Performance Analysis of Similar Appliances Identification using NILM Technique under Different Data Sampling Rates

Abstract

Similar power/similar loads identification using a non-intrusive load monitoring (NILM) system is a challenging task due to overlapping characteristics of similar appliances. Therefore, we need to identify robust features that are capable of distinguishing the events of similar appliances effectively. Further, the developed NILM system/methodology need to be examined for different sampling rates, since the performance evaluation of the algorithm with single sampling rate alone is not sufficient. In this work, we used four compact fluorescent lamps (CFL) which are having same specifications configured in different electrical network combinations. We explored locality constrained linear coding (LLC) based deep neural networks (DNN) to express the input feature vectors in terms of load independent basis vectors in a higher dimensional feature space to make the features robust for effective identification of similar loads in a NILM system. Further, we examined the effectiveness of LLC-DNN approach for the different sampling rates, 10 kHz, 25 kHz and 50 kHz respectively. From our study, it is observed that the performance of the LLC-DNN has been consistent and improves the similar appliances identification accuracy significantly with the increase in sampling rates.