This paper presents a novel approach to aid the operational decision-making of scheduling activities in a real-world pipeline, transporting heavy oil derivatives, which are products of less aggregate value, such as fuel oils, e.g. marine fuel. These products present special characteristics that influence their transport as the impossibility of being transferred at room temperature, due to their viscosity, or the use of shared tanks for different products. Thus, during the transport of such products, the entire pipeline network (and the tanks) must be maintained heated during all the pumping process. Such characteristics imply that a specific model oriented to this type of problem must be developed. The approach proposed in this work develops a decomposition procedure that uses a sequence of mathematic programming models and heuristics to solve the problem in hand. The proposed approach is tested using a real-world scenario, composed of a pipeline tree system.
The use of pipelines to transport petroleum and its derivatives is a commonly used solution within the oil supply chain. Despite the high cost of initial investment, pipelines are the most economical way to transport large volumes of products at great distances when compared to other modes such as railroads, ships or roads. Several pipeline configurations exist in the entire petroleum industry supply chain: from a single source (usually a harbour) to a single destination (a refinery), from a single source (the refinery) to several destinations (consumption and distribution centres), from multiple sources to multiple destinations, in which case the system is commonly named a multi-pipeline system (Cruz, Andrés, Herrán, Besada, & Fernández, 2003). Solutions of planning and scheduling to all these configurations have been recently proposed, in which authors emphasize the complexities that characterize the tackling of all constraints and details. However, these works propose solutions only to simplified versions of the general problem. Rejowski and Pinto (2004), Magatão, Arruda, and Neves Jr. (2004), and Zyngier and Kelly (2009) used discrete Mixed Integer Linear Programming (MILP) formulations, dividing the planning horizon into equal and fixed time intervals. This decision lead to a situation where obtained solutions are only useful if the discretization period is very small, and thus the computational time to find the optimal solution, even for small planning horizons (75–120 h), is high. Cafaro and Cerdá (2004) proposed a rolling horizon technique to this kind of problem, enabling larger time periods. Relvas, Barbosa-Póvoa, and Matos (2009) proposed a MILP model that addresses the inventory management at the final destination, but only for a multiproduct pipeline with a single origin and only one destination area. Proposals that approached more realistic versions of the problem, with several origins and destinations over a complex pipeline network, have been recently presented (Cafaro & Cerdá, 2009), applying both structural and temporal decomposition techniques to achieve a solution for the planning and scheduling problems in reasonable computational times (Neves et al., 2007 and Boschetto et al., 2010).
In this paper, we follow this later tendency and a structural decomposition solution is proposed, using MILP models and heuristic approaches to solve the complex problem of heavy oil derivatives transportation. The presented approach is applied to a less common situation, in which several refineries must send their production to a single receiving area, in this case a harbour, for exportation. The approach is a development of the work first proposed by Neves Jr. et al. (2007), and further developed by Boschetto et al. (2010). However, the solution herein developed incorporates some real-world constraints that make it more applicable to real operational scheduling, such as restrictions on tank availability and heating constraints – restrictions present when heavy oil products are transported, cases not commonly treated in the published literature.
This paper is organized as follows. Section 2 provides the problem description including the problem constraints and a block diagram of the proposed hierarchical decomposition. In Section 3 the proposed solution is detailed, comprising the optimization and heuristic models. In Section 4 some results are presented, and finally in Section 5 the conclusions are drawn.
This paper presents a decomposition solution to the scheduling of pipelines used to deliver heavy oil derivatives. A real case study was solved as motivating problem. This involves the transportation of oil derivatives from 4 refineries to a harbour, in a real-world pipeline. Multiple products must be scheduled for transport over a multi-origin pipeline, but some particularities that were never properly treated in the literature are here considered. These particularities are namely: (1) the presence of heating constraints that impose a maximum transit time for each batch over each pipeline and (2) the possibility of use the same tank for different products during the scheduling horizon. To cope with these problems, a decomposition solution is proposed, based on MILP mathematical models, and heuristic models, that treat each part of the problem separately. The interplay of the blocks is able to, after two feedback loops, achieve feasible solutions for the real world considered scenarios, coping with the above mentioned particularities. The obtained solution is applicable for a scheduling horizon of 30 days, and it was obtaining in less than 300 s of computing time, on a standard computing platform. The developed decomposition solution has proved to be a useful tool to help experts plan the scheduling of the heavy-oil derivatives where the need of accounting for maximum pumping time due to heating constraints to avoid solidification of pipelines coupled with the presence of heated tanks is required.
