Research Of Discrete And Continuous Time Models For The Crude Oil Scheduling Problem

Optimal scheduling on unloading, transfer and charging operations of crude oilenables high profitability and process stability in refineries. This thesis models andsolves the crude oil scheduling problem in three different types of refineries: the in-land refinery, the coastal refinery and a multi-regional refinery. The problem can beposedasamixed-integernonlinearprogram(MINLP)usingdiscreteorcontinuoustimeformulations.For the inland refinery problem, three state-of-the-art models, i.e., the event-basedmodel, the unit slot model, and the multi-operations sequence (MOS) model, are pre-sented, implemented, analyzedandmodifiedtofurtherimprovetheefficiency. Compu-tational results indicate that optimal solutions can be obtained with less computationalefforts by the proposed variants of the MOS model.The present MOS model cannot be applied to the coastal refinery problem, mainlybecause of the distinct operational features of the coastal refinery, and the intractabilityof the large-scale nonconvex MINLP problems. To model the problem, the MOS mod-el is tailored to incorporate various operating rules of the coastal refinery problem. Tosolve the MINLP model effectively, a global optimization algorithm based on the im-proved multi-parametric disaggregation technique is proposed. For most test problems,solutions with small relative gaps can be obtained within limited time.Aformaldefinitionofthemulti-regionalrefineryanditscrudeoilschedulingprob-lem are presented. As the problem is large-scale with many nonlinear constraints rep-resenting the practical features, it is intractable from both the modeling perspective andthecomputationalperspective. Inordertoreducethecomputationalburden,ahierarchi-caldecompositionapproachisproposed. Inthehierarchicaldecompositionscheme, theupper-level model distributes the crude oil among regions without taking into accountof detailed operations. Lower-level problems of each region are then solved to obtainthedetailedscheduleswithineachregion. Atwo-stagedecompositiontechniqueforthe upper-levelproblemisproposedtofurtherreduce thesolutiontime. Forthelower-levelproblem, acomputationallyefficientheuristicalgorithmisalsodeveloped. Satisfactoryand flexible schedules are obtained by applying the above hierarchical decompositionmethodologyandtheheuristicalgorithmtorealinstancesofthemulti-regionalrefinery.The results of this study have not only contributed to the theoretical research men-tality of optimal crude oil scheduling, but also made the positive exploration of theapplication of mathematical modeling and optimization techniques to practical large-scale crude oil scheduling problems.