电力工程技术2024,Vol.43Issue(1) :165-173,211.DOI:10.12158/j.2096-3203.2024.01.018

基于先验统计模型的非侵入负荷辨识算法

Resident non-invasive load identification algorithm based on prior statistical model

赵成 宋彦辛 周赣 冯燕钧 郭帅 李季巍
电力工程技术2024,Vol.43Issue(1) :165-173,211.DOI:10.12158/j.2096-3203.2024.01.018

基于先验统计模型的非侵入负荷辨识算法

Resident non-invasive load identification algorithm based on prior statistical model

赵成 1宋彦辛 1周赣 2冯燕钧 2郭帅 1李季巍1
扫码查看

作者信息

  • 1. 国网北京市电力公司电力科学研究院,北京 100080
  • 2. 东南大学电气工程学院,江苏南京 211189
  • 折叠

摘要

针对传统非侵入负荷辨识技术中电热细分能力不足的问题,文中提出了一种基于先验知识与统计学习模型的居民非侵入式负荷辨识算法.文中对洗衣机辅热、电水壶、电饭锅、电热水器等设备进行了电热细分研究,通过设备运行关联算法实现了辅热设备的细分,并在用户有限反馈信息和专家标注的基础上,实现了非辅热设备分类的模型训练.实验结果表明,文中所提技术框架在事件检测负荷辨识算法的基础上实现了电热设备的细分,且在运行状态分解的F1分数指标中取得了 0.9以上的优异效果.

Abstract

In this paper,a non-intrusive load identification algorithm for residents based on prior knowledge and statistical learning model is proposed to solve the problem of insufficient electric heating subdivision capability in traditional identification technology.In this paper,the electric heating subdivision research is carried out for the auxiliary heating equipment of washing machine,electric kettle,electric rice cooker,electric water heater.The subdivision of auxiliary heating equipment is realized through the equipment operation association algorithm,and the model training of non-auxiliary heating equipment classification is realized based on the limited feedback information of users and expert annotation.The experimental results show that the technical framework proposed in this paper realizes the subdivision of electric heating equipment on the basis of the event detection load identification algorithm and F1 socre above 0.9 is achieved in the decomposition of operation state.

关键词

非侵入负荷监测(NILM)/事件检测/电热细分/统计分析/高斯混合聚类(GMM)/支持向量机(SVM)

Key words

non-intrusive load monitoring(NILM)/event detection/subdivision of electric heating equipment/statistical analysis/Gaussian mixture model(GMM)/support vector machine(SVM)

引用本文复制引用

基金项目

国网北京市电力公司科技项目(SGBJDKOOJLJS2250128)

出版年

2024
电力工程技术
江苏省电力公司 江苏省电机工程学会

电力工程技术

CSTPCDCSCD北大核心
影响因子:0.969
ISSN:2096-3203
参考文献量25
段落导航相关论文