Battery Storage Co-Optimization in Day-Ahead and Real-Time Markets with Bayesian Optimization
创建于 更新于
摘要
ARBO-DART jointly optimizes day-ahead (DA) commitments and real-time (RT) recourse for standalone battery energy storage (BESS) by wrapping Bayesian optimization around a closed-loop RT solver with exogenous DA prices and a mean-reverting RT price process. It uses a blockwise, adaptive refinement of the DA profile guided by RT feedback, achieving substantial speedups (2–4x) over full-dimension BO while recovering economically meaningful DA schedules that leverage intraday price volatility [page::2][page::3][page::4]. ARBO-DART demonstrates robustness across price-curve shapes and DART spread structures and yields interpretable, parsimonious DA profiles with limited nonzero blocks that efficiently coordinate DA and RT actions [page::2][page::3][page::4].
