, in data processing, according to the rules to select the curve, making data output has a random jump, jump out of local minimum value to a certain degree, using the smooth processing of data, to further improve the prediction accuracy. In this paper, the model and method are simulated and predicted by the actual data of a wind farm in North China, and compared with the traditional Elman neural network, the results show that the proposed model can significantly improve the prediction accuracy.On the basis of the proposed algorithm, the paper also realize combine of the algorithm and the actual combination, to achieve the hardware realization. The whole process of collecting, processing and displaying are successfully realized. In the ordinary PC machine can get the remote wind power generator at the next moment to predict the position of the wind.