Design and Experimental Investigation of Temperature Control for a 10 kW SOFC System Based on an Artificial Neuronal Network

Abstract

For a future carbon-neutral energy economy, fuel cells play an important role due their high efficiency. Especially the Solid Oxide Fuel Cell (SOFC) with demonstrated efficiencies beyond 60 % 1 can contribute to reasonable roundtrip efficiencies for hydrogen and e-fuels 2.To match the fluctuating electricity demand in a future electricity grid, dominated by renewable energy sources like wind and photovoltaic, dynamic operation of fuel cells is required. The research has shown that the degradation und therefore the lifespan of Solid Oxide Cell (SOC) stacks shows a significant dependency on the operating conditions and dynamic load changes 3. However, some research suggests that the degradation is not caused by the load changes itself but spatial temperature gradients during load changes 4–6.Therefore, controlling the temperature gradients in the stack during load changes can have a significant impact on the lifespan of SOC stacks. This is typically more dramatically for stacks with larger cell sizes or multiple cells per layer. Furthermore, a tight temperature control allows for running the stack at maximum efficiency without the risk of stack damage due to exceeding temperature limits.In this work the authors designed and experimentally evaluated different controller topologies for fuel cell operation (SOFC) of a reversible solid oxide cell (rSOC) system described previously 7. The controller design incorporates an artificial neuronal network (ANN) for real time state predictions. The training data for the ANN was generated by a dynamic model of this system. This model is implemented in Matlab Simulink and was validated against experimental data. The generated training data consists of about 1,000 simulated days of dynamic system operation with a sample interval of 10 s. Additionally, data for 16,000 different steady state operating conditions of the system were generated.One focus of this work is the robustness of the controller under real world conditions despite inaccuracies of the underlying model and SOC degradation effects over time. To compensate the aging of the stack, the ANN is trained on variable degradation states. The degradation state is then tracked by the controller during operation to maintain an accurate prediction. First system experiments showed promising results in this respect (Fig. 1).Fig 1. Maximum temperature (red) and its setpoint (red, dashed) as well as 8 other temperatures distributed over the stack (black) in response to a given current density profile (green) and the air flow (blue) set by the controller AcknowledgmentsThe authors would like to thank their colleagues at the Forschungszentrum Jülich GmbH, who helped realize this work, and the Helmholtz Society for financing these activities as part of the Living Lab Energy Campus. References(1) Peters, Ro.; Frank, M.; Tiedemann, W.; Hoven, I.; Deja, R.; Kruse, N.; Fang, Q.; Blum, L.; Peters, R. Long-Term Experience with a 5/15kW-Class Reversible Solid Oxide Cell System. J. Electrochem. Soc. 2021, 168 (1), 014508. https://doi.org/10.1149/1945-7111/abdc79.(2) Heydarzadeh, Z.; McVay, D.; Flores, R.; Thai, C.; Brouwer, J. Dynamic Modeling of California Grid-Scale Hydrogen Energy Storage. ECS Trans. 2018, 86 (13), 245–258. https://doi.org/10.1149/08613.0245ecst.(3) Kim, Y.-D.; Lee, J.-I.; Saqib, M.; Park, K.-Y.; Hong, J.; Yoon, K. J.; Lee, I.; Park, J.-Y. Degradation of Anode-Supported Solid Oxide Fuel Cells under Load Trip and Cycle Conditions and Their Degradation Prevention Operating Logic. J. Electrochem. Soc. 2018, 165 (9), F728–F735. https://doi.org/10.1149/2.1391809jes.(4) Hagen, A.; Høgh, J. V. T.; Barfod, R. Accelerated Testing of Solid Oxide Fuel Cell Stacks for Micro Combined Heat and Power Application. Journal of Power Sources 2015, 300, 223–228. https://doi.org/10.1016/j.jpowsour.2015.09.054.(5) Nakajo, A.; Wuillemin, Z.; Van herle, J.; Favrat, D. Simulation of Thermal Stresses in Anode-Supported Solid Oxide Fuel Cell Stacks. Part I: Probability of Failure of the Cells. Journal of Power Sources 2009, 193 (1), 203–215. https://doi.org/10.1016/j.jpowsour.2008.12.050.(6) Jiang, W.; Luo, Y.; Zhang, W.; Woo, W.; Tu, S. T. Effect of Temperature Fluctuation on Creep and Failure Probability for Planar Solid Oxide Fuel Cell. Journal of Fuel Cell Science and Technology 2015, 12 (5), 051004. https://doi.org/10.1115/1.4031697.(7) Peters, R.; Tiedemann, W.; Hoven, I.; Deja, R.; Kruse, N.; Fang, Q.; Blum, L.; Peters, R. Development of a 10/40kW-Class Reversible Solid Oxide Cell System at Forschungszentrum Jülich. ECS Trans. 2021, 103 (1), 289–297. https://doi.org/10.1149/10301.0289ecst. Figure 1