A NEW TECHNIQUE COMBINING SEMI-SUPERVISED AND ACTIVE LEARNING FOR NON-INTRUSIVE LOAD MONITORING

Abstract

The current work introduces a new technique that leverages both the semi-supervised and active learning together to the benefit of non-intrusive load monitoring (NILM), which is the procedure used to disaggregate the contributions of different appliances in a building. The main idea is that semi-supervised learning improves the results of active learning aiming to decrease the need to the user. Two different approaches were utilized, one used active and reactive power features and the other used current waveform harmonics to use them later in the machine learning model.