Machine Learning-Driven Load Prioritization for Reliable Residential Smart Grids

Ensuring the reliability of smart grid systems during abnormal conditions is critical, particularly in residential areas where power outages can have severe consequences. This study focuses on improving the resilience of a small smart grid system, integrating a traditional utility grid, solar photovoltaic (PV) generation, and wind energy, in a residential area of North Carolina. The proposed approach utilizes Machine Learning (ML) models to classify electrical loads based on priority, aiming to disconnect only nonessential loads during electrical faults rather than entire subzones. Five ML models were evaluated based on load behavior and frequency of usage. Decision tree model achieved the highest accuracy, with a performance rate of 100%. This selective disconnection strategy enhances grid stability while ensuring that essential loads, particularly those supporting individuals with medical dependencies, remain powered. The analysis was carried out by PSS®E software. The findings underscore the potential of intelligent load management in modern smart grids, offering a robust solution for maintaining critical services during power disruptions.