基于二分递推SVD分解的居民电器开关事件检测方法  

Method of Switching Events Detection of Residential Appliances Based on Binary Recursive SVD Decomposition

作　　者：林顺富 于俊苏 李东东 符杨 LIN Shunfu;YU Junsu;LI Dongdong;FU Yang(College of Electrical Engineering,Shanghai University of Electric Power,Yangpu District,Shanghai 200090,China)

机构地区：[1]上海电力大学电气工程学院,上海市杨浦区200090

出　　处：《电网技术》2020年第4期1534-1542,共9页Power System Technology

基　　金：上海市科委科技项目(16020501000);上海市曙光计划项目(15SG50);上海市人才发展资金资助项目(2018004)。

摘　　要：针对非侵入式负荷监测技术中居民电器开关事件检测准确率不高的问题,提出一种基于二分递推奇异值分解(singular value decomposition,SVD)方法。对电器总有功功率数据进行二分递推SVD分解,通过计算奇异熵确定最佳分解层数,以不同层次空间的细节信号体现原始信号的突变点特征;采用硬阈值滤波函数对细节信号进行滤波,实现总有功功率中突变点位置的准确检测。利用不同噪声条件下实验数据和REDD(reference energy disaggregation data)数据集进行验证,所提方法克服了小波分解方法的检测位置偏移和非参数化双边滑动CUSUM变点检测方法的漏检、误检问题,有效提高了居民电器开关事件检测的准确率。Aiming at the problem of low accuracy of switching event detection of residential appliances in non-intrusive load monitoring technology(NILM), a method based on bisection recursive singular value decomposition(SVD) is proposed for switching event detection. It performs bipartite recursive SVD on the total active power data of residential appliances to obtain the detailed components of each decomposition layer. The optimal decomposition layer is determined by calculating singular entropy. Hard threshold filtering is adopted to denoise the detailed components. Based on experimental data and REDD database, the proposed method is verified and compared with wavelet decomposition method and non-parametric bilateral sliding CUSUM method. Results show that the proposed method improves the detection accuracy of switching events of residential appliances.

关 键 词：非侵入式负荷监测 居民电器 突变点检测 SVD分解 奇异熵 

分 类 号：TM721[电气工程—电力系统及自动化]