This work focuses on tests for control reserve of a novel Power-to-Gas-to-Power platform based on proton exchange membrane technologies and on pure oxygen instead of air in the re-electrification process. The technologies are intended as a further option to stabilize the power system, therefore, helping integrating renewable energy into the power system. The tests are based on the pre-qualification tests used by Swissgrid, but are not identical in order to capture the maximum dynamics by the plants. The main characteristics identified are the ramping capabilities of ±8% per unit per second for the electrolyzer system and ±33% per unit per second for the fuel cell system. The ramping capabilities are mainly limited by the underlying processes of polymer electrolyte membrane technologies. Additionally, the current and projected round-trip efficiencies for Power-to-Gas-to-Power of 39% in 2025 and 48% in 2040 are derived. Furthermore, during the successful tests, the usage of oxygen in the present Power-to-Gas and Gas-to-Power processes and its influence on the dynamics and the round-trip efficiency was assessed. In consequence, fundamental data on the efficiency and the dynamics of the Power-to-Gas-to-Power technologies is presented. This data can serve as basis for prospective assessments on the suitability of the technologies investigated for frequency control in power systems.

We present a method that operates an electrolyzer to meet the demand of a hydrogen refueling station in a cost-effective manner by solving a model-based optimal control problem. To formulate the underlying problem, we first conduct an experimental characterization of a Siemens SILYZER 100 polymer electrolyte membrane electrolyzer with \SI{100}{\kilo \watt} of rated power. We run experiments to determine the electrolyzer's conversion efficiency and thermal dynamics as well as the overload-limiting algorithm used in the electrolyzer. The resulting detailed nonlinear models are used to design a real-time optimal controller, which is then implemented on the actual system. Each minute, the controller solves a deterministic, receding-horizon problem which seeks to minimize the cost of satisfying a given hydrogen demand, while using a storage tank to take advantage of time-varying electricity prices and photovoltaic inflow. We illustrate in simulation the significant cost reduction achieved by our method compared to others in the literature, and then validate our method by demonstrating it in real-time operation on the actual system.