Energy storage is of central importance in order to be able to reliably provide renewable energy, the availability of which is naturally subject to fluctuations. To this end, the question must be answered as to how these storage facilities can be used optimally. In energy systems, e.g. energy centers of urban districts, there are potentially several storage facilities of different size and use (short-term and long-term storage). This makes operational management a complex problem, as long-term considerations for charging and discharging long-term storage have to be reconciled with short-term needs at any given time.

The DISTEL - District Storage Intelligence project aims to develop algorithms that always operate such energy systems optimally, with maximum efficiency and minimum emissions. To this end, DISTEL will combine classical methods of optimization with advanced methods of artificial intelligence. For example, consumption and yield profiles over longer periods of time must be available for long-term planning, and it must be possible to model the behavior of the storage facilities in terms of energy losses in sufficient detail to be able to estimate what costs energy stored at the current time will save in the future. These long-term simulations in particular usually require a high degree of computing resources. This is where theory-driven machine learning methods come in handy, as they can approximately describe the behavior in much less time. Coupled with a model-predictive control system that takes this information into account, it should be possible to make the right decision at any point in time.