.png){kind=icon}
.png){kind=icon}
Return
Data-driven robust optimization for integrated refinery planning and blending scheduling under multi-mode uncertainty
.png){kind=icon}
PRE
AI
W
X
Y
X
M
DOI：10.1016/j.cjche.2026.02.007.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Integrating refinery planning and product blending scheduling under uncertainty is challenging due to multi-mode operating conditions and mismatched temporal scales. This work proposes a bilevel optimization framework that explicitly coordinates long-term refinery planning with short-term blending scheduling. To represent mode-dependent uncertainty in blending behavior, a data-driven robust optimization (DDRO) strategy is developed, in which mode-associated uncertainty sets are constructed from historical data using a multi-stage clustering (MSC) algorithm combined with weighted support vector machines (SVM). This enables uncertainty modeling that is adaptive to operating modes and less conservative than conventional holistic robust approaches. To solve the resulting bilevel robust model efficiently, a robust-buffer-based decomposition (RBD) algorithm is proposed, which iteratively introduces a specification buffer at the planning level to preserve scheduling feasibility under uncertainty. In the case study, the proposed multi-mode robust framework improves the net profit by approximately 1.0% compared to holistic robust model, while maintaining 100% feasibility in quality constraints and reducing computational time by more than 50% relative to the single-level formulation. These results demonstrate that the proposed framework effectively balances economic performance, robustness, and computational tractability in integrated refinery planning and scheduling.
Keywords:
refinery planning
blending scheduling
multi-mode uncertainty
data-driven robust optimization
bilevel optimization
Journal
IF:
3.7
Papers:
5.1K
Citations:
1.1W
Organization
Cited Papers