Multi-objective drilling trajectory optimization using decomposition method with minimum fuzzy entropy-based comprehensive evaluation
Introduction
Horizontal wells are the main type of resource exploitation, such as oil, natural gas, and shale gas. Pre-drill trajectory design is one of the key points to improve the safety and efficiency of drilling. Measuring every point on a trajectory in drilling practice is very difficult, so a drilling trajectory’s actual shape is always unknown. Therefore, it is necessary to raise assumptions to calculate the coordinates of points on a drilling trajectory [1]. The commonly used mathematical models are curvature radius method [2], circular arc method [3], minimum curvature method [4], and nature curve method [5]. Based on these models, the optimization of a drilling trajectory tries to find the optimal trajectory under some specific indicators and constraints.
In a definite drilling technology and geological environment, decreasing the trajectory length can reduce the cost of a drilling project. The description parameters (coordinates, inclination angles, and azimuth angles, etc.) of a trajectory model always have nonlinear relationships, leading to nonlinearity in constraint functions. Comparing with methods that need gradient information of the mathematical model, evolutionary algorithms only need the value of objective functions to guide the search. They are popular in drilling trajectory optimization problems to minimize a trajectory length [6], [7], [8].
However, a small trajectory length sometimes refers to a complex well structure, such as a rapid variation of inclination angles and azimuth angles, and significant dog-leg severity. They will increase the risk of drilling accidents. Thus, drilling trajectory optimization should be considered as a multi-objective optimization problem (MOP).
Some researchers introduced the target error as the second objective [9], [10], [11]. However, they did not study the complexity of a trajectory and only considered the range of decision parameters as constraints.
Mansouri et al. [12] introduced drill-string torque as the second objective. Furthermore, they applied the multi-objective genetic algorithm to solve the MOP. Based on this MOP model, Zheng et al. [13] introduced well-profile energy as the third objective. Although both of these indicators can reflect the complexity of the drilling trajectory, the target error was not considered. Besides the constraint of decision parameters’ range, only some non-negative constraints are studied.
Wang et al. [14] proposed a MOP of a sidetracking horizontal well trajectory. Their study considered the trajectory length, the complexity of the trajectory, and the target error. Limiting the location of the trajectory endpoint is set as a constraint. To solve the MOP, they transformed tri-objective functions into a single objective function by a linear weighted sum method. On the one hand, the transformation may lose some information of the MOP. On the other hand, the algorithm cannot obtain non-dominated solutions in a single run. Furthermore, decision-makers have to intervene in the optimization process. In our previous work [15], we aimed at these disadvantages and tried to solve the MOP proposed by Wang et al. [14] without considering the target error. Table 1 shows the objectives and constraints used in previous studies.
For the MOP mentioned above, constraint handling is a critical part. Some multi-objective optimization algorithms deal with constraints and objectives separately, such as constrained NSGA-II [16] and C-MOEA/D [17]. In these two algorithms, the feasible individuals are always dominating infeasible ones. Infeasible individuals are sorted by their constraint violation. These methods may lose fitness information of infeasible individuals, which affects the population convergence. As an improvement of this, an epsilon constraint handling method is studied [18], [19]. In these methods, the domination relationship depends on the epsilon value and the constraint violation. Penalty methods are convenient to combine with different optimization algorithms and effectively applied to engendering problems [20]. Especially adaptive penalty function (APF) methods can direct the population to the feasible area according to the current population status [21], [22]. It is appropriate to combine APF methods with MOEA/D (multi-objective evolutionary algorithm based on decomposition) [23] to handle constraints in MOPs. The combination can take advantage of applying APF methods to scalar optimization algorithms.
Multi-objective evolutionary optimization algorithms can obtain a set of solutions in a single run. But applying every solution to drilling practice is impossible. Therefore, it is necessary to provide preference solutions to decision-makers. Many researchers used multiple criteria decision making (MCDM) algorithms such as TOPSIS (technique for order preference by similarity to an ideal solution) [24], LINMAP (the linear programming technique for multidimensional analysis of preference) [25], and Shannon Entropy [26] to address these problems. These methods rank solutions and select the best one as the final solution [27], [28]. However, the preference of a decision-maker is a qualitative evaluation. A fuzzy comprehensive evaluation method transforms the qualitative evaluation into the quantitative evaluation, and its application in engineering problems was studied [29]. The membership degree theory of fuzzy mathematics can quantify the preferences of a decision-maker.
Considering the characteristics of the MOP and the strength and the weakness of existing algorithms, the main contribution of this work includes three parts:
- (1)
The nonlinear, multi-objective, and multi-constraint of a drilling trajectory optimization problem are taken into consideration. Combining an adaptive penalty method with MOEA/D, adaptive penalty function based MOEA/D (APF-MOEA/D) is proposed. The proposed algorithm is effective for MOP in drilling practices and has a higher hypervolume (HV) comparing with other methods.
- (2)
Because of the ambiguity of decision-maker’s preference, a comprehensive evaluation method based on minimum fuzzy entropy (CE/MFE) is proposed. The membership functions are established, and solutions are ranked according to an evaluation matrix. CE/MFE can obtain the solutions that have a high membership degree in good under different weight vectors.
- (3)
A drilling trajectory optimization problem is studied to ensure the efficiency and safety of the drilling process. The application of MOEA/D with CE/MFE on this MOP could give control parameters that satisfying different preference situations in drilling practice.
A parameter setting is done to obtain better performance on this MOP. The performance of APF-MOEA/D is compared with NSGA-II, C-MOEA/D. We also combined an epsilon constraint handling method [19] with MOEA/D and named it SaE-MOEA/D as a comparison. To illustrate the parameter sensitively, we studied the other two engineering problems in drilling practice. Some popular decision-making methods are used as a comparison: TOPSIS, LINMAP, Shannon Entropy.
The paper is organized as follows. Section 2 formulates the multi-objective drilling trajectory optimization problem of sidetracking horizontal wells. Section 3 describes the optimization method and the decision-making method to solve the MOP. Section 4 studies the application of the proposed method on drilling practice problems and presents a comparison of proposed methods with some existing methods. Finally, conclusions are made in Section 5.
Section snippets
Model formulation of multi-objective drilling trajectory optimization
This part mainly illustrates the formulation of the drilling trajectory model, which includes three objectives and several constraints.
Sidetracking horizontal wells always face complex formation when applied to the revitalization of shut down wells and low-production wells. Comparing with general directional wells, sidetracking wells often have a more complex structure. This research aims to optimize a sidetracking horizontal well model proposed by Wang et al. [14]. The blue line in Fig. 1
Multi-objective optimization and evaluation
The optimization model described above is a nonlinear constrained MOP with tri-objective, and the objectives have a much different scale. According to the characteristics of MOP, MOEA/D and adaptive penalty function (APF) are combined to form the proposed APF-MOEA/D. To deal with the ambiguous preference of decision-makers, we propose a method named comprehensive evaluation based on minimum fuzzy entropy (CE/MFE).
MOEA/D decomposes a MOP into multiple scalar optimization problems to reduce the
Case study
In this section, the MOP of a sidetracking horizontal well trajectory is studied. Before the optimization, the parameter setting is discussed to specific the algorithm to the application problem. APF-MOEA/D is applied to solve the MOP, and a set of solutions is obtained. To demonstrate the parameter sensitivity when instances varied, we studied the other two instances from drilling engineering. The MOP solutions are selected through CE/MFE, offering a final solution to the decision-maker.
Conclusions
Aiming to a tri-objective drilling trajectory optimization problem, we proposed APF-MOEA/D and CE/MFE in this study. APF-MOEA/D is used to find solutions under a series of nonlinear constraints. CE/MFE is used to choose a preference one from a set of solutions. The application of APF-MOEA/D with CE/MFE method and the comparison with other algorithms got satisfying results. Several conclusions can be drawn according to the results obtained by this study:
- •
Simulation results show the APF-MOEA/D has
CRediT authorship contribution statement
Wendi Huang: Conceptualization, Methodology, Writing - original draft, Software, Validation. Min Wu: Supervision, Writing - review & editing. Luefeng Chen: Methodology, Writing - review & editing. Xin Chen: Writing - review & editing. Weihua Cao: Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the National Natural Science Foundation of China under Grant 61733016, the National Key R&D Program of China under Grant 2018YFC0603405, the Hubei Provincial Technical Innovation Major Project, China under Grant 2018AAA035, the 111 Project under Grant B17040, and the Fundamental Research Funds for the Central Universities, China under Grant CUGCJ1812.
References (33)
- et al.
Designing and optimizing deviated wellbore trajectories using novel particle swarm algorithms
J. Natural Gas Sci. Eng.
(2014) Metaheuristic profiling to assess performance of hybrid evolutionary optimization algorithms applied to complex wellbore trajectories
J. Natural Gas Sci. Eng.
(2016)- et al.
Stochastic optimal control and algorithm of the trajectory of horizontal wells
J. Comput. Appl. Math.
(2008) - et al.
Three-dimensional trajectory design for horizontal well based on optimal switching algorithms
ISA Trans.
(2015) - et al.
3-D well path design using a multi objective genetic algorithm
J. Natural Gas Sci. Eng.
(2015) - et al.
Multi-objective cellular particle swarm optimization for wellbore trajectory design
Appl. Soft Comput.
(2019) - et al.
Multi-objective sidetracking horizontal well trajectory optimization in cluster wells based on DS algorithm
J. Petrol. Sci. Eng.
(2016) - et al.
A multi-objective differential evolutionary algorithm for constrained multi-objective optimization problems with low feasible ratio
Appl. Soft Comput.
(2019) - et al.
Performance comparison of low-grade ORCs (organic Rankine cycles) using R245fa, pentane and their mixtures based on the thermoeconomic multi-objective optimization and decision makings
Energy
(2015) - et al.
A novel multi criteria decision making model for optimizing time-cost-quality trade-off problems in construction projects
Expert Syst. Appl.
(2015)
Multi-objective inventory planning using MOPSO and TOPSIS
Expert Syst. Appl.
Multi-objective control char design optimization using NSGA-III and MOPSO enhanced with DEA and TOPSIS
Expert Syst. Appl.
Fuzzy comprehensive evaluation of virtual reality mine safety training system
Saf. Sci.
Horizontal well’s path planning: An optimal switching control approach
Appl. Math. Model.
Multi-objective optimization of water distribution system based on an improved NSGA2 algorithm
Control Eng. China
An improved method for computing directional surveys
J. Pet. Technol.
Cited by (17)
On the estimation of the core for TU games
2024, Expert Systems with ApplicationsApplication of fast marching method and quality map to well trajectory optimization with a novel well parametrization
2023, Geoenergy Science and EngineeringGranularity self-information based uncertainty measure for feature selection and robust classification
2023, Fuzzy Sets and Systems