Performance comparison of models for fast short-term wind speed prediction

Abstract

Because a microgrid system has a smaller inertia than traditional utility, accurate prediction of wind speed is an effective way to rationally adjust the scheduling strategy and to improve the operation stability and economy of the microgrid. Impacts of different model parameters on the predicted results are investigated based on the analysis of the characteristics of ridgelet, error back propagation, radial basis function artificial neural networks and support vector machine model. In addition, the four models mentioned above are used to predict the wind speed for microgrid systems with different wind speed characteristic and sample size. Quantitative evaluation results of the four models are acquired according to the proposed evaluation indexes. The results show that both artificial neural networks and support vector machine models can be used to predict the short-term wind speed for a microgrid. But the former operates faster while the latter is more accurate; Configuration of model parameters and training sample size affects the speed and accuracy of prediction to varying extents. The conclusions are instructive for the microgrid users with different prediction targets to select a proper short-term wind speed prediction model.