State-of-the-art one-stop handbook on wind forecasting technologies: An overview of classifications, methodologies, and analysis

摘要

Wind energy has the inherent nature of intermittence and randomness, such that its accurate prediction is extremely critical to ensure safe and stable operation of power system with increased wind power integration. This paper aims to present a thorough state-of-the-art one-stop handbook on various approaches applied in wind forecasting based on three perspectives: wind speed and power forecasting, uncertainty forecasting, and ramp events forecasting. Firstly, four classifications of wind data according to data source, along with twenty-seven data pre-processing technologies which can efficiently improve prediction precision are carefully addressed. Then, a systematic literature review of model principle based wind speed and power forecasting strategies is investigated, which is categorized into three groups: physical approaches, statistical approaches, and combined approaches. Peculiarly, it has reported that combined approaches can realize more than 70% accuracy improvement compared with single model. Besides, three novel technologies developed in recent years are also discussed, e.g., spatial correlation forecasting, regional forecasting, and offshore forecasting. Moreover, thirty-seven evaluation criteria of wind speed and power forecasting are thoroughly summarized for performance verification. Additionally, uncertainty forecasting and ramp events forecasting which can provide more risk information for operators to handle decision-making issues in power system are also elaborated. Lastly, eight recommendations for further development in wind forecasting are also proposed. The prominent merit of this work is that a total of one hundred and seven wind forecasting methods from three perspectives are comprehensively summarized and compared based on inputs, time-scale, space-scale, forecasting variables, metrics, and features, which aims to help readers more effectively utilize these approaches for future in-depth researches. (C) 2020 Elsevier Ltd. All rights reserved.