Support Vector Machine-Assisted Improvement Residential Load Disaggregation

Considering the importance of energy and the necessity of its management, this paper examines residential energy disaggregation/non-intrusive load monitoring. Support Vector Machine (SVM) has been proposed as one of the most powerful machine learning applications to solve this problem. The advantage of this method over other methods is the feature extraction of data and their classification based on recognized patterns. The proposed method is conducted on two REDD and AMPds datasets, which are related to real-world measurements. In the proposed method, SVM is trained by each of the characteristics of a particular electrical appliance. Then, the trained network shows the closest recognition to identify the given electrical appliance and predict the total power consumption of homes. The accuracy obtained for the datasets shows the applicability of the proposed method for load disaggregation.