„Aufbau eines Batteriemanagementsystems auf Basis selbstlernender neuronaler Netze„
„Battery-storage-control based on self learning neural networks “
The composition of the german electricity production is changing significantly. The goal is protect the environment by replacing old climate-damaging power plants with renewable energy sources. The decentralized structure and volatile power feed of renewable energy sources causes an increased need of storage systems and power-flexibility. The integration of large battery-storage-systems can be helpful to balance the energy flow and support a stable operating electric power system. Therefore a battery-storage-system was implemented into an household. Including a photovoltaic system, the household is able to reduce its power-feed of the electric grid as well as to improve power system stability by offering primary control to the grid.
The controllable power flows are strongly dependent on time, weather and frequency and are tough to manage with an rule-based control system. Machine Learning with Reinforcement Learning algorithms are offering new interesting opportunities to improve this kind of power flows in such an uncertain environment. The advantage of control systems based on Machine Learning is the independent learning of complex patterns and rules, by just receiving simple rewards. Later on those power flows can be compared and analyzed to data of a similar project called SWARM.
Ort: Raum 04.137, Martensstr. 3, Erlangen