A Binary Weight-based Energy Disaggregation Framework for Residential Electricity Consumption

Abstract

Non-intrusive Load Monitoring (NILM) is a technique for monitoring the aggregate power consumption of a residential or commercial building from the main power supply. NILM enables users to gain insights into their power consumption patterns and can aid in optimizing energy usage. In this paper, a binary weight-based energy disaggregation framework for monitoring the electrical consumption of a household is proposed. This framework entails data preprocessing, followed by event detection via peak identification and k-means clustering. Subsequently, an energy disaggregation algorithm based on binary weights was developed. The algorithm is able to segregate total electrical consumption into device-level power consumption. The method was evaluated on data from house 1 in the Reference Energy Disaggregation Data set (REDD) sampled at a frequency of 1 Hz and 1/6 Hz. The proposed algorithm successfully separated power consumption into four device-level power consumptions with a precision of 99.5%, a recall of 89% and f1-score of 93%. This resulted in an improvement in overall performance with a 25% improvement in average precision, 35% in average recall, and 45% in the f1-score compared to an existing technique.