基于改进kShape聚类的谐波污染分区方法
Harmonic pollution partitioning method based on improved kShape clustering
张敏 1樊瑞 1祗会强 1张世锋 1李慧蓬 1赵军1
作者信息
- 1. 国网山西省电力公司 电力科学研究院,山西 太原 030001
- 折叠
摘要
针对大量电力电子负荷接入后,谐波源数量大幅增加,全网谐波源位置追溯困难的问题,利用电能质量监测数据,提出一种基于双层聚类的谐波分区溯源方案.首先,使用基于波形相似性的kShape时间序列聚类算法,通过计算谐波电压序列的形态距离来度量数据波动相似性,挖掘谐波污染关联信息;然后,引入自适应密度峰值聚类改进kShape算法,解决初始聚类中心随机选取导致的局部最小化问题,实现最佳聚类数目的自适应选择.该方法能够有效实现多谐波源的区域化定位,缩小主导谐波源的嫌疑范围,适用于大规模谐波源接入场景的溯源分析.基于IEEE 123 节点网络和监测平台的实测数据,验证了方法的有效性和实用性.
Abstract
With the integration of numerous power electronic loads,the number of harmonic sources has significantly increased,making it more challenging to trace the locations of harmonic sources across the entire network.This paper proposes a harmonic partition tracing scheme based on a two-layer clustering approach using power quality monitoring data.First,the kShape time series clustering algorithm,based on waveform similarity,is employed to calculate the shape distance between harmonic voltage sequences,measuring data fluctuation similarity and uncovering information related to harmonic pollution.Then,adaptive density peak clustering is introduced to improve the kShape algorithm,addressing the issue of local minimization caused by the random selection of initial cluster centers and enabling the adaptive selection of the optimal number of clusters.This method effectively achieves regional localization of multiple harmonic sources,narrows down the suspect range of dominant harmonic sources,and is suitable for tracing analysis in scenarios with large-scale integration of harmonic sources.The effectiveness and practicality of the proposed method are validated using measured data from the IEEE 123-node network and the monitoring platform.
关键词
谐波溯源/谐波污染分区/kShape聚类/时间序列相似性/自适应密度峰值聚类Key words
harmonic tracing/harmonic pollution zoning/kShape clustering/time series similarity/adaptive peak density clustering引用本文复制引用
基金项目
国网山西省电力公司科技项目(52053022000A)
出版年
2024
国家科技期刊平台
NSTL
万方数据