A Real-time Prediction Algorithm for Wind Power Generator

In recent years, many literatures used history test data, build real-time prediction algorithm for wind power generator and analyze the prediction error influence by wind turbine assembly. Based on this situation, this paper proposed three kinds of prediction model of wind power: BP neural network with self similar network model, time series analysis of ARMA (3,3) model, wavelet neural network prediction model. For the same set of fan, the three models of real-time prediction error were controlled in 40%, 10% and 7%.By analysis we find that, the aggregation number of wind turbines increases, the relative prediction error percentage will decreases. In order to improve the accuracy of real-time forecasts of wind power, we use filter method to process ARMA (3, 3) model, we find a new model with higher in real-time prediction accuracy, Analysis of the forecast results, this method can improve the accuracy of forecasting.