A new event detection technique for residential load monitoring

Abstract

In the process of non-intrusive load monitoring, the first step is to find the switching on/off events as accurate as possible. However, there are extensive disturbances in the recorded data, due to the voltage fluctuations, electrical noise and etc, resulting in either false detection or event missing. To solve this problem, a new event detection technique is proposed in this paper. This technique uses median filter algorithm and the proposed ripple mitigation algorithm to effectively remove the unexpected disturbances and extract the real signal of switching on/off events. The proposed algorithms are very easy to implement in the low cost devices. The lab data and BLUED data tests show more than 90% accurate rate of event detection.