基于数据驱动的风电功率预测误差解耦评价方法

Data-driven Decoupling Evaluation Method of Wind Power Prediction Error

针对风电功率预测中多环节交互影响的复杂性,文中提出一种风电功率预测误差的精细化评价方法,旨在利用数值天气预报、气象观测数据、风电运行数据等多源异构信息,定量分析功率预测各关键环节对预测总误差的影响程度。首先,解析了风电功率预测运行机理,将预测过程分解为数值天气预报、风能-功率转换建模、预测结果校正3个关键环节。然后,设计了基于核密度估计的风电异常数据分区间辨识方法,建立了风资源-电力特性的简化模型。最后,基于气象、电力等多源运行数据驱动,提出功率预测各环节等效误差的量度方法。算例结果表明,所提方法可定量评估各环节预测对功率预测误差的影响。

For the complexity of multi-link interaction in wind power prediction, this paper proposes a refined evaluation method for wind power prediction error. It aims to quantitatively analyze the effect of each key link of power prediction on the total prediction error by using numerical weather prediction, meteorological observation data, wind power operation data and other multi-source heterogeneous information. Firstly, the operation mechanism of wind power prediction is analyzed, the prediction process is divided into three key links, i.e., numerical weather prediction, wind energy-power conversion modeling and prediction result correction. Secondly, based on the kernel density estimation, the segmented identification method is designed for wind power anomaly data, and a simplified model of wind resource-power characteristics is established. Finally, based on the multisource operation data-driven in meteorology and electric power, a method for measuring equivalent error in each link of power prediction is proposed. The case results show that the proposed method can quantitatively evaluate the effects of each link on the power prediction error.