A common refinery's objective is to ally operation safety and product quality to the process profitability. In this sense, computational resources in distribution logistics, planning and scheduling activities of petroleum refineries have been presented in the context of the Operational Research since the 1950's through the application of linear programming (LP) techniques [1], which remain as the most currently used mathematical programming approach for modeling this type of chemical plant. We tried to solve this problem using a Mixed-integer programming approach [2], but it turned out after incorporating all the specification of the problem this alternative became impracticable. In this work, we address the LPG transfer and storage scheduling in petroleum refinery by means of a genetic algorithm approach. Moreover a solution representation based on real numbers allows a continuous representation of the time leading to a more realistic solution.

The Liquefied Petroleum Gas (LPG) transfer and storage operations at the Capuava Refinery (RECAP) in São Paulo are a challenged problem due to the large amount of product transferred and its lack of storage capacity. Furthermore, the time windows constraint imposes an additional difficult to find a good solution without recourse to a computational tool.

The usual approach to solve scheduling problem by genetic algorithm is the discretization of the time and associate chromosomes to activities where each gene represents the state of an equipment in a given time [3]. Another important issue that arises when applying genetic algorithm is the feasibility of the solution generated. There are a lot of techniques to deal with infeasible solution, but as a rule of thumb this is something that should be avoided. Our algorithm overcomes these two problems pointed out by using a solution codification using real number and an algorithm for constructing solutions guaranteed feasible.

Our model was applied in some test instances of the problem and the results show a great improvement compared to a random search. Moreover, as the construction algorithm is based on the knowledge of the specialists, the solutions have more chance to be accepted in the refinery. Another advantage of our approach is the consideration of production specification in the constraint set of the problem can be made nearly in a transparent way.

References

[1] Symonds, G.H. (1955). Linear Programming: The Solution of Refinery Problems. New York, Esso Standard Oil Co.

[2] Joly, M. ; Aires, M. ; Rocha, R. ; Smania Filho, P. (2005). Short-term Scheduling of the RECAP Refinery LPG area: A Modeling and Solution Algorithm. ENPROMER – 2nd Mercosur Congress on Chemical Engineering and 4th Mercosur Congress on Process Systems Engineering, Rio de Janeiro.

[3] Sikora, R. (1996). Integrating the Lot-sizing and Sequencing Decisions for Scheduling a Capacitated Flow Line. Computers Industrial Engineering, 30, 4, 969-981.