Innovative ensemble system based on mixed frequency modeling for wind speed point and interval forecasting

Abstract

Wind speed forecasting can improve wind energy utilization and is thus highly significant for wind power systems; however, it is a challenging process. Forecasting techniques from previous research were mainly developed using single-frequency-based time series modeling, which has recently become a popular approach. Although mixed-frequency-based modeling has significant potential, research applying this approach to wind speed forecasting is almost nonexistent. To fill this research gap, we developed an innovative ensemble system based on mixed frequency modeling to perform wind speed point and interval forecasting, comprising data preprocessing, mixed frequency modeling, and ensemble forecasting modules. Unlike the methods used in most previous studies, this system exploits the tremendous potential of mixed frequency data. In addition, a multi-objective optimizer-based ensemble forecaster was devised to provide deterministic and uncertain information regarding future wind speed. Furthermore, data preprocessing based on different strategies was applied to improve the performance of multiple forecasting models. An empirical study based on two wind farms shows that the system outperforms benchmark techniques and can be employed for data monitoring and analysis in wind farms or other fields.