考虑新能源电力系统频率响应时空分布特性的实时分区方法  

A Real-Time Partitioning Method Considering the Spatial and Temporal Distribution Characteristics of Renewable Power System Frequency Response

作　　者：徐艳春[1] 任建新 宋文宇 席磊[1] MI Lu XU Yanchun;REN Jianxin;SONG Wenyu;XI Lei;MI Lu(Hubei Provincial Key Laboratory for Operation and Control of Cascaded Hydropower Station(China Three Gorges University),Yichang 443002,Hubei Province,China;Wuhan Power Supply Company Hongshan Power Supply Center,State Grid Hubei Electric Power Co.,Ltd.,Wuhan 430010,China;Department of Electrical and Computer Engineering,Texas A&M University,College Station 77840,Texas,USA)

机构地区：[1]梯级水电站运行与控制湖北省重点实验室(三峡大学),湖北省宜昌市443002 [2]国网湖北省电力有限公司武汉供电公司洪山供电中心,武汉市430010 [3]德州农工大学电气与计算机工程系,美国德克萨斯州卡城77840

出　　处：《电力建设》2025年第3期128-145,共18页Electric Power Construction

基　　金：国家自然科学基金项目(52277108)。

摘　　要：新能源渗透电网使得系统频率响应时空分布特性愈发凸显,现有固定分区方法难以确保任何时段内频率响应的相似性,实时分区方法提供了解决思路,但又存在可能的错误分区和耗时问题。首先,提出一种频率曲线趋势和数值距离,并与动态时间弯曲距离、欧式距离和符号聚合近似距离作对比。然后,使用层次聚类和k-means聚类算法实现系统实时分区,分区数目由silhouette-coefficient(S-C)指数、Calinski-Harabas(C-H)指数和Davies-Bouldin(D-B)指数三个指标评估得到。上述四种距离、两种聚类方法和三种指数互相组合得到24种实时分区模型,对比后选定计算频率曲线之间趋势和数值近似距离,采用层次聚类算法进行聚类,基于峰值S-C指数确定分区数目这一实时分区模型。最后,在IEEE 39系统、IEEE 118系统和实际系统进行仿真,结果表明该实时分区模型可在不同时段针对不同扰动类型和位置实时变更分区,确保了区域内频率响应的相似性,有效提高了系统分区的准确性和快速性,且在区域惯量评估方面表现较好。The integration of renewable energy into the grid significantly influences the spatial and temporal distribution characteristics of system frequency response.Existing fixed partitioning methods struggle to maintain the similarity of frequency responses across different time periods,while real-time partitioning methods often encounter challenges such as partitioning errors and excessive computational demands.To address these issues,this paper introduces a trend and value approximation method to measure the distance between frequency curves.This approach is compared against dynamic time warping,Euclidean distance,and symbolic aggregate approximation distance.Hierarchical clustering and k-means clustering algorithms are then employed to achieve real-time system partitioning.The number of partitions is evaluated using three indices:the silhouette-coefficient(S-C)index,Calinski-Harabas(C-H)index,and Davies-Bouldin(D-B)index.These four distance metrics,two clustering methods,and three evaluation indices are combined to form 24 real-time partitioning models.The optimal model is identified through comparative analysis,which involves calculating trend and value approximation distances,clustering via the hierarchical clustering algorithm,and determining the number of partitions based on the peak S-C index.Finally,simulations are performed in the IEEE 39-bus system,IEEE 118-bus system,and actual power system.The results demonstrate that the proposed real-time partitioning method dynamically adjusts partitioning results in response to different disturbances,locations,and time periods.It ensures similarity in regional frequency responses,enhances the accuracy and speed of system partitioning,and performs effectively in evaluating regional inertia.

关 键 词：频率响应 时空分布 频率曲线趋势和数值近似方法 聚类算法 实时分区 

分 类 号：TM73[电气工程—电力系统及自动化]