Betriebsführung von Batteriesystemen in Industriebetrieben mit Reinforcement Learning

Battery storage systems play a crucial role in the system integration of renewable energy. Applying various use cases can increase the profitability of a battery, but requires an intelligent energy management. Reinforcement Learning (RL) proves effective in solving complex problems without the necessity of manual modification of input parameters as the system changes. This paper investigates the economic potential of an RL algorithm as a battery control strategy for an industrial company for the multi-use case energy arbitrage and atypical grid usage. Applying it on the investigated data set, the RL controller manages to save between € 85,000 and € 98,000 per year, with capacity costs reduced by 40 to 46% compared to the reference case.