Research on Dynamic Response and Predictive Smoothing Control of AEM Water Electrolyzers Considering Renewable Energy Fluctuations

Abstract

To address current-density ramping, cell-voltage increase, and elevated operating stress in anion exchange membrane water electrolyzers (AEMWEs) under fluctuating wind and photovoltaic (PV) power inputs, this study develops a dynamic AEMWE model that couples voltage losses, thermal dynamics, and water-management states. A forecast-assisted reference governor based on short-term power prediction and dynamic constraints is further proposed. Using German Open Power System Data (OPSD) wind and PV, the effects of power-command correction on current density, voltage efficiency, specific energy consumption (SEC), voltage-limit exceedance, and a degradation-related stress proxy are analyzed. The steady-state benchmark distinguishes a separate 21-point empirical polarization fit (same-set RMSE 0.0059 V and MAPE 0.292%) from the mechanistic voltage-loss model actually used in the dynamic simulations (same-set RMSE 0.5063 V and MAPE 28.42%); consequently, the dynamic results are treated as model-conditioned rather than experimentally validated. In the 24 h case, explicit state-constrained smoothing reduced the maximum cell voltage from 2.2479 to 2.1501 V and SEC from 59.4755 to 58.1968 kWh kg−1, but hydrogen yield and renewable-energy utilization decreased from 1.51997 to 0.54293 kg d−1 and from 99.52% to 34.78%, respectively. The forecast envelope did not bind on the selected smooth day; across 30 representative days it reduced the stress proxy relative to the no-look-ahead state-constrained baseline on only 3–6 days depending on wind–PV composition, while in a diagnostic ramp-down case it activated 10 times and reduced cumulative ramping by 6.7% at the cost of a 16.9% hydrogen-yield loss. These results provide model-based exploratory evidence and separate the benefit of state projection from the incremental value of prediction.