Fault Aware AI Based Load Scheduling for Industrial Facilities Using Machine Learning–Driven Demand Side Management

Industrial demand side management (DSM) has become a key enabler for reducing electricity costs, mitigating peak demand, and improving renewable energy utilization. However, most existing DSM frameworks assume ideal operating conditions and neglect the impact of electrical disturbances, such as voltage deviations and supply outages, which are common in real industrial power systems. This paper proposes a fault aware artificial intelligence–based load scheduling framework for industrial facilities that explicitly integrates electrical disturbance states into the DSM decision making process. A comprehensive industrial case study is developed using a realistic synthetic dataset comprising 61,488 hourly records for seven industrial loads, an 80-kW photovoltaic system, dynamic electricity pricing, and modeled disturbance conditions, including voltage drops and outages representing a 15% annual disturbance rate consistent with industrial reliability benchmarks. A Random Forest classification model is employed to predict feasible operating states under combined temporal, economic, solar, and disturbance conditions. The model outputs are incorporated within a constraint aware scheduling strategy that preserves production requirements while avoiding infeasible operation during fault conditions. Simulation results demonstrate that the proposed fault aware DSM framework maintains the economic benefits of intelligent scheduling, achieving 11.7% annual electricity cost reduction and more than 95% peak hour operation reduction, while significantly enhancing operational robustness under non ideal grid conditions. The findings confirm that integrating disturbance awareness into AI based DSM is essential for bridging the gap between theoretical optimization and practical industrial deployment.