Wind speed forecasting based on fuzzy-neural network combination method

Wind speed forecasting is significant for power system operation. Accuracy and rapidity are essential requirements. This paper is proposed a combination method based on a fuzzy neural network. The combination weight of the proposed hybrid model which combines an Autoregressive Integrated Moving Average (ARIMA) model and a Radial Basis Function Neural Network (RBFNN) model is adjusted online according to the previous error. To validate the proposed method effectiveness and performance, two cases of wind data have been utilized. The first one is a simulation wind case and created by TurbSim code, which is developed by National Renewable Energy Laboratory. The other one is a real-time wind speed case and its wind farm is located in northern Gulf of Mexico. Comparing with the single models of ARIMA, RBFNN, hybrid mean strategy and variance strategy, the result of fuzzy-neural network combination method reduce both mean square and mean absolute errors.