Maximizing Efficiency in Commercial Power Systems with an Optimized Load Classification and Identification Method Using Deep Learning and Ensemble Techniques

Abstract

Due to the continuous rise of energy demand and electricity costs, the need for a detailed metering option has become crucial. Non-intrusive load monitoring is such an approach that requires less hardware compared to the other load monitoring options, significantly improving consumer comfort. Due to this reason, researchers are encouraged to implement more advanced machine learning techniques capable of accurate load classification and identification; among them, most focus on residential applications due to fewer complications. However, commercial power systems present considerable challenges compared to residential power systems due to the greater diversity of loads and significant imbalances. In order to overcome these challenges, we introduce a novel neural network design that incorporates sequence-to-sequence, WaveNet, and Ensembling techniques to identify and classify single-phase and three-phase loads in commercial power systems. We tested our approach by identifying and classifying nine appliances - five single-phase and four three-phase - for three months, revealing a significant improvement in accuracy.