Multi-step forecasting of wind speed using IOWA operator

It is difficult to accurately predict wind speed because of its random and non-stationary property. In order to improve accuracy of wind speed forecasting, this paper researched multi-step combination forecasting of wind speed on the base of traditional wind prediction models. BP neural network, RBF neural network, ELMAN neural network, support vector machine (SVM) and GMDH neural network were respectively used to establish wind speed forecasting models. Then induced ordered weighted averaging (IOWA) operator-based forecasting method was used to build combination forecasting model for multi-step forecasting of wind speed. Wind speed data of ten minutes sampling and one hour sampling was used to test the forecasting performance. The results show that the combination forecasting model significantly improves the accuracy and reliability of wind speed prediction.