Improving Forecasting Performance of Back Propagation to Wind Speed by Square Root Transformation

Abstract

In order to improve the forecast accuracy and reliability of wind speed with strong randomness, this paper suggests a method. First, apply square root transformation to the wind speed time series, which is different from familiar normalization. Second, BP neural network is employed to forecast the future wind speed time series. Finally, invert the forecasted time series by the inverse square root transformation. The experiment result shows the effectiveness of the proposed method.