Leveraging HPC for ML Surrogate Models and Data-Driven MPC in PEMFC Optimization

Short description of the business experiment

Genport, with the engineering support of POLIMI, proposes a business experiment to leverage High-Performance Computing (HPC) for developing a novel control method optimising PEMFC performance. Our project aims to develop and implement Machine Learning (ML) surrogate models, trained on high-fidelity 3D PEMFC models running on HPC resources, using the existing POLIMI PEMFC 3D model implemented with the open-source OpenFoam Computational Fluid Dynamics (CFD) software. These surrogate models will be integrated into a data-driven Model Predictive Control (MPC) strategy for real-time optimisation of fuel cell operation. Creating data-driven models solely based on experimental data would be prohibitively expensive and time-consuming compared to using surrogate models. Thus, the HPC-based approach will enable the efficient development of a highly competitive control product that increases PEMFC efficiency by up to 15% and extends durability by up to 50%.

Organisations involved: 

End User: GENPORT Srl

Technology expert: Politecnico di Milano - Department of Energy