Short-term wind power forecasting using adaptive neuro-fuzzy inference system combined with evolutionary particle swarm optimization, wavelet transform and mutual information

The non-stationary and stochastic nature of wind power reveals itself a difficult task to forecast and manage. In this context, with the continuous increment of wind farms and their capacity production in Portugal, there is an increasing need to develop new forecasting tools with enhanced capabilities. On the one hand, it is crucial to achieve higher accuracy and less uncertainty in the predictions. On the other hand, the computational burden should be kept low to enable fast operational decisions. Hence, this paper proposes a new hybrid evolutionary-adaptive methodology for wind power forecasting in the short-term, successfully combining mutual information, wavelet transform, evolutionary particle swarm optimization, and the adaptive neuro-fuzzy inference system. The strength of this paper is the integration of already existing models and algorithms, which jointly show an advancement over present state of the art. The results obtained show a significant improvement over previously reported methodologies. •A new hybrid evolutionary-adaptive methodology is proposed for wind power forecasting.•Mutual information, wavelet transform, EPSO and ANFIS are combined in a proficient way.•Results from a real-world case study in Portugal are reported, along with a comprehensive comparison.•Higher accuracy and less uncertainty in predictions are attained, alongside low computational burden.