A review of predictive uncertainty modeling techniques and evaluation metrics in probabilistic wind speed and wind power forecasting
Yun Wang, Fan Zhang, Hongbo Kou, Runmin Zou, Qinghua Hu, Jianzhou Wang, Dipti Srinivasan
Abstract
Given the significant variability in wind resources, addressing the inherent uncertainty in wind energy forecasting is crucial. As a result, numerous probabilistic models have been developed, offering valuable insights into wind variability and improving forecast accuracy. This paper aims to analyze the significance of different types of uncertainty in predictive uncertainty and provides a comprehensive review of probabilistic methods for wind speed and wind power forecasting. Notably, a detailed examination of representative model structures employed for generating prediction intervals, which serve as a universal representation of predictive uncertainty, is also presented. Furthermore, this review examines the evaluators used to assess the quality of probabilistic forecasts and provides an analysis of their expression, time complexity, and usage conditions. These evaluators play a crucial role in determining the reliability and accuracy of the forecasted results. The paper also identifies five key challenges that need to be addressed to achieve accurate probabilistic wind speed and wind power forecasting. In an effort to tackle these challenges, six future trends in enhancing probabilistic forecasting performance are summarized.