Cost effective disaggregation mechanism for the NIALM system

Abstract

Non-intrusive appliance load monitoring (NIALM) technology enables to detect the amount of power usage of each home appliance with single point sensing. In order to improve the performance (i.e., disaggregation/classification accuracy), features such as real power and harmonic components of electrical current of appliances are considered. However, taking more features into an account for better performance also increases the total cost of NIALM-related system. In this paper, a cost effective approach for NIALM is presented. With measured real power data sampled at low speed (about 1~2 Hz) from a single power line, classification of home appliances with high accuracy (about 80%) is achievable based on the proposed NIALM mechanism. The proposed mechanism is based on the real power time window pattern during a transient state and the real power level in a steady state after transition. The performance of the proposed mechanism is evaluated through simulation; classification accuracy is improved by 28% compared to the conventional NIALM approach based on real power delta scheme and 17% compared to the time window pattern.