Short-Term Wind Speed Prediction Based on the Weighted Regular Extreme Learning Machine

Abstract

Accurate wind speed is the basis of the wind power prediction. And it is of great significance to the parallel operation of the wind farm. So it is to the maintenance of the safety and stability of the power system. In the view of the strong volatility and randomness of the wind speed time series, it is difficult to be predicted. A new method of short-term wind speed prediction is established based on the weighted regular extreme learning machine (Weighted Regularized Extreme Learning Machine, WRELM). First, the wind speed time series and the wind direction time series which have high correlation with the wind speed are taken account. Besides the meteorological factors are also taken as the candidate sets such as temperature, pressure, humidity and so on. Then the maximal relevance minimal redundancy (mRMR) principle is used to select the maximum serial correlation properties. And those are taken as the prediction inputs. Afterwards, the train set and test set of the prediction network are fixed to establish WRELM. Then, the network parameters are trained with the train set data. And the WRELM prediction model is built. Finally, the WRELM network is adopted to predict the short-term wind speed and the future wind speed are obtained. The data from the wind farm is carried out to do the experiment. And it wants to prove the effectiveness of the new method.