Day-ahead electricity market bidding for hybrid energy storage aggregators involves navigating multiple simultaneous price uncertainties: locational marginal prices, regulation market clearing capacity prices, and regulation market performance prices. Traditional seasonal autoregressive integrated moving average (SARIMA) forecasting treats each price and hour independently, missing critical correlations that occur across a full 24-hour operating day.
Researchers have introduced Day-SARIMA, an enhanced approach that combines QUBO (quadratic unconstrained binary optimization) with SARIMA modeling to capture complete-day price relationships. Rather than calibrating forecast errors separately for each hour and price, the method represents each historical day as a 24-hour joint residual matrix. QUBO then intelligently selects which past days are most relevant to the target forecast day, considering both relevance and the price co-movement patterns while penalizing redundant residual signatures.
The selected subset of historical days defines price- and hour-specific deviation ranges and adaptive uncertainty budgets that feed into a two-stage robust bidding optimization model. This approach preserves cross-price correlations implicitly rather than imposing them as explicit constraints, allowing the algorithm to learn natural market dynamics.
Testing on PJM Interconnection data spanning October 2024 through February 2025 yielded 56 matched days for comparison. Day-SARIMA reproduced price correlations more accurately than alternative selection methods such as fast forward selection or K-medoids clustering. Compared to conventional SARIMA, Day-SARIMA increased mean net revenue by $53.21 per day—a gain supported by paired statistical testing. The method also reduced modeled cycling-degradation costs by 16% and supercapacitor equivalent-full-cycle costs by 20.5%.
Interestingly, a block-based variant using approximately 13 times more QUBO variables than Day-SARIMA produced inferior scenario similarity and coherence, suggesting that complete-day information capture is more valuable than raw variable count. The results demonstrate that quantum-inspired optimization can meaningfully improve energy storage aggregate profitability and asset longevity in competitive power markets.
