Systematic Comparison of Feature Selection Methods for Solar Energy Forecasting

Abstract

Over the last few years, solar energy forecasting has been increasingly adopted as a sustainable low energy solution for a smart environment. In addition to the various explainable artificial intelligence tools that have been applied to solar energy forecasting, a feature selection process has become an absolute prerequisite to improve the effectiveness of the model building. In this chapter, we raise interest in the potential of feature selection processing by providing a fundamental taxonomy of some feature selection techniques and examining their usefulness, variety, and capability to handle a solar energy prediction problem based on meteorological and geographical data. The experimental results show that the investigated feature selection methods can significantly improve the prediction process. The features selected differ from one method to another, depending on the considered data constraints.