Di, 18. Mai. 2021   Lehrstuhl SO

Neue Publikation: Optimizing trading decisions of wind power plants with hybrid energy storage systems using backwards approximate dynamic programming

Benedikt Finnah und Jochen Gönsch entwickeln Künstliche Intelligenz (KI) für mehr erneuerbare Energie

Das Papier ist hier kostenlos verfügbar (bis 7. Juli): https://authors.elsevier.com/a/1d5No3ISCVChra


On most modern energy markets, electricity is traded in advance and a power producer has to commit to deliver a certain amount of electricity some time before the actual delivery. This is especially difficult for power producers with renewable energy sources that are stochastic (like wind and solar). Thus, short-term electricity storages like batteries are used to increase flexibility. By contrast, long-term storages allow to exploit price fluctuations over time, but have a comparably bad efficiency over short periods of time.

In this paper, we consider the decision problem of a power producer who sells electricity from wind turbines on the continuous intraday market and possesses two storage devices: a battery and a hydrogen based storage system. The problem is solved with a backwards approximate dynamic programming algorithm with optimal computing budget allocation. Numerical results show the algorithm's high solution quality. Furthermore, tests on real-world data demonstrate the value of using both storage types and investigate the effect of the storage parameters on profit.