基于时域分析特征的非侵入式负荷事件检测方法  

A non-intrusive load event detection method based on time-domain analysis features

作　　者：赵学明 王宏伟 吴宇彤 张军[1] 史成豪 米东风 ZHAO Xueming;WANG Hongwei;WU Yutong;ZHANG Jun;SHI Chenghao;MI Dongfeng(State Grid Tianjin Electric Power Company,Tianjin 300171,China;School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China)

机构地区：[1]国网天津市电力公司,天津300171 [2]天津大学电气自动化与信息工程学院,天津300072

出　　处：《浙江电力》2025年第3期62-70,共9页Zhejiang Electric Power

基　　金：国网天津市电力公司科技项目(城东-研发2023-03)。

摘　　要：准确检测事件的完整暂态过程可提升NILM(非侵入式负荷监测)的最终效果,然而现有的事件检测方法难以在多场景下维持其优良性能。为此,提出基于时域分析特征的非侵入式负荷事件检测方法。首先,对各功率采样点计算自适应阈值,并据此初步定位事件位置;随后,多次移动稳态搜索滑窗,定位事件前后的稳态时段以确定事件所处的时段,并通过变点检测确定事件的启止时间;然后,设计了后处理步骤,根据波形特征识别并移除波动事件;最后,在私有和公开的BLUED数据集上,将所提方法同其他3种检测方法进行对比实验,实验结果验证了所提方法的有效性。Accurately detecting the complete transient process of an event can improve the performance of nonintrusive load monitoring(NILM).However,existing event detection methods struggle to maintain high perfor⁃mance across multiple scenarios.To address this issue,a non-intrusive load event detection method based on timedomain analysis features is proposed.First,adaptive thresholds are calculated for each power sampling point,which are then used to preliminarily locate event positions.Next,a sliding window search method is employed to repeat⁃edly move through steady-state periods before and after the event,determining the time span of the event.The start and end times of the event are identified through change point detection(CPD).Subsequently,a post-processing step is designed to recognize and remove fluctuation events based on waveform characteristics.Finally,comparative experiments with three other detection methods are conducted using both private and publicly available BLUED da⁃tasets.The experimental results validate the effectiveness of the proposed method.

关 键 词：有功功率 非侵入式负荷监测 负荷事件检测 状态检测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]