A refinery consists of crude oil operations, unit operations and product blending. In crude oil operations the refinery deals with the availability of crude oil and crude oil management, whereas product blending deals with the fluctuating market requirement of blended products. In the whole process we can see that the information of demand flows towards the suppliers whereas the material of products flows towards the customers. However, because of volatile raw material prices, fluctuating product demands, and other changing market conditions, many parameters in a planning/scheduling model are uncertain. Moreover many times, due to unavailability of crude oil or accidents in pipelines or unlikely events, the problem becomes more difficult and in some cases infeasible. One objective for oil operations in a refinery is to optimise the cost for crude oil inventories and product inventories. The current picture of the refining industry is characterised by stiff competition, stricter environmental regulations and heavier, sourer, costly crude oils, uncertainty in crude oil prices and uncertainty in the availability of the crude oil. Therefore, to maintain the profit margins in this ever-changing market environment, refiners need to have smarter operation strategies as well as keeping the backup options on the other hand in case of uncertainties.

A novel approach for inventory management of a refinery under uncertainty of the availability of the crude oil is presented in this work. Considered problems involve the uncertainty in the crude oil availability, its transfer to storage tanks and the charging schedule for each crude oil mixture to crude distillation units. The state of the art for the scheduling problem is to solve it either by discrete-time formulation or continuous-time formulation. The existing methods using discrete-time formulation results with high computational demand due to large amount of binary variables whereas continuous-time formulation results with a lower computational demand but not able to provide global optimum solution. To calculate the expected cost of the crude oil scheduling problem under uncertainty (includes late/early arrival of crude vessels), piecewise linear approximation of loss function is applied.

A mixed integer linear/nonlinear optimisation model is developed for the inventory management problem. The proposed model incorporates the uncertainty issue in the availability of crude oil and an effective solution algorithm is developed which provides good quality solutions and also save the computational efforts. The existing methodologies available in the literature provide single point optimum solution and many times it has been found that it is not possible to go with the solution. To consider the flexibility issue we have proposed an “operating window” concept which provides a solution space instead of a single point solution and decisions can be changed dynamically to avoid losses. During the study it has been found that mixed integer linear formulation may result in the form of discrepancy in properties. To avoid the property discrepancy in the solution, an algorithm is incorporated that iteratively solves two mixed integer linear programming models and a nonlinear programming model. A case study is carried out to demonstrate the effectiveness of the developed algorithms. As our proposed approach provides an operating window so the schedule can be carried out in more ways for the same optimum cost.