Experimental Investigation and Comparative Analysis of an Efficient Machine Learning Algorithm for Distribution System Reconfiguration

Abstract

This study studies the implementation of machine learning (ML) algorithms to improve power distribution in an industrial context, concentrating on the essential issue of anticipating energy consumption. Various ML models, including Support Vector Machine (SVM), Artificial Neural Network (ANN), Decision Trees (DT), and Random Forests (RF), were extensively examined and compared for their usefulness in anticipating demand patterns within a sector encompassing machining, forging, CNC, and packaging stations. The models revealed various strengths, with SVM leading with an accuracy of 95.6%, closely followed by ANN at 94.33%, while DT and RF displayed accuracies of 87.6% and 85.6%, respectively. The research additionally gives a thorough comparison of actual vs expected demand levels over hourly intervals, illustrating the models' responsiveness to dynamic energy use patterns throughout the day.