Oil production optimization solved by piecewise linearization in a Branch & Price framework

• Published in: Computers & operations research Vol. 39; no. 11; pp. 2469 - 2477
• Main Authors: Gunnerud, V., Foss, B.A., McKinnon, K.I.M., Nygreen, B.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.11.2012; Elsevier; Pergamon Press Inc
• Subjects: Applied sciences; Branch & Price; Column generation; Computer programs; Continental shelves; Crude oil; Decomposition; Exact sciences and technology; Inventory control, production control. Distribution; Linearization; Mathematical problems; Mathematical programming; Mixed integer linear programming; Natural gas; Oil fields; Oil production planning; Oil rim reservoir; Operational research and scientific management; Operational research. Management science; Operations research; Optimization; Optimization techniques; Petroleum production; Studies; Norway; Branch & Price; Column generation; Mixed integer linear programming; Oil rim reservoir; Oil production planning; Branching; Production system; Mass production; Oil field; Mixed integer programming; Distributed system; Modeling; Gas oil; Optimization; Coastal zone; Production management; Production planning; Piecewise linearization
• ISSN: 0305-0548; 1873-765X; 0305-0548
• DOI: 10.1016/j.cor.2011.12.013

This paper presents a method for optimizing oil production on large scale production networks such as the Troll west field in the North Sea. The method is based on piecewise linearization of all nonlinearities, and on decomposition of the full scale problem into smaller subproblems. Column generation in a Branch & Price framework is used to solve the decomposed problem. The method differs from most Branch & Price methods by branching only on continuous quantities and by solving the subproblems using commercial MILP software. The method is applied to a realistic model of an oil field, the Troll oil and gas field at the Norwegian Continental Shelf, a petroleum asset with severe production optimization challenges due to rate dependent gas-coning wells. This study shows that the method is capable of solving instances of practical size to proven optimality.