Spatio-temporal independent applicability of one time trained machine learning wind forecast models: a promising case study from the wind energy perspective

Abstract

The extensive grid integration of wind electricity necessitates accurate wind speed forecasting to maintain grid stability and quality of power. In the usual practice of wind speed forecasting using machine learning models, the models are trained and tested for the same location and the testing is performed adjacent to the training data. In this work, we investigate the prospect of employing one time trained machine learning predictive models for wind speed forecasting independent of time and location. It is revealed in our investigations that machine learning models trained with historical wind speed data from one location are capable of predicting wind speeds effectively at other locations of interest within a wide geographical region. The findings in this study are useful to the wind energy industry for wind speed forecasting at multiple locations with less computational effort and enhanced speed and productivity.

Keywords

• wind energy
• wind speed forecasting
• machine learning
• cross-prediction
• mutual information