A Least Squares Method for Ensemble-based Multi-objective Oil Production Optimization

Despite a significant potential to improve industrial standards, practical applications of production optimization are impeded by geological uncertainty. As a mean to handle the associated financial risks, the oil literature has devised a range of ensemble-based strategies that seek to optimize proper combinations of sample-estimated risk measures to balance the opposing objectives of risk and reward. Many of the associated optimization problems are naturally formulated in terms of multi-objective optimization (MOO). Ideally, MOO problems should be solved by generating an approximation to the efficient frontier of optimal trade-offs between risk and return. However, the large-scale nature of real-life oil reservoirs implies that formation of the frontier often becomes computationally intractable in practice. To meet this challenge, this paper introduces a generalized least squares (LS) approach that provides an efficient and unified solution strategy for ensemble-based multi-objective optimization problems. At its core, the LS method uses an a priori characterization of desirable trade-offs that allows the method to focus on a single Pareto optimal point. Consequently, the LS approach avoids the need to generate a representative of the efficient frontier. In turn, this significantly reduces computational complexity compared to most MOO methods. As a result, the LS method poses a practical alternative to conventional strategies when the efficient frontier is unknown and computationally intractable to generate. (C) 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.