Wind power forecasting based on variational mode decomposition and high-order fuzzy cognitive maps

Abstract

Accurate wind power forecasting can effectively reduce the adverse effects of wind power forecasting errors on wind power grid integration and power dispatch. However, current wind power forecasting technology, such as the method based on machine learning, belongs to the black box model and is not solvable. Variational mode decomposition (VMD) is a decomposition technique based on the time–frequency characteristics of the original time series, which has a mathematical theoretical foundation. Besides, fuzzy cognitive map (FCMs) is a kind of soft computing method with strong knowledge representation and reasoning ability. Therefore, to enhance the forecasting accuracy of wind power, in this paper, a novel time series forecasting method based on improved VMD (IVMD) and high-order FCM (HFCM), namely IVMDHFCM is proposed. IVMD can effectively extract the features in the raw time series depending on the time–frequency characteristics of the time series. Then, the subseries obtained by IVMD are modeled and forecasted by HFCM, and the Bayesian ridge regression method is adopted to learn the weight of HFCM. Finally, the differential evolution (DE) algorithm is used to get the optimal hyperparameters of IVMDHFCM. The performance of IVMDHFCM is verified by comparing it with that of state-of-the-art methods on ten publicly available datasets. Moreover, the proposed IVMDHFCM is compared with the existing HFCM based method on ten actual wind power datasets. The results show that the IVMDHFCM can effectively improve the accuracy of wind power forecasting and reduce the forecasting error. Besides, the IVMDHFCM can also effectively explore the fluctuations of wind power.