Adaptive MPC lowers PEM electrolyzer temperature overshoot by 4.2 °C versus PID in simulation

A research team has proposed an online method for regulating the temperature of proton exchange membrane water electrolyzers (PEMWE) powered by fluctuating wind and solar generation. According to the journal Carbon Neutrality, the method combines an electro-thermal-mass performance prediction model with an adaptive model predictive control (AMPC) algorithm. The model was validated in experiments on a 100 kW PEM electrolyzer and predicted temperature with an accuracy of 5.3% springer.comCarbon Neutralityspringer.comOpen source ↗.

The control comparison with the conventional approach was run under simulated fluctuating wind-solar power input, not in a field installation. The authors report that AMPC shortened the temperature rise response time by 0.55 h relative to PID control.

Why PID control falls short

The widely used proportional-integral-derivative (PID) method depends heavily on manual tuning experience. The authors state that it cannot precisely regulate the strongly nonlinear and tightly coupled electrolysis process, which brings risks of temperature overshoot and response lag.

With direct coupling to wind and solar, the current through the stack changes quickly. If heat is not removed in time, local hot spots can damage the membrane electrode assemblies, while slow thermal response limits reaction rates in cold conditions.

Model, control variable and results

The prediction model accounts for electric conduction, heat transfer and multi-phase mass transport. The control algorithm supports real-time updating of the electrolyzer operating states springer.comCarbon Neutralityspringer.comOpen source ↗. The experiments were carried out at the Daxing Hydrogen Energy Park in Beijing on an industrial stack from CSSC Peric Hydrogen Technology.

The researchers also found that the temperature of the circulating cooling water varies more mildly than the water flow rate. For this reason it is better suited as the control variable for temperature regulation under dynamic operating conditions.

Under the simulated fluctuating wind-solar input, the AMPC method brought the electrolyzer temperature to the target value within 10.7 min. Its temperature overshoot was 4.2 °C lower than that of PID control.

The authors say the work is intended to provide theoretical references and technical tools for applying PEM water electrolysis safely and reliably in future large-scale hydrogen production systems coupled directly to wind and solar power springer.comCarbon Neutralityspringer.comOpen source ↗.