A two-layer robust model predictive control-based energy management system for a zero-carbon DC microgrid with hybrid hydrogen-battery energy storage

Hybrid hydrogen-battery energy storage systems (HHBESS) integrated with renewable energy sources (RES) present a solution for zero-carbon microgrids. However, the real-time operation of HHBESS encounters a trade-off between equipment degradation, computational complexity, and conservativeness of uncertainty based on the control time scales. Conventional model predictive control (MPC) can address these issues; however, it typically neglects large RES forecast errors between the control time steps, potentially causing operational failures or accelerated degradation of the HHBESS. In this study, we propose an enhanced two-layer robust MPC-based energy management system for the real-time operation of zero-carbon DC microgrids. In the proposed system, the affinely adjustable robust optimization integrated into the MPC framework dynamically captures the stable degradation-minimizing power adjustment range of the HHBESS under real-time forecast errors between the control time steps. This information is shared between the two layers of MPC controllers. The upper layer ensures long-term stability, whereas the lower layer operates over shorter horizons with reduced cost conservatism. This hierarchical structure balances cost conservatism and computational complexity while ensuring robust operation. The proposed approach was validated using a benchmarked DC microgrid. Results showed that, compared to conventional MPC-EMS, the method reduced HHBESS degradation costs by over 50%, achieving a constraint violation rate of 0.02%. Compared to a single-layer structure using the same algorithm, it also reduced computational time by over 90%. These findings highlight the feasibility and efficiency of the proposed EMS for real-time operation in zero-carbon microgrids.