Neural network based modeling of metal-hydride bed storages for small self-sustaining energy supply systems

The electrical supply of remotely located objects - such as telecommunication relay stations, alpine huts, or farms and small settlements in developing countries - requires autonomously operated micro-grids, favorably based on renewable energy sources. For decoupling of fluctuating renewables based generation and consumption, energy storage is needed. Hydrogen paths, especially those based on metal-hydride beds, have proven well as long term energy storage in particular for such small supply systems, combining the advantages of low operating pressures and no storage losses. The performance of such metal-hydride bed storages plays an important role in designing and operating the complete energy system; rather, physical modeling of them for simulative system studies is awkward in consequence of the high grade of non-linearity and the multitude of internal parameters to be considered, which are partly unknown. Therefore, neural network based modeling of metal-hydride bed storages was successfully developed and verified, which is described in the present paper.