This chapter will address the problem of scheduling operations in refineries. Such a problem is to define the sequence and the time of completion of each activity related to oil refining and derivatives. The system under consideration can be defined as a multiproduct system of three stages in series, consisting of tanks, production units and mixing units (mixers). Such a system is subject to operational constraints and availability of resources, and operate continuously. In this context, in order to support the decision process, an optimization model is proposed capable of representing the operating environment of such a system, which is applied to the particular context of a refinery. Two approaches are proposed for the model solution, the first based on Mixed Integer Linear Programming (PLIM) and the second based on Genetic Algorithms (GA). The model's main objective is to meet the expected demand for manufactured products, in compliance with the operational constraints of the refinery and minimizing the number of operational changes during the horizon of programming. To validate the proposed approach, we used a set of real data related to production of fuel oil and asphalt in a large refinery. The results suggest that both approaches are able to get good solutions to the problem at hand, and the AG more efficient in terms of goals of care, which is due to the incorporation of a multi-criteria approach which is able to update the weights each goal adaptively during the evolutionary process.
↧