含照明负载用户供电智能监测算法改进的研究  

Study on the Improvement of Intelligent Monitoring Algorithm for Power Supply Containing Lighting Loads

作　　者：潘明明 李义民 游元通 唐勇兵 PAN Mingming;LI Yimin;YOU Yuantong;TANG Yongbing(China Electric Power Research Institute Co., Ltd., Beijing 100085, China;State Grid Xiamen Power Supply Company, Xiamen 361004, China)

机构地区：[1]中国电力科学研究院有限公司,北京100085 [2]国网厦门供电公司,福建厦门361004

出　　处：《照明工程学报》2021年第6期210-217,共8页China Illuminating Engineering Journal

摘　　要：电力负荷的在线监测、分解及分类识别,有利于感知负荷的变化规律和发展趋势,也有利于对电力负荷的科学管理,是建设园区智能电网的重点研究方向之一。近年来,非侵入式的负荷监测及识别引起了众多学者的关注,该算法可以节省安装和维护所需要的时间和金钱,更适合于综合型园区多类型用户特征下的碳核查、故障定位及节能诊断。照明电器是园区负载的重要组成部分,数量大,且其V-I特性和小家电区分困难,如能提高其辨识率,可以大幅度提升园区负载监控准确性。本文首先针对多类型园区用户的场景,选择了无需训练过程的k最近邻算法作为负荷监测模型,相对传统算法减少时间,然后对标准k最近邻算法容易误判少数类的缺陷,采用加权方式进行改进算法,并根据V-I轨迹缺失数值特征的不足,提出了基于综合相似度的类别判决方法,其次通过幅值特征分析进一步改进了算法,最后特别选择了照明电器、空调、风扇等小家电与其他多种不同设备负荷特征混合的数据集和实验室数据,验证了改进后算法在含照明负载时监测性能的提升。On-line monitoring,decomposition,classification and identification of power load is conducive to the perception of load changes and development trends,and is also conducive to the scientific management of power loads.It is one of the key research directions for building smart grids in the park.In recent years,non-intrusive load monitoring and identification have attracted the attention of many scholars.This algorithm can save time and money required for installation and maintenance,and is more suitable for carbon verification,fault location and fault location under the characteristics of multiple types of users in comprehensive parks.Energy saving diagnosis.Lighting appliances are an important part of the park’s load,and the number is large,and its V-I characteristics are difficult to distinguish from small appliances.If the recognition rate can be improved,the accuracy of park load monitoring can be greatly improved.This paper first selects the k-nearest neighbor algorithm without training process as the load monitoring model for the scene of multiple types of park users,which reduces the time compared with the traditional algorithm.Then,the standard k-nearest neighbor algorithm is easy to misjudge the minority defects,and the weighted method is adopted.Improve the algorithm,and according to the lack of the missing numerical features of the VI trajectory,a category judgment method based on comprehensive similarity is proposed.Secondly,the algorithm is further improved through amplitude feature analysis.Finally,small appliances such as lighting appliances,air conditioners,fans,and others are selected.A variety of data sets and laboratory data mixed with different equipment load characteristics verify the improved algorithm’s monitoring performance when the lighting load is included.

关 键 词：照明负载 非侵入式负荷监测 多类型园区用户 智能监测算法 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]