Decision-Centric Uncertainty Sets for Storage Arbitrage: Integrating Estimation and Optimization

Traditional uncertainty optimization methods employ a two-stage paradigm that separates uncertainty estimation from decision-making, often yielding suboptimal solutions. To address this issue, this letter proposes a decision-centric uncertainty set approach for risk-averse storage arbitrage, which integrates estimation and optimization through a tailored decision-centric gradient based on the implicit function theorem and a statistically valid size calibration method. The proposed framework achieves 35.1% higher profitability than robust optimization (RO) and 12.2% higher than distributionally robust optimization (DRO), while ensuring the desired risk coverage.