As an important renewable energy source, wind energy is significant for realizing energy transition and reducing carbon emissions. With the increasing penetration of wind energy in the global energy system, higher prediction accuracy is needed to ensure the safe and stable operation of the power grid. However, the existing wind power prediction methods are constantly pursuing model improvement, ignoring the importance of data quality to the prediction performance, resulting in a stagnation of the upper limit of prediction accuracy. In this paper, we establish a comprehensive wind power prediction system based on correct multi-scale clustering ensemble, similarity matching, and an improved whale optimization algorithm. Firstly, multiple classification algorithms combined with meteorological data are used to correct the extreme scenarios in the clustering results. Secondly, a library of typical fluctuation patterns is established based on the clustering ensemble results, and the optimal training dataset is determined by similarity matching. Finally, complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) is used to extract further the power data's local features and time-frequency characteristics and to predict the modal components using the improved whale optimization algorithm(IWOA)-optimized BiLSTM network. The results of the three sets of experiments show that the proposed model is able to improve more than 10% in terms of MAE, RMSE, and MAPE compared to other models, and the model robustness is high.
