Abstract In this paper, a hybrid wind power forecasting approach based on Bayesian model averaging and Ensemble learning (BMA-EL) is proposed. Firstly, SOM clustering and K-fold cross-validation are used to… Click to show full abstract
Abstract In this paper, a hybrid wind power forecasting approach based on Bayesian model averaging and Ensemble learning (BMA-EL) is proposed. Firstly, SOM clustering and K-fold cross-validation are used to generate multiple sets of the training subsets with the same distribution from the training set of meteorological data to increase the difference of the input samples of the base learners. These training subsets are imported into three base learners, i.e. BPNN, RBFNN, and SVM, to train the model. Then, the BMA combining strategy is trained based on the outputs of the three base learners on the validation set. Finally, the test set is combined by the BMA through the outputs of the three base learners to obtain the WPF results. By comparing the simulation error and curve between the base learner and other literature approaches, our proposed method can accurately and reliably forecast the wind power outputs under different meteorological conditions, with higher precision and reliability.
               
Click one of the above tabs to view related content.