Wind Power Forecasting Using Parallel Random Forest Algorithm

Abstract

Wind power keeps on developing and it is broadly observed as the sustainable power source best capable to compete with fossil fuel electricity generation. In the near past wind power forecasting has improved the situation for the estimation of power production in wind farms. In general wind power forecasts are generated in two different ways one namely using Numerical Weather Prediction (NWP) and the other one using physical forecasting methods. Physical forecasting is deeply dependent on meteorological facts and the data from the NWP. The approach of physical method is vulnerable by the way that wind speeds are estimated a few feet over the ground can fluctuate. Statistical scheme encompasses models likewise ANN, SVM, and etc. less reliant on accuracy of Numerical Weather Predictions (NWP), yet relies more extremely on historical information of wind speed at respective areas. To compute large accurate wind energy forecast a good amount of real-time observations of historical observations from the wind farms becomes essential. Wind power forecasts using Support Vector Machines (SVM) and Artificial Neural Networks (ANN) suffers from slow training speed, and poor generalization ability. This paper aims at conducting experiments to assess the performance and test the suitability of the Random Forest Algorithm for wind power forecasting. The prediction results are seeming to be close with the actual wind power generated at the wind farms and it is more accurate when compared to the results of the ANN.