Energy Disaggregation of Residential House via Event Based Optimization Technique

Abstract

Non-intrusive load monitoring (NILM) is a technique for disaggregating the total energy consumption of a building into individual appliance-level energy consumption. Event detection is a critical component of NILM systems as it involves the identification and classification of different electrical events from the aggregate power signal. In this article an event detection method for NILM systems has been proposed that is based on the analysis of the statistical properties of the aggregate power signal. Specifically, we use a sliding window approach and K-Means clustering to detect number of devices from the power signal and then apply a threshold-based algorithm to detect electrical events. We evaluate the proposed method on a public dataset and demonstrate its effectiveness in accurately detecting electrical events. The proposed method has the potential to improve the accuracy with recall of 98.84% carried out on Pecan Street Datanort Inc.