A novel adaptively combined model based on induced ordered weighted averaging for wind power forecasting

Accurate wind power prediction is a vital factor in day-ahead dispatch and increasing the high penetration of renewable energy integration. Some combined prediction models based on weight are proposed to provide uncertain information about wind power. However, the optimal weight issues to be addressed are still a tough problem. This paper proposes an adaptively combined prediction model based on the induced ordered weighted averaging operator. First, four classical prediction models are used as a predictor to forecast wind power. Second, the optimal weight is expressed as an optimization problem to minimize the prediction errors with the nonnegative constraints and summation constraint of weights, and the optimal optimization problem is transformed into a quadratic programming model, which can be easily solved to determine the optimal weights for multiple prediction models. Third, an interval prediction mode based on errors distribution produced by the proposed combined model with different confidence intervals is constructed, which can perform deterministic and probabilistic prediction of wind power simultaneously. Furthermore, the model's effectiveness is comprehensively evaluated using both deterministic and probabilistic approaches. This analysis is further enriched with a comprehensive examination of a year-round dataset obtained from a wind farm cluster in Ningxia, China. The data, recorded at 15-min intervals, offer robust evidence of the combined model's superior performance. The results show that the proposed approach can achieve higher accuracy than other benchmark prediction models, further verifying its reliability for wind power dispatch.