Towards the adoption of neural-network-based solvers for coupled mh-pcm hydrogen storage devices

In the present work, recent advancements are shown in the development of neural network-based (NN) numerical solvers for the simulation of the coupled chemistry and physics typical of intermetallic Metal Hydride – Phase Change Material (IMH-PCM) hydrogen storage tanks. The final aim of this research activity is to obtain fast and reliable numerical tools, which should be able to span a large number of different IMH-PCM configurations and thus effectively support the optimal design of the tank in terms of energy density and specific power delivery. The results shown here include: i) a NN training phase, based on a subset of CFD-generated data, and ii) initial tests of the NN-based solver which is applied out of the training set points to estimate the MH-PCM performance in terms of heat fluxes and hydrogen mass flow rate. Our findings show great potential in NN-based approaches to reduce simulation turnaround times for this class of complex hydrogen storage systems, thus paving the way for their efficient integration in real-time modelled hydrogen energy systems and their adoption as design tools for IMH-PCM devices.