Efficient heuristics for the workover rig routing problem with a heterogeneous fleet and a finite horizon

Onshore oil fields may contain hundreds of wells that use sophisticated and complex equipments. These equipments need regular maintenance to keep the wells at maximum productivity. When the productivity of a well decreases, a specially-equipped vehicle called a workover rig must visit this well to restore its full productivity. Given a heterogeneous fleet of workover rigs and a set of wells requiring maintenance, the workover rig routing problem (WRRP) consists of finding rig routes that minimize the total production loss of the wells over a finite horizon. The wells have different loss rates, need different services, and may not be serviced within the horizon. On the other hand, the number of available workover rigs is limited, they have different initial positions, and they do not have the same equipments. This paper presents and compares four heuristics for the WRRP: an existing variable neighborhood Glaydston Mattos Ribeiro Transportation Engineering Program, Federal University of Rio de Janeiro, Brazil Tel.: +55-21-25628132 Fax: +55-27-25628131 E-mail: [email protected] Guy Desaulniers Department of Mathematics and Industrial Engineering and GERAD, École Polytechnique de Montréal, Canada E-mail: [email protected] Jacques Desrosiers Department of Management Sciences and GERAD, HEC Montréal, Canada E-mail: [email protected] Thibaut Vidal Massachusetts Institute of Technology and CIRRELT, MIT, USA E-mail: [email protected] Bruno Salezze Vieira Department of Computer Engineering and Electronics, Federal University of Esṕırito Santo, Brazil E-mail: [email protected] 2 Glaydston Mattos Ribeiro et al. search heuristic, a branch-price-and-cut heuristic, an adaptive large neighborhood search heuristic, and a hybrid genetic algorithm. These heuristics are tested on practical-sized instances involving up to 300 wells, 10 rigs on a 350period horizon. Our computational results indicate that the hybrid genetic algorithm outperforms the other heuristics on average and in most cases.