AI-driven smart grid optimization for hospital energy systems integrating renewable generation, predictive maintenance, and resilient infrastructure.

Efficient and resilient electrical systems are vital for hospital operations, where uninterrupted power is critical for patient safety and clinical continuity. This study develops an AI-driven smart grid optimization framework for a tertiary hospital in Kuala Lumpur, Malaysia, integrating renewable generation, load forecasting, predictive maintenance, and HVAC energy optimization. A detailed survey of the hospital’s electrical infrastructure including 158,305 m2 of built-up area, 1,500 beds, and over 200 electrical appliances—was used to construct an appliance-level load model capturing both deterministic and stochastic energy behavior. Advanced Long Short-Term Memory (LSTM) forecasting and Reinforcement Learning (RL) algorithms were employed to predict dynamic load fluctuations, optimize renewable dispatch, and manage hospital load uncertainty arising from variable occupancy and equipment usage. The proposed HVAC energy efficiency strategy, incorporating adaptive set-point control, occupancy-based variable-air-volume scheduling, and renewable-aligned operation, achieved an 11.6% reduction in HVAC energy consumption. Simulation results show an optimized daily demand of 91,080 kWh, with renewable sources supplying 86% from rooftop solar PV, 1.2% from wind, and 0.2% from battery storage, reducing grid dependence to 12.6%. Overall, the AI-driven system improved total energy efficiency by 25%, reduced unplanned downtime by 30%, and enhanced resilience for critical hospital zones such as ICUs and operating theatres. The findings demonstrate a scalable pathway toward sustainable, data-driven, and self-adaptive hospital energy systems aligned with Malaysia’s National Energy Transition Roadmap (NETR).