Reliability and Comfort-Aware Operation Optimization for Hydrogen-Based Building Energy Systems in Off-Grid Mode

Hydrogen-based energy systems have gained wide attention due to their significant potential for relieving energy crises and environmental problems. However, existing studies neglect occupant comfort control in hydrogen-based building energy systems (HBESs). In this paper, we investigate a reliability and comfort-aware operation optimization problem of an HBES in off-grid mode. Achieving the above aim encounters several challenges due to multi-source uncertainties, implicit indoor environmental dynamics, temporally and spatially coupled constraints, and nonlinear constraints. To address the above challenges, we propose a physics-embedded neural network (PENN)-assisted hierarchical model predictive control (MPC) algorithm. Specifically, the PENN architecture is adopted to capture the indoor dynamics accurately. Then, the PENN-assisted upper-level MPC implements occupant comfort control and optimizes the multi-energy demands. Next, the lower-level MPC optimizes the system operation cost and reliability with known multi-energy demands decided by the upper-level MPC. Simulation results show that the proposed algorithm outperforms baselines in terms of operation cost, system reliability, and indoor occupant comfort.