DERIEL (De-Risking Electrolysis) forms the link between the two aspects of series production of water electrolyzers and the application of series-produced water electrolyzers in the industrial environment. A prerequisite for reliable series production and a low-friction market ramp-up is a fundamental understanding of the degradation, failure and interface mechanisms at all technical and economic levels. In DERIEL, the technical level is fundamentally investigated in relation to the operation of the electrolyzer up to the single module size on a scale of 0.75 - 1 MW. All components of the electrolyzer are considered, starting with the electrocatalysts, membrane electrode assembly (MEA), cell, cell stack and electrolyzer-specific process technology. A complementary project to exploit the results along an entire value chain (electricity to kerosene, total electrolyzer size 8 MW) is in planning (DERIVA - De-Risking Value Chain). The scaling of the electrolyzer into the application is done exclusively by parallelization of single modules with adapted overall process technology (8 units). This type of qualification is generally referred to in the industrial sector as de-risking. In automotive engineering, such pre-series models are called Erlkönige (prototypes), which are characterized above all by comprehensive sensor equipment.
Together with its partners, OFFIS is developing a digital twin that combines data-driven models with physicochemical models to best represent the behavior of the DERIEL test facility. The functions of the twin include process analysis/diagnosis and aging modeling. OFFIS focuses on the following topics. First, on a comparison of architectural concepts for the construction of the twin (especially the coupling of the different model types). Second, on the data-driven development of submodels for the twin (e.g. with respect to aging). Third, on concepts for real-time model selection between data-driven and physico-chemical modeling. Fourth, on the comparison of conventional approaches such as parameter learning for the physico-chemical model with approaches from the field of machine learning, and, fifth, on simulations to identify relevant parameter sets in distinction of initial/design phase, production and operation.