Quantile regression based probabilistic forecasting of renewable energy generation and building electrical load: A state of the art review

With the increasing penetration of renewable energy in smart grids and the increasing building electrical load, their accurate forecasting is essential for system design, control and associated optimizations. To date, probabilistic forecasting methods have attracted increasing attentions as they can assess various uncertainty impacts. Among them, quantile regression based probabilistic forecasting methods are more popular and experience fast developments. However, there is little review that systematically covers their similarities and differences in the aspects of mechanism, feature and effectiveness in applications. This paper, therefore, provides a comprehensive review of quantile regression-related methods for renewable energy generation and building electrical load. Firstly, according to their principles/mechanisms, existing quantile regression based probabilistic forecasting methods are classified into two major categories, namely statistic-based methods and machine learning-based methods. Meanwhile, their respective strengths and limitations are comparatively analyzed and summarized. Next, their practical applications and effectiveness are systematically reviewed. On the basis of the above review part, a discussion focusing on the current research gaps and potential research opportunities is presented regarding quantile regression future developments. The timely review can help improve researchers’ understanding and facilitate further improvements of the quantile regression based probabilistic forecasting methods.