(A) knowledge-based approach to crude oil delivery scheduling

Abstract

Although there have been many mangement science applications to part of pertroleum industry, only a few integrated planning and scheduling models can be found. Moreover the acquisition of crude oil by ocean transportation has been overlooked. This comes from the inherent complexity of overall production scheduling in oil refinery. To tackle this problem a detailed daily schedule for crude oil delivery is generated under a monthly master plan. But even this problem is too complex to solve by traditional mathematical approach. So it is natural to try to solve this problem by Artificial Inteligence techniques. The problem is decomposed into three parts, usable inventory calculation, processing at crude distillation unit, and crude oil delivery, and then a heuristic search is applied. when the inventory of each crude oil is given the usable inventory at that thme is calculated. The usable inventory is reduced by the charge schedule of each CDU and the necessity of crude oil delivery is detected. A master plan is used to calculate those numbers. The contents of crude oil delivery is determined by domain knowledge. Complicated circumstances in mathematical approach is represented by simplified knowledge. The knowledge determines an assignment of crude oils to a vessel which would minimize the cost from delivery and quality loss. This procedure is continued until a schedule is accomplished. UNIK-PCS is developed to realize the proposed method and implemented in field. This system has three subsystems: usable inventory calculator, inventory checker and vessel dispatcher. Especially the vessel dispatcher is composed of production rules which represent the normative and expert knowledge for vessel assignment. UNIK-PCS is developed using LISP programming languae and uses AI tools like UNIK-FRAME and UNIK-FWD for knowledge representation and inference.