Q-stacking prediction model considering wind speed error correction

Abstract

In order to improve the accuracy of wind speed forecasting in wind farms, an ensemble-enhanced combined forecasting model is proposed considering error correction. First establish five independent base learners, build a two-layer Stacking ensemble model to fuse the prediction results of each base learner, and divide the input data by cross-validation to improve the generalization ability of the model. Then use the model-free learning framework Q learning selects the optimal model in the base learner to correct the preliminary prediction error and obtain the final prediction result. Select the actual wind farm measured data in different seasons to simulate the prediction effect of the model, and verify the prediction ability of the proposed model through comparative analysis. The results show that the model has high prediction accuracy with ε = 0.093.