Degradation-Aware Model Predictive Dispatch of Modular Pem Electrolyzers Under Fluctuating Wind Power

Abstract

The high-load transients, inefficient partial-load operation, thermal cycling, and start-stop stresses in protonexchange membrane (PEM) electrolyzers are caused by rapid wind-power fluctuations. In this paper, degradation-sensitive model predictive dispatch scheduling of a simulated 1-MW plant (four 250-kW PEM stacks) has been described. A supervisory layer decides possible stack commitments and a continuous quadratic MPC allocates stack power and cooling. of stack power and cooling on a 20 -min prediction horizon. The goal is a combination of operating-time, temperature, ramping, start-up and low-load stress with explicit stack-health-balancing. Comparisons in five 24-h wind scenarios, and 30 realizations of stochastic forecast errors indicate that compared to conventional MPC, the proposed controller causes a 19% reduction in the normalized stress index, a 57% reduction in final stack-health dispersion, a reduction in curtailment of 5.1 to 4.3, and a greater simulated hydrogen production of 427 to 429 kg day ${}^{-1}$. The sensitivity analysis determines the trade-off between operation stress and hydrogen production and proves that gains reported can be determined by the choice of coefficient and MPC weights. These numbers are not experimentally calibrated lifetime predictions, but comparatively simulated simulation results; physical-stack, hardware-in-the-loop, and long-duration verification is needed before lifetime or bankability claims can be made.