Discrete-Time MILP Formulation for the Optimal Scheduling of Maintenance Tasks on Oil and Gas Production Assets Victoria G. Achkar, †,‡ Vanina G. Cafaro, †,‡ Carlos A. Me ́ ndez, †,‡ and Diego C. Cafaro* ,†,‡ † INTEC (UNL-CONICET), Gü emes 3450, 3000 Santa Fe, Argentina ‡ Facultad de Ingeniería Química (UNL), Santiago del Estero 2829, 3000 Santa Fe, Argentina * S Supporting Information ABSTRACT: In this work, we address the scheduling of general maintenance tasks on oil and gas wells and surface facilities, using a mixed integer linear programming (MILP) formulation. The problem involves multitask workorders with precedence relations, comprising both preemptive and nonpreemptive operations, using parallel heterogeneous resources. Crews with different capabilities and work shifts are accounted for. The model introduces novel ideas for expanding the scope of discrete-time formulations. We solve a scheduling problem that includes crew traveling decisions to minimize operational costs and production losses. Different instances of real- world examples are efficiently solved in reasonable times for the industry. Results demonstrate the benefits from optimizing the maintenance schedule of oil and gas production assets and the effectiveness of the solution approach. 1. INTRODUCTION Crude oil and natural gas are the main energy sources worldwide and also important feedstocks for a large number of industrial products. 1 Onshore oil and gas fields are usually composed of a set of wells geographically distributed over an exploitation area. The extraction of oil and gas usually requires costly materials and equipment, including production and surface facilities, which need regular maintenance in order to keep efficiency sufficiently high. These activities comprise safety, environmental protection, and regulatory tasks, as well as preventive, predictive, and corrective maintenance. Some- times, the production from the well needs to be stopped, yielding loss of profits. 2 Moreover, many tasks require very expensive resources such as qualified crews and specialized rigs. In practice, tasks are grouped into workorders, and there are many costs associated with the start time and the length of an order, mainly related to production losses and risks of failure. This paper addresses the scheduling of maintenance tasks on onshore oil and gas assets of an Argentine exploitation area. There are very few works in the literature focused on the optimal operation of onshore wells considering general maintenance tasks and multiple heterogeneous resources. Villagra et al. 3 account for oil and gas well maintenance operations by resembling the single machine scheduling problem. They propose an evolutionary heuristic algorithm to obtain the schedule of visits to the wells. In fact, several works have tackled the maintenance planning problem in the oil and gas industry but just focusing on one specific resource: the workover rig. This expensive and scarce resource can make various tasks, such as cleaning and repairing the wells, involving interventions of several days. Ferná ndez Pé rez et al. 4 remark that workover rigs have different capabilities depending on their size and type, and a well can be serviced only by the rigs that can perform the required tasks. The same authors highlight that the rig planning problem has been addressed following two approaches: the workover rig scheduling problem (WRSP) and the workover rig routing problem (WRRP). Both problems can be seen as the scheduling of jobs on parallel heterogeneous machines, but the second one also includes the time required to move the rigs from well to well. Generally, the first approach is adopted when traveling times are in the order of minutes or hours and well interventions are in the order of days or weeks. Different solution techniques for the WRRP can be found in the literature. Aloise et al. 5 present the problem of planning workover rigs to visit a set of wells demanding maintenance services so as to minimize the oil production loss in Brazilian fields. They propose a continuous-time formulation and make use of a VNS (Variable Neighborhood Search) heuristic to solve it. Heterogeneous resources and traveling distances are taken into account. Mattos Ribeiro et al. 2 consider the WRRP as a variant of the vehicle routing problem with time windows (VRPTW), seeking to minimize arrival times at wells weighted by the associate production loss. They propose iterated local search, clustering, and adaptive neighborhood heuristics to Received: February 13, 2019 Revised: April 16, 2019 Accepted: April 19, 2019 Published: April 19, 2019 Article pubs.acs.org/IECR Cite This: Ind. Eng. Chem. Res. XXXX, XXX, XXX-XXX © XXXX American Chemical Society A DOI: 10.1021/acs.iecr.9b00861 Ind. Eng. Chem. Res. XXXX, XXX, XXX-XXX Ind. Eng. Chem. Res. Downloaded from pubs.acs.org by BOSTON COLG on 05/08/19. For personal use only.