Efficient rolling horizon approach to a crude oil scheduling problem for marine-access refineries

We aim to address a crude oil scheduling problem for a marine-access refinery considering the storage, blending and processing of complex types of crude and the demands of feedstock from downstream units. Since diverse crude oils present significant variations in compositions, properties, and yields, the incoming crude oils should be appropriately stored and blended in refinery tanks and a proper charging schedule of tanks feeding crude distillation units (CDUs) should be determined to meet quality specifications and production demands. Meanwhile, complex types of crude can eventually lead to a complex mixture in the storage tanks, which affects the model size and final solution of the mathematical model. In this paper, a full-space mixed integer nonlinear programming (MINLP) model is proposed with new constraints represented by (1) a better optimization for the operation of crude unloading and blending, and (2) an application of heuristic rules to deal with the complex mixtures in the tank. An efficient rolling horizon approach with variable fixing strategy and constraint pruning strategy is developed for instances of large model sizes. Case studies are performed to show the efficiency of this approach compared with the other methods, and the influence of new constraints on the final solution. The results show that it is efficient to find a feasible solution of reasonable quality.