An End-to-end Prediction Method for Energy Storage System Arbitrage with Battery Degradation Costs

Energy storage system (ESS) can achieve arbitrage by charging at low electricity prices and discharging at high prices, thus electricity price prediction is important in ESS arbitrage optimization problem. Traditional methods follow "Predict then Optimize" framework, which divides prediction and optimization into two independent processes and focuses on prediction accuracy. However, these methods ignore the inconsistency between prediction error and decision error, and a lower average prediction error may not necessarily generate a better decision. In addition, the battery will suffer degradation during ESS operation, which needs to be considered in arbitrage process. To achieve better modeling of ESS arbitrage and further improve arbitrage benefits, this paper realizes an end-to-end closed-loop method for ESS arbitrage considering battery degradation costs. The electricity price model and ESS arbitrage optimization model are built and simplified to fit into the end-to-end prediction framework. Then the prediction and optimization stages are combined to form a complete closed-loop algorithm. Finally, the proposed method is evaluated by an experiment on a real electricity price dataset, and the results show an enhancement in ESS arbitrage benefit.