Well placement optimization is one of the major challenging factors in the field development process in the oil and gas industry. This chapter aims to survey prominent metaheuristic techniques, which solve well the placement optimization problem. The well placement optimization problem is considered as high dimensional, discontinuous, and multi-model optimization problem. Moreover, the computational expenses further complicate the issue. Over the last decade, both gradient-based and gradient-free optimization methods were implemented. Gradient-free optimization, such as the particle swarm optimization, genetic algorithm, is implemented in this area. These optimization techniques are utilized as standalone or as the hybridization of optimization methods to maximize the economic factors. In this chapter, the authors survey the two most popular nature-inspired metaheuristic optimization techniques and their application to maximize the economic factors.
TopIntroduction
Metaheuristics are generally implemented to tackle problems when no satisfactory problem-specific algorithm is available to solve them. “Conventional” mathematical programming methods such as gradient-based methods have not provided satisfactory results in solving optimization problems which comprise of multi-modality, discontinuity, and non-smooth cost function. So, Metaheuristic has emerged as an alternative to traditional methods. In recent years Stochastic algorithms are widely used in different complex problems in the real world.
In the current epoch, the Well placement optimization problem has become a major concern in the field development process in the Oil and gas industry. This optimization problem is viewed as a challenging problem cause the objective function is multimodal, nonconvex and discontinuities (J. E. Onwunalu & Durlofsky, 2010). Previously Classical methods had been utilized to tackle well placement optimization problems(Rosenwald & Green, 1974), (Ma, Plaksina, & Gildin, 2013), (Pan & Horne, 1998), (Li & Jafarpour, 2012),(Jansen & Fluids, 2011), (Bangerth, Klie, Wheeler, Stoffa, & Sen, 2006),(Zhang et al., 2016) . On the other hand, Non-classical derivative-free stochastic techniques do not require the calculation of derivatives and they are less likely to get stuck in local optimal (Isebor, Durlofsky, & Ciaurri, 2014), (Giuliani & Camponogara, 2015), (Fahim Forouzanfar, Reynolds, & Engineering, 2013). Derivative-free stochastic techniques have the capacity to avoid local optima due to their inherent stochasticity. Meta-heuristics approaches are probabilistic in nature and controlled by parameters, e.g., population, elite population size, the number of generations, etc. It is true that the performance of Metaheuristic is largely dependent on parameter tuning. Due to computational expenses, proper tuning of parameters for Well placement optimization is challenging and improper tuning may result in increased computation expenses or local optima. The Recent attention from the researcher in well placement optimization is summarized in the Table 1.
Table 1. Recently applied methods on well placement optimization
| No. | Reference work | Year | Diligence |
| 1. | (S. Ding et al., 2019) | 2019 | Incorporated DMPP technique with PSO. |
| 2. | (Redouane, Zeraibi, & Nait Amar, 2018) | 2019 | An intelligent neighborhood search mechanism is combined with GA to improve the quality of proxy models. |
| 3. | (Janiga, Czarnota, Stopa, & Wojnarowski, 2019) | 2019 | Authors developed a clustering approach that significantly reduces the convergence time of the particle swarm optimization algorithm. |
| 4. | (Miyagi & Yamamoto, 2018) | 2018 | CMA-ES is included with mixed-integer support for Well Placement optimization. |
| 6. | (Jang, Oh, Kim, Park, & Kang, 2018) | 2018 | The search space is successively refined using an artificial neural network. |
| 7. | (Chen et al., 2018) | 2018 | Combined cat swarm optimization (CSO) algorithm, particle swarm optimization (PSO) mesh adaptive direct search (MADS) algorithm. |
| 8. | (Chen et al., 2017) | 2017 | An analytical expression used to calculate the cost function using the cat swarm optimization (CSO) algorithm. |
| 9. | (Hamida, Azizi, & Saad, 2017) | 2017 | PUNQ-S3 and Brugge field datasets are used to test the robustness of the Genetic Similarity Algorithm (GSA). |
| 10. | (Pouladi, Keshavarz, Sharifi, & Ahmadi, 2017) | 2017 | By combining FMM-based methods with Particle Swarm Optimization algorithms, authors introduced a surrogate model that includes cost functions. |
| 11. | (Sayyafzadeh, 2017) | 2017 | Development of a computationally cheaper self-adaptive model management strategy using a proxy support algorithm. |
| 12. | (Jesmani, Jafarpour, Bellout, Hanea, & Foss, 2016) | 2016 | Authors adopted the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm using an efficient stochastic gradient approximation. |