基于AI的电网母线负荷典型特征提取技术  

Typical Feature Extraction Technology of Power Grid Bus Load Based on AI

作　　者：李豹[1] 王巍[1] 张蔷[1] 李海坤 LI Bao;WANG Wei;ZHANG Qiang;LI Haikun(Power Dispatch and Control Center of China Southern Power Grid,Guangzhou 510623,China;Beijing Tsingsoft Technology Co.,Ltd.,Beijing 100085,China)

机构地区：[1]中国南方电网电力调度控制中心,广东广州510623 [2]北京清软创新科技股份有限公司,北京100085

出　　处：《电工技术》2022年第5期71-73,76,共4页Electric Engineering

基　　金：南方电网公司重点科技项目(编号ZDKJXM 20190020)。

摘　　要：母线负荷量级小,母线曲线特征在不同时空下的差异较明显。传统技术中,通常对呈现相对固定特征的曲线开展分析,忽略了关键的“异常用电曲线”,实用性较差。针对此种问题构建了基于聚类技术的电力负荷特征提取分析综合框架,基于海量母线负荷数据,首先利用基于密度的聚类算法提取母线典型负荷曲线,然后利用K-means算法对母线典型负荷曲线进行聚类,最后利用LOF算法对聚类结果中的异常数据进行检测,通过人工干预的方法对各异常检测结果进行单独分析,实现了对“典型”和“异常”用电曲线的全覆盖。通过对广东省内1062条实际母线进行算例验证,表明该技术框架具有可行性及实际意义。The magnitude of bus load is small,and the characteristics of bus curve are obviously different in different time and space.In the traditional technology,the analysis is usually focused on the curve with relatively fixed characteristics,ignoring the key"abnormal power consumption curve",which has poor practicability.To solve this problem,a comprehensive framework of power load feature extraction and analysis based on clustering technology is constructed.Based on the massive bus load data,firstly,the typical bus load curve is extracted by density-based clustering algorithm.Then the typical bus load curve is clustered by K-means algorithm.Finally,the abnormal data in the clustering results is detected by LOF algorithm.Through the method of manual intervention,the abnormal detection results are analyzed separately,and the full coverage of"typical"and"abnormal"power consumption curves is realized.An example of 1062 actual buses in Guangdong Province shows that the technical framework is feasible and practical.

关 键 词：母线负荷 典型负荷曲线 综合框架 提取 聚类 检测 

分 类 号：TM714[电气工程—电力系统及自动化]