Computationally efficient identification of databased models applied to a milk cooling system

Along with the progressing climate change, the use of renewable energy becomes more important. Farms have a high energy demand as well as space for the installation of renewable energy plants and hence there is a high potential for reducing the use of fossil energy sources by using self-produced renewable energy at availability. To best use that energy, smart energy management systems can reschedule tasks with high energy demand and charge or discharge storages. For such a system, models describing the behavior for all devices of a farm are required.

We designed a software module using a black box approach to identify general databased models at low computational cost. It can be used to calculate forecasts for arbitrary generator and storage devices. Using real world data we apply this to model a milk cooling system's temperature and generate additional inputs to improve the model. With these, the temperature of that system can be forecasted well and at low computational cost.