Classification of domestic electrical appliances based on starting transient using artificial intelligence methods

Abstract

With the rising implementation of Home Energy Management Systems (HEMS), active studies had been done relative to power monitoring alternatives. Load monitoring is an essential block of HEMS; therefore, the improvement of simplicity and convenience in load monitoring is crucial for the HEMS market expansion. This paper proposes the use of Artificial Neural Network models for the classification of common electrical appliances based on the extracted distinctive current starting transient features of electrical appliances. This research’s main challenges are: conducting reliable instrumentation practice with an appropriate choice of instruments, extracting distinctive features in the current transient, and analyzing the ANN classifier for good performance using artificial intelligence methods. The analysis would compare the performance of time-domain inputs and frequency-domain inputs to the ANN classifier. By selecting appropriate frequency-domain features as input to the ANN classifier, it was shown that up to 86% classification accuracy could be obtained using the proposed method, justifying our hypothesis that multiple non-intrusive load monitoring using a single sensor is indeed plausible.