基于Stacking模型融合算法的风电功率预测方法  

Wind Power Prediction Method Based on Stacking Model Fusion Algorithm

作　　者：张雪原 蔡思烨 刘巧宏 朱坚 包晓炜 夏玉剑 陈极 ZHANG Xueyuan;CAI Siye;LIU Qiaohong;ZHU Jian;BAO Xiaowei;XIA Yujian;CHEN Ji(State Grid Jiading Power Supply Company,SMEPC,Shanghai 201800,China)

机构地区：[1]国网上海市电力公司嘉定供电公司,上海201800

出　　处：《电力与能源》2025年第1期61-66,共6页Power & Energy

摘　　要：随着新能源在新型电力系统中渗透率的日益增加,对风电场功率预测的准确性能要求也不断提升。为提高风电功率预测的准确性和可靠性,设计了以线性回归、K邻近、随机森林算法为特征提取层,以轻量梯度提升机为回归预测层的Stacking模型融合算法。以某风电场近年运行数据为案例,验证了该基于Stacking模型融合算法的预测方法相较于任一单一机器学习算法都具有更高的预测精度。With the increasing penetration of renewable energy in the new power system,the requirements for the prediction performance of wind farm power have gradually improved.To enhance the accuracy and reliability of wind power forecasting,a Stacking model fusion algorithm is designed,with linear regression,k-nearest neighbors(KNN),and random forest algorithms as the feature extraction layer,and a lightweight gradient boosting machine as the regression prediction layer.Using operational data from a wind farm in recent years as a case study,it is verified that the prediction method based on the Stacking model fusion algorithm achieves higher accuracy com-pared to any single machine learning algorithm.

关 键 词：风力发电 Stacking模型融合算法 随机森林 K邻近 负荷预测 

分 类 号：TM61[电气工程—电力系统及自动化]