Experimental demonstration of dynamic demand response scheduling for PEM-electrolyzers

The use of renewable energy sources, such as wind power and photovoltaics is expected to produce fluctuating electricity prices. These fluctuations give PEM electrolyzers the opportunity to reduce costs, as they can adapt their production rates rapidly. Moreover, typically slow temperature dynamics of electrolyzers increase their flexibility for effective operational management strategies. With a defined temperature trajectory during scheduling optimization, overload operation of the electrolyzer for a given amount of time is possible. However, the temperature dynamics are typically nonlinear. In conjunction with discrete on/off decisions, temperature dynamics lead to mixed-integer nonlinear optimization problems for scheduling that are highly challenging to solve in real time. In this study, we experimentally validate the dynamic ramping scheduling optimization method that precisely linearizes nonlinear temperature dynamics using a flatness-based coordinate transformation. Utilizing the available information from the dynamic scheduling optimization a 100 kW PEM electrolyzer was operated by studying three stack temperature control methods, rejecting disturbances from load variations. Identifying a suitable control method was essential to guarantee the desired temperature tracking performance of the optimization. Our experiments show a 3.8 % cost reduction compared to the benchmark without overload operation. The designed PEM electrolyzer model also deviated only 0.6 % in costs from the experiment. Simulative scaling of PEM electrolysis to 2 MW demonstrates even higher cost reductions with the dynamic ramping method, as the larger electrolyzer has slower dynamics.