Applying Machine Learning Techniques to Predict Optimal Capacitor Locations in Distribution Power Grids

Abstract This research aims at investigating how Machine Learning (ML) can help decision makers in solving the problem of optimal capacitor allocation for transmission loss minimization in power systems. ML is a branch of Artificial Intelligence (AI) which can predict and draw inference from recognized patterns in data. The data used in this study, to train the ML models, inherent the structure of power grids. Four regressive predictive models are used and compared. The suggested investigation assesses the prediction performance compared to the actual values through comparing each predicted busbar rank to its actual rank. Busbar rank is defined based on the power loss reduction achieved by capacitor connection to this busbar. The busbar giving maximum loss reduction is ranked as 1. By ranking the busbars, the control system of the grid can select the most appropriate group of buses whose capacitors should be switched on for a certain loading condition. The results show that ML models can reach a ranking accuracy that exceeds 80%. Hence, it is shown through this research that ML techniques can help in reducing the search space of the optimal capacitor allocation problem.