Abstract
As a response to rapidly increasing penetration of wind power generation in modern electric power grids, accurate prediction models are crucial to deal with the associated uncertainties. Due to the highly volatile and chaotic nature of wind power, employing complex intelligent prediction tools is necessary. Accordingly, this article proposes a novel improved version of empirical mode decomposition (IEMD) to decompose wind measurements. The decomposed signal is provided as input to a hybrid forecasting model built on a bagging neural network (BaNN) combined with K-means clustering. Moreover, a new intelligent optimization method named ChB-SSO is applied to automatically tune the BaNN parameters. The performance of the proposed forecasting framework is tested using different seasonal subsets of real-world wind farm case studies (Alberta and Sotavento) through a comprehensive comparative analysis against other well-known prediction strategies. Furthermore, to analyze the effectiveness of the proposed framework, different forecast horizons have been considered in different test cases. Several error assessment criteria were used and the obtained results demonstrate the superiority of the proposed method for wind forecasting compared to other methods for all test cases.