基于改进型密度峰值算法的电力负荷聚类分析  

Clustering Analysis of Power Load Based on Improved Density Peak Algorithm

作　　者：曾四鸣 李铁成 李顺[2] 梁纪峰 范辉 杨军[2] 吴赋章 ZENG Si-ming;LI Tie-cheng;LI Shun;LIANG Ji-feng;FAN Hui;YANG Jun;WU Fu-zhang(Electric Power Research Institute,State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050001,China;School of Electric Engineering,Wuhan University,Wuhan 430072,China;State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050001,China)

机构地区：[1]国网河北省电力有限公司电力科学研究院,石家庄050001 [2]武汉大学电气与自动化学院,武汉430072 [3]国网河北省电力有限公司,石家庄050021

出　　处：《科学技术与工程》2022年第25期11032-11040,共9页Science Technology and Engineering

基　　金：河北省省级科技计划新一代电子信息技术创新专项(20314301D)。

摘　　要：在海量异质灵活资源参与含高比例新能源电网的运行调节背景下,针对用户用电特性分析的准确性、鲁棒性、计算效率的高要求问题,提出了一种基于特征指标完善和改进型密度峰值算法的电力负荷聚类分析方法。首先,通过提取9个完备的特征指标进行指标降维和完善以代替日负荷曲线组成的功率向量作为聚类输入;其次,采用熵权法对各项特征指标赋予权重保证负荷曲线的形态特征;最后,采用一种改进型密度峰值聚类算法对日负荷进行聚类分析。基于某地区实际负荷数据进行算例分析,结果表明,所提方法在鲁棒性、聚类质量等方面相比于传统电力负荷聚类算法均具有优越性,聚类结果能真实有效地反映用户的实际用电特性,为制定精准的电力用户画像、需求侧响应策略提供了态势感知基础。Under the background of massive heterogeneous flexible resources participating in the operation regulation of power grid with high proportion renewable energy, an analysis method for power load clustering based on complete feature index and improved density peak algorithm to the high requirements of accuracy, robustness and computational efficiency in the analysis of user power consumption characteristics was proposed. Firstly, nine complete characteristic indexes were extracted for index reduction and improvement to replace the power vector composed of daily load curves as clustering input. Secondly, the entropy weight method was used to assign weight to each characteristic index to ensure the morphological characteristics of load curves. Finally, an improved density peak clustering algorithm was applied in clustering analysis of daily load. Case studies were carried out based on a certain area actual load data. The results show that the proposed method is superior to the traditional power load clustering algorithm in terms of robustness and clustering quality. The clustering results can reflect the actual power consumption characteristics of users truly and effectively, which will provide a situational awareness basis for formulating accurate user portrait and demand response strategy.

关 键 词：电力负荷聚类 特征指标 改进型密度峰值算法 海量异质灵活资源 高比例新能源 

分 类 号：TM714[电气工程—电力系统及自动化]