(373am) A Predictive-Reactive Framework for Large-Scale Dynamic Crude Oil Scheduling

Production scheduling optimization is pivotal in enhancing refineries' economic performance and competitive edge [1]. Among the primary components of refinery scheduling, crude oil blend scheduling assumes a significant role as it directly influences the stability of subsequent production processes [2]. However, the expansion of modern refineries has led to the emergence of large-scale dynamic crude oil scheduling problems (COSPs), characterized by thousands of discrete and continuous variables, nonlinear constraints, and various unexpected events such as delays in vessel arrivals, unavailability of tanks, and demand fluctuations. It is nontrivial to make feasible crude oil scheduling decisions and effectively cope with unforeseen events within a limited time [3,4]. Therefore, large-scale dynamic COSPs have attracted extensive attention from both academia and industry.

Over the past decades, mathematical programming (MP) has been widely applied to tackle COSPs [5-7]. Notably, the presence of bilinear terms in blending constraints leads to the formulation of complex mixed-integer nonlinear programming models, which are challenging to solve by commercial solvers [8]. To overcome this challenge, several solving strategies have been proposed, such as rolling horizon [9,10] and relaxation-based iterative decomposition [11]. However, these methods are only effective for small- or medium-scale scenarios with few types of crude oil and cannot thoroughly solve the composition concentration discrepancy [12]. As scheduling scales increase, the expansion of discrete variables and nonlinear terms leads to exponential growth in computational time [13].

Additionally, in light of various uncertainties inherent in scheduling, the existing literature mainly presents two methodologies: proactive [4,14] and reactive [15,16] scheduling approaches. The former attempts to incorporate future uncertainties, but even with the most robust proactive scheduling, it is impossible to completely overcome all uncertainty that may arise in the future. The latter approach involves modifying the original scheduling policy or devising a new one after realization of the uncertainty, which is suitable when proactive scheduling lacks sufficient information to describe the uncertain parameters accurately. Nevertheless, reactive scheduling optimization methods still face significant challenges regarding solution quality and responsiveness, especially when grappling with large-scale problems. Moreover, existing reactive methods have yet to strike the trade-off between optimality and stability in rescheduling [17].

In this work, we investigate various common unexpected events (i.e., vessel arrival delays, demand fluctuations, residue consumption fluctuations, and tank malfunctions) emerging in the large-scale crude oil blend scheduling process. We propose a knowledge transfer-based dynamic scheduling evolutionary algorithm (KT-DSEA) to effectively address large-scale dynamic COSPs within a predictive-reactive framework. This study improves the responsiveness and rescheduling quality when tackling large-scale COSPs in dynamic environments, achieving a balance between optimality and stability.

The novelty of the proposed KT-DSEA involves three main aspects. First, multiple subpopulations are generated according to diverse reactive scheduling starting points spanning from the detection of uncertainty to the appearance of infeasibility. Herein, individuals will have different optimization scopes and initial conditions due to the different rescheduling points. The cooperation of subpopulations aims to provide a more adaptable way to balance the optimality and stability of results. Furthermore, knowledge transfer techniques are integrated into the initialization phase of the subpopulations. We devise two knowledge transfer mechanisms for dynamic crude oil scheduling problems. One mechanism is knowledge transfer based on a repair strategy to assist the transition of solutions from a static environment (source domain) to a dynamic environment (target domain). The other extracts knowledge from a static scheduling environment to effectively enhance the responsive speed of the algorithm in a dynamic environment. In order to improve the convergence of the algorithm, MP is utilized to optimize continuous variables throughout the evolutionary process. The hybridization of MP and evolutionary algorithms promotes the efficient computing performance of KT-DSEA.

Keywords: Large-scale optimization, evolutionary algorithm, dynamic scheduling, crude oil scheduling

Acknowledgments

Wanting Zhang acknowledges the financial support from the China Scholarship Council Program under Grant 202306740022. Pedro Castro acknowledges support from Fundação para a Ciência e Tecnologia (FCT) through project UIDB/04028/2020.

References

[1] Shah NK, Li Z, Ierapetritou MG. Petroleum Refining Operations: Key Issues, Advances, and Opportunities. Industrial & Engineering Chemistry Research 2011;50(3):1161–1170.

[2] Xu J, Zhang S, Zhang J, Wang S, Xu Q. Simultaneous scheduling of front-end crude transfer and refinery processing. Computers & Chemical Engineering 2017;96:212–236.

[3] Wu N, Bai L, Zhou M. An Efficient Scheduling Method for Crude Oil Operations in Refinery With Crude Oil Type Mixing Requirements. IEEE Transactions on Systems, Man, and Cybernetics: Systems 2016;46(3):413–426.

[4] Li J, Misener R, Floudas CA. Scheduling of crude oil operations under demand uncertainty: A robust optimization framework coupled with global optimization. AIChE Journal 2012;58(8):2373–2396.

[5] Cerdá J, Pautasso PC, Cafaro DC. Efficient Approach for Scheduling Crude Oil Operations in Marine-Access Refineries. Industrial & Engineering Chemistry Research 2015;54(33):8219–8238.

[6] Castro PM. Source-based discrete and continuous-time formulations for the crude oil pooling problem. Computers & Chemical Engineering 2016;93:382–401.

[7] Assis LS, Camponogara E, Grossmann IE. A MILP-based clustering strategy for integrating the operational management of crude oil supply. Computers & Chemical Engineering 2021;145:107161.

[8] Zhang S, Xu Q. Refinery continuous-time crude scheduling with consideration of long-distance pipeline transportation. Computers & Chemical Engineering 2015;75:74–94.

[9] Reddy PCP, Karimi IA, Srinivasan R. Novel solution approach for optimizing crude oil operations. AIChE Journal 2004;50(6):1177–1197.

[10] Zimberg B, Camponogara E, Ferreira E. Reception, mixture, and transfer in a crude oil terminal. Computers & Chemical Engineering 2015;82:293–302.

[11] Assis LS, Camponogara E, Zimberg B, Ferreira E, Grossmann IE. A piecewise McCormick relaxation-based strategy for scheduling operations in a crude oil terminal. Computers & Chemical Engineering 2017;106:309–321.

[12] Yadav S, Shaik MA. Short-Term Scheduling of Refinery Crude Oil Operations. Industrial & Engineering Chemistry Research 2012;51(27):9287–9299.

[13] Castro PM, Grossmann IE, Zhang Q. Expanding scope and computational challenges in process scheduling. Computers & Chemical Engineering 2018;114:14–42.

[14] Dai X, Zhao L, He R, Du W, Zhong W, Li Z, et al Data-driven crude oil scheduling optimization with a distributionally robust joint chance constraint under multiple uncertainties. Computers & Chemical Engineering 2023;171:108156.

[15] Panda D, Ramteke M. Reactive scheduling of crude oil using structure adapted genetic algorithm under multiple uncertainties. Computers & Chemical Engineering 2018;116:333–351.

[16] Zhang S, Xu Q. Reactive Scheduling of Short-Term Crude Oil Operations under Uncertainties. Industrial & Engineering Chemistry Research 2014;53(31):12502–12518.

[17] Zhang S, Wang S, Xu Q. A New Reactive Scheduling Approach for Short-Term Crude Oil Operations under Tank Malfunction. Industrial & Engineering Chemistry Research 2015;54(49):12438–12454.