An Optimal Combined SVM Model for Short-term Wind Speed Forecasting

A high precise wind speed forecasting method is one of current wind power research hotspots. This paper presented a combined wind speed forecasting model based on support vector machine (SVM) optimized by particle swarm optimization (PSO) using historical data of wind speed at the site. The model took the results of back propagation neural network (BPNN), radial basis function neural network (RBFNN), genetic neural network (GNN) and wavelet neural network (WNN) as the inputs, and adopted the actual wind speed as the output. Meanwhile, particle swarm optimization was used to optimize model parameters. Apply this model in hourly prediction of wind speed using historical data from a wind farm in Shanxi Province. It is observed that its prediction accuracy was not only higher than that of any of its single network but higher than traditional linear combined forecasting model and neural network combined forecasting model.