CC BY-NC 4.0Groß, ArneArneGroßKreilgaard, LarsLarsKreilgaardKöpfer, BenediktBenediktKöpferKühnbach, MatthiasMatthiasKühnbach2023-06-152024-05-272023-06-152023Note-ID: 0000796Ehttps://publica.fraunhofer.de/handle/publica/442882https://doi.org/10.24406/h-44288210.2991/978-94-6463-156-2_1410.24406/h-442882With decreasing prices of batteries, business cases of electric storage systems reach profitability in commercial applications. However, for commercial storage systems to become economically attractive, many studies indicate that different use cases both behind and in front of the meter must be addressed simultaneously. Shaving load peaks to reduce grid surcharges is considered in most presented setups. However, the success of weighing peak-shaving with other use cases is highly dependent on a precise load forecast. In the literature, perfect foresight of the future load profile is assumed for most multi-use applications. In a real-life setup, forecasts have sizable uncertainty specifically for load peaks. Consequently, the performance of a storage system in a realistic setting could differ vastly. Addressing this topic, this article presents an Energy Management System (EMS) for a battery storage combining peak-shaving with other use cases. The EMS relies on machine learning techniques tailored specifically to forecast load peaks and a heuristic control scheme using these forecasts in operation. Furthermore, we compare the performance using our load forecast algorithm to the operation with perfect foresight.enPV Battery SystemsForecast of Electrical LoadPeak ShavingMulti-Useconference paper