In the oil industry, any improvement in the planning and execution of the associated operations (e.g., production, storage, distribution) can generate considerable profits. To achieve this, the related activities need to be optimized. Within these activities, planning and scheduling occur at the different levels of the oil supply chain, from the strategic to the operational levels looking from global networks to sets of individual resources. This work looks into the planning, namely the assignment/ sequencing of activities that occur in a multiproduct, multipipeline system. The aim is to contribute to the definition of generic models that can help the decision-making process characterized by a high level of complexity. An approach formed by two mixed integer linear programming (MILP) formulations that act in sequence is proposed. The first generic MILP planning model calculates volumes for attending the necessary requirements on inventory management of the producer and consumer areas. As a result, this model defines the products and the total volumes to be transported in order to attain storage goals, while respecting operational constraints, demands of consumers, and pipeline capacity. Then, the planning model results are used by an MILP assignment and sequencing model, which splits the total volume into operational batches and determines the sequence of pumping for the batches during the available horizon. The developed approach is applied to a real-world pipeline network that includes 30 bidirectional multiproduct pipelines associated with 14 node areas: four refineries, two harbors, six depots/parks of pumps and valves, and two final clients.

No abstract

This work presents an approach for scheduling of operational activities in a large real-world pipeline network, where oil derivatives and ethanol are transported and distributed among refineries, terminals, depots, and final clients. The hierarchical decomposition approaches to solve the pipeline-scheduling problem presented by Boschetto et al. [Ind. Eng. Chem. Res. 2010, 49, 5661] and Magataõ et al. [Ind. Eng. Chem. Res. 2012, 51, 4591], which are based on the integration of mixed integer linear programming (MILP) models and a set of heuristic modules, are merged and compounding blocks are also improved. Thus, a novel decomposition approach for scheduling product distribution through a pipeline network is proposed. In addition, this work presents a new MILP approach for the last hierarchical level: the timing block (timing model). This paper expands and improves the former MILP model, which was the timing block core. A series of operational constraints were considered within a continuous time representation in order to determine the exact time instants that products should be pumped into the pipelines and received in the operational areas during a scheduling horizon of, typically, 1 month. Within the new MILP timing model, turn shift constraints, local constraints, and surge tank constraints are improved; immediate pumping constraints are proposed. In addition, a decomposition approach for the new MILP model is also proposed within this article. This decomposition is based on a relax-and-fix heuristic implemented by a sequential run of two MILP models: MLC (Model with Local Constraints) and MST (Model with Seasonal costs and Turn shift constraints). The MILP decomposition goal is to reduce the computational load, if seasonal costs and turn shift constraints are active, without quality solution losses. The proposed approach is applied to the solution of real case studies of a pipeline network that includes 30 bidirectional multiproduct pipelines associated with 14 nodes (four refineries, two harbors, six depots, and two final clients). Computational results have been attained in a reasonable computational time (from seconds to a few minutes) for the addressed pipeline network.

This paper presents an optimization model to reduce the gap between planning and scheduling activities in pipeline networks for supply of petroleum derivatives, so as to effectively operate the plans developed at higher decision levels. Most of the information taken in the petroleum supply chain is unidirectional, from the strategic level passing through the tactical level and finally reaching the operational level. However, unviable schedules are often obtained from an operational standpoint, since operational criteria are not considered when decisions are made at the strategic and tactical levels. In particular, in the supply chain of refined petroleum products, the production campaigns of the refineries are defined for different regions based on the consumer market and no operational conditions related to distribution are accounted at this stage. This fact that may lead to operational infeasibilities. To overcome this type of problem the present paper proposes a Mixed Integer Linear Programming (MILP) approach to validate the refineries' production plans where operational conditions are accounted for, and as a consequence, a decrease in the existent gap between the scheduled and executed operations is obtained. The proposed methodology is applied to a real world case study of a Brazilian Oil company.

This paper presents a collaborative approach to the assignment and sequencing of batches in pipeline networks. The approach is based on the integration of a heuristic algorithm with a mixed integer linear programming (MILP) model. The pipeline-scheduling problem is solved using a hierarchical decomposition [Ind. Eng. Chem. Res.2015545077], but a new collaborative approach is proposed for assignment/sequencing tasks. At a first step, the proposed heuristic algorithm (assignment module) determines priorities for sending batches in order to respect deadlines. The algorithm encompasses an analysis of production and demand plans, inventories, and input and output of products in terminals, trying to use resources, namely tanks and pipelines, in an optimized form. This algorithm is used in cooperation with a proposed MILP sequencing model, which allows overcoming computation difficulties previously indicated by a traditional scheduling approach that tried to aggregate into the same monolithic MILP model assignment and sequencing decisions [Ind. Eng. Chem. Res.2012514591]. The proposed assignment/sequencing collaborative approach can be used to define operational batches with their volumes and routes in pipeline networks. Thus, the lot-sizing problem of batches in pipeline networks is addressed within the proposed paper. Tests were made in pipeline networks of different topologies. First, a small, but representative pipeline network is proposed and a data set for this network is made available for reproducibility purposes. Second, tests are made in a real-world pipeline network and results have been attained in computational times from seconds to few minutes.