关键词: ＷＲＦ模式, 随机森林, 天鹰优化算法, 非参数核密度估计, 风电功率区间预测

Abstract: To solve the problems that the pre-set parameters of the random forest algorithm depend on empirical settings and the randomness of wind power deterministic prediction is difficult to describe, this paper proposes an interval prediction method based on a combination of the Aquila optimizer (AO), Random Forest (RF) and nonparametric kernel density estimation (NKDE). Firstly, the AO is combined with RF for power single-point value prediction, on which basis NKDE is introduced for wind power interval prediction to provide more information for grid scheduling and optimal allocation. According to the proposed method, a comparison experiment is then conducted on the wind speed data derived from a wind farm in Rudong County, Jiangsu Province using WRF model forecasts. The experiments show that the AO-RF-NKDE interval prediction model can provide wind power fluctuation intervals with a better comprehensive performance, which has a higher application value for reducing wind power uncertainty and weakening grid fluctuation.