基于聚类边缘筛选和改进SBR的风电极端场景提取方法  

Wind Power Extreme Scenario Extraction Method Based on Clustering Edge Screening and Improved SBR

作　　者：侯佳辰 孙际友 余洋 罗倩 王现勋 HOU Jia-chen;SUN Ji-you;YU Yang;LUO Qian;WANG Xian-xun(College of Resources and Environment,Yangtze University,Wuhan 430100,Hubei Province,China;Henan Water Conservancy Investment Group Co.,Ltd.,Zhengzhou 450000,Henan Province,China;Middle Changjiang River Bureau of Hydrology and Water Resources Survey,Wuhan 430010,Hubei Province,China)

机构地区：[1]长江大学资源与环境学院,湖北武汉430100 [2]河南水利投资集团有限公司,河南郑州450000 [3]长江水利委员会水文局长江中游水文水资源勘测局,湖北武汉430010

出　　处：《中国农村水利水电》2025年第3期225-230,共6页China Rural Water and Hydropower

基　　金：长江水利委员会智慧水文智能控制重点实验室开放基金资助项目(KLICSH-001,KLICSH-002)。

摘　　要：基于现有风电极端场景提取方法存在提取小概率场景的针对性不足以及描述极端场景发生概率不准确等问题,提出一种基于K-means赋值聚类边缘筛选方法和改进同步回代缩减(Simultaneous Backward Reduction,SBR)算法相结合的极端场景提取方法。首先通过形维、量维特征值将风电日出力场景集赋值为空间点集并进行K-means聚类,接着依照给定概率筛选出与各聚类中心点加权距离较大的边缘场景。其次使用改进SBR算法以保留边缘场景的原则进一步去除相似场景。最后采用我国西北某省电网风功率实际数据进行了算例分析,通过布莱尔分数(Brier score,BS)指标验证了所提方法的有效性,该方法能够提高SBR极端提取精度约60%及其计算效率,将为电力系统规划与运行提供更准确的数据支撑。Existing methods for extracting extreme scenarios of wind power generation face issues such as insufficient pertinence in extracting low-probability scenarios and inaccuracies in describing the probabilities of occurrence of extreme events.To address these issues,this paper proposes a method for extracting extreme scenarios that combines K-means assignment clustering edge screening approach with an improved simultaneous backward reduction(SBR)algorithm.Firstly,using contoured and quantitative feature values,the daily wind power output scenarios are assigned to spatial point sets and subjected to K-means clustering.Then,edge scenarios with larger weighted distances from each cluster center are selected according to a given probability.Secondly,the improved SBR algorithm is used to further eliminate similar scenarios based on the principle of retaining edge scenarios.Finally,an example analysis is conducted using actual wind power data from a provincial grid in Northwest China.The effectiveness of the proposed method are validated using the Brier score(BS)indicator,demonstrating that the proposed method can improve the extreme scenario extraction accuracy of SBR by approximately 60%and enhance computational efficiency,which will provide more accurate data support for power system planning and operation.

关 键 词：风电出力 K-means赋值聚类 同步回代缩减 BS指标 

分 类 号：TM614[电气工程—电力系统及自动化]