Selection of Optimal Well Trajectory Using Algorithm

Arabian Journal for Science and Engineering (2023) 48:16831–16855
https://doi.org/10.1007/s13369-023-08149-1
RESEARCH ARTICLE-PETROLEUM ENGINEERING
Selection of Optimal Well Trajectory Using Multi-Objective Genetic

Algorithm and TOPSIS Method
Hossein Yavari1 · Jafar Qajar2,3 · Bernt Sigve Aadnoy4 · Rasool Khosravanian4
Received: 2 June 2023 / Accepted: 16 July 2023 / Published online: 7 August 2023
© The Author(s) 2023
Abstract
This study presents a novel approach for optimizing well paths in extended reach drilling (ERD) wells. Different trajectories
can be used for ERD wells, each with its pros and cons. Previous research overlooked certain objective functions in single-
objective optimization and lacked an autonomous method for selecting the best solution from Pareto optimal solutions in
multi-objective optimizations. Furthermore, they lacked comparing different profiles in well design. Risk assessment and
operational factors, which greatly influence optimization and drilling success, were insufficiently considered. This study
utilized the multi-objective genetic algorithm (MOGA) and the technique for order preference by similarity to an ideal solution
(TOPSIS) method to select the optimal well path based on torque, wellbore length, risk (e.g., keyseat), and required tools.
First, all possible trajectories were determined, and MOGA identified the optimal path with minimal torque and length. The
fuzzy decision-making method automatically selected the best solution from the Pareto optimal solution set. The associated
risks and required tools are evaluated for each trajectory. Finally, the TOPSIS method selected the optimal trajectory based on
torque, length, risks, and required tools. The case study demonstrated that the undersection path was the most advantageous
trajectory for ERD wells, with a 60% closeness to the ideal state. The multiple build trajectory achieved 57% closeness, while
the build and hold and double build paths had lower closeness values (43 and 28%, respectively). Consequently, it can be
inferred that in the context of ERD wells, it is preferable to carry out the deviation process at deeper depths.
Keywords Multi-objective genetic algorithm · Optimization · Torque and drag · Extended reach drilling · TOPSIS
List of Symbols MOGA Multi-objective genetic algorithm

TOPSIS Technique for order preference by
ERD Extended reach drilling similarity to an ideal solution
TVD (ft) True vertical depth
B Jafar Qajar MD (ft) Measure depth
j.qajar@uu.nl; jqajar@shirazu.ac.ir EOB End of build
Hossein Yavari I (degree) Inclination
h.yavari@aut.ac.ir East (ft) Easting
Bernt Sigve Aadnoy North (ft) Northing
bernt.aadnoy@uis.no r (ft) Drill pipe radius
Rasool Khosravanian R (ft) Radius of curvature
rasool.khosravanian@uis.no T (N.ft) Torque on drill string
F (kN) Axial tension load or drag
1 Department of Petroleum Engineering, Amirkabir University
θ (degree) Azimuth
of Technology, Tehran, Iran
B (non-dimensional) Buoyancy factor
2 Department of Earth Sciences, Utrecht University, Utrecht,
β (degree) Total angle change
The Netherlands
w (kN/ft) Weight of unit length
3 Department of Petroleum Engineering, School of Chemical
W (kN) Weight
and Petroleum Engineering, Shiraz University, Shiraz, Iran
N (kN) Normal force
4 Department of Energy and Petroleum Engineering, University
L (ft) Length
of Stavanger, Stavanger, Norway
123
16832 Arabian Journal for Science and Engineering (2023) 48:16831–16855
ppg Pound per gallon of a reservoir where an additional platform is not economi-
μ (non-dimensional) Friction factor cally feasible. ERD is predicted to become more widespread
HD (ft) Horizontal departure as the cost of platforms in deeper water and severe envi-
KOP (ft) Kick off Point ronments increases [3, 6]. Previous research suggests that
DLS (degree/100ft) Dogleg severity the cost of drilling a horizontal well is approximately 1.4
ECD (ppg) Equivalent circulation density times that of drilling a vertical well [8]. The advantage of
MW (ppg) Mud weight drilling horizontal and directional wells is their ability to
P (psi) Frictional pressure drops access a larger volume of the reservoir and traverse the
OD (in) Outside diameter highest quality zones more effectively than vertical wells,
ID (in) Inside diameter leading to higher production and recovery rates [9]. Effi-
M (in4 ) Moment of inertia ciently planning the trajectory of directional wells is crucial
Fcr (lb) Critical buckling force for minimizing drilling expenses and reducing the potential
Fhel (lb) Helical buckling force negative effects of drilling issues. This process presents com-
E (psi) Young’s modulus plex multi-objective optimization challenges [10].
d (in) Spatial distance between the bore and In recent decades, optimization techniques have been
the wall of the borehole. in widely employed in the oil and gas industry for various
Pc Crossover probability purposes such as transport scheduling, process plant opti-
NDM Non-dimensional matrix mization, well placement optimization, and different aspects
V Weighted decision matrix of drilling operations [11–15]. Shokir et al. [11] used the
PIS Positive ideal solution genetic algorithm to design the well path, and Atashnezhad
NIS Negative ideal solution et al. [12] used a single-objective particle swarm optimiza-
S Separation distance tion algorithm to minimize wellbore length within defined
CL Relative closeness constraints. Yasari et al. [16] applied multi-objective genetic
OF Objective function optimization to determine the best and most efficient sys-
χ Normalized membership function tem of water injection wells, and Guria et al. [13] utilized
Nobj Number of objective functions an elitist non-dominated sorting genetic algorithm to min-
SM Spacing metric imize drilling cost, drilling time, and maximize drilling
MS Maximum spread depths with the constraint of fractional drill bit tooth wear.
Mansouri et al. [9] proposed another application of multi-
objective genetic optimization to optimize a horizontal well
trajectory scenario. They provided a detailed description of
1 Introduction how their model functioned and analyzed the results for the
specific wellbore trajectory chosen. Furthermore, Khosra-
The drilling industry’s primary objective of directing a well vanian et al. [17] optimized the casing string placement in
trajectory toward a distant geological target has necessitated oil wells in the presence of geological uncertainty. Khosra-
the development of tools and methods for identifying the vanian et al. [18] conducted a comparative study to assess
location and path of the wellbore during drilling. Initially, multiple metaheuristic algorithms for optimizing complex
wells were drilled straight down from the drilling rig to the three-dimensional well-path designs. Mansouri et al. [14]
target spot. However, directional drilling techniques were conducted an optimization study aimed at mitigating colli-
developed to drill wells with non-vertical trajectories and sion risk in directional drilling by maximizing the separation
reach destinations that were not directly beneath the surface. factor. Their approach sought to optimize the directional well
Directional drilling offers an effective approach to accessing trajectory to ensure minimal overlap with existing wells,
challenging targets that cannot be easily reached by vertical thereby reducing the likelihood of collision and the asso-
drilling. Over time, numerous instruments and methods have ciated risks. Biswas et al. [19] employed objective functions
been developed for directional drilling, with many companies such as true measured depth (TMD), torque, and strain energy
now providing tools for deflecting and guiding wellbores as to assess the effectiveness of wellbore trajectory design. To
well as assessing their inclination and azimuth [1–4]. address the optimization challenges posed by the 17 tuning
The challenging nature of harsh environments presents variables involved in drilling, they devised a novel approach
significant obstacles in the drilling of wells, requiring a that combined the cellular automata (CA) technique with
high degree of technical expertise and specialized equipment the gray wolf optimization (GWO) and particle swarm opti-
to overcome [5]. One application of directional drilling is mization (PSO) algorithms. This hybridization successfully
extended reach drilling (ERD) [6]. ERD is utilized in devel- resolved the optimization objectives associated with drilling
oping reservoirs with fewer platforms [7] or smaller sections
123
Arabian Journal for Science and Engineering (2023) 48:16831–16855 16833
and provided a comprehensive solution for achieving opti- The problem of torque and drag, which results from the
mal results. Huang et al. [20] approached the discrepancy friction between the drill string and the borehole wall, is one
between a planned trajectory and the actual trajectory as a of the major challenges of extended reach drilling (ERD)
multi-objective optimization problem (MOP) characterized [26]. The design of the well has a significant impact on the
by parameter uncertainties. To address this challenge, they drilling time, cost, and the torque on the drill string [3, 9].
developed a novel methodology called outlier removal (OR- In the present investigation, a multi-objective genetic algo-
NSGA-II) within the framework of nondominated sorting rithm (MOGA) is employed to minimize both well bore
genetic Algorithm II. This innovative approach effectively length and drill string torque during drilling operations.
resolves the optimization problem by simultaneously opti- Moreover, a fuzzy decision-making approach was utilized to
mizing multiple objectives and removing outliers, thereby select the best solution from Pareto optimal solutions [27,
enhancing the accuracy and reliability of the results. Wendi 28]. The TOPSIS method is then utilized to identify the
Huang et al. [21] introduced an innovative optimization best trajectory based on multiple criteria, such as torque on
method that merges an adaptive penalty function with a the drill string, wellbore length, likelihood of keyseat, and
multi-objective evolutionary algorithm according to decom- number of required tools. The advent of directional drilling
position. This technique effectively tackles the challenges technology, particularly the widespread use of diverse down-
posed by contradictory objectives and nonlinear constraints. hole motor tools, measure while drilling tools (MWD), drag
Wang et al. [22] introduced a cutting-edge optimization tech- reduction oscillators, rotary steerable systems, and agitators,
nology for well trajectory and implemented it in 42 wells. The has greatly accelerated the drilling efficiencies of horizontal
findings of their research demonstrate that this technology is wells, extended reach wells, multi-branch wells, and other
capable of delivering a low-cost and high-efficiency develop- complex-structured wells [29]. However, the implementation
ment of horizontal wells in the Cangdong sag. Furthermore, of these new tools significantly escalates the cost of drilling
the research outcomes hold significant reference value for the operations. Hence, it is advantageous to select a trajectory
development of analogous regions. Biswas et al. [23] intro- that necessitates fewer tools.
duced the “Modified Multi-Objective Cellular Spotted Hyena In this research, the utilization of a genetic algorithm for
Optimizer” as an innovative solution to address uncertainty in optimization purposes is discussed. Previous studies have
well design. Unlike previous models, this approach excels in successfully employed the genetic algorithm to optimize the
identifying isolated minima and exhibits a high convergence trajectory of a well, confirming its effectiveness in determin-
rate, making it particularly effective for solving problems ing the optimal path [9, 11–13, 16, 18, 20]. Additionally,
with multiple variables. Huang et al. [24] conducted an anal- the TOPSIS method was employed to select the optimal tra-
ysis of the challenges involved in drilling extended-reach jectory, taking into account various criteria such as risk. The
wells, using the drilling limit theory as a framework for their flexibility of the TOPSIS method is one of its key advantages,
investigation. Wood [25] conducted research that highlighted making it a highly suitable option for the well design process.
the benefits of utilizing multiple constraints related to differ- During the well design process, different scenarios may arise.
ent Dogleg severity (DLS) characteristics and the absolute In some cases, a wealth of data is available, with numerous
changes in the tool-face angle for optimizing measure depth wells having been drilled in the field. In such cases, risk,
(MD) of Bezier curves. uncertainty, and other criteria can be accurately assessed and
Previous studies have primarily concentrated on a single modeled using complex methodologies. Conversely, there
well trajectory, overlooking the provision of a methodol- are situations where very limited data is accessible, and for
ogy to ascertain the optimal configuration by evaluating example the estimation of risk might rely solely on consid-
various potential drilling trajectories. Furthermore, previ- ering the tension at the Kick-Off Point (KOP), indicating the
ous studies have not introduced an automated method, free likelihood of a keyseat occurrence. The TOPSIS method is
from the necessity of manual settings and parameters, to capable of accommodating both scenarios. It can be seam-
choose the finest solution from the set of Pareto optimal lessly integrated with sophisticated mathematical risk models
solutions. Additionally, the operational aspects, encompass- to derive an optimal solution, or it can simply employ a
ing risk assessment and the requisite drilling tools, have straightforward numerical value to assess parameters like
been disregarded in earlier investigations. Therefore, it is risk. This versatility is highly valuable in well design, as
of significant importance to present a method capable of specific data may not be obtainable for every well. There-
simultaneously examining and comparing distinct trajecto- fore, the genetic algorithm and TOPSIS are highly suitable
ries, automatically identifying the best solution from the and effective approaches for the well design process. The
array of optimal choices, and determining the optimal tra- efficacy of this method for selecting the optimal ERD well
jectory for diverse trajectories while considering both risk path is demonstrated through a case study.
factors and the specialized drilling tools required.
123
use of an agitator and rotary steerable system (RSS) [33–35].

Furthermore, drilling a completely horizontal layer is a chal-
lenging task that requires utmost precision since the drill bit
and drill string tend to deviate under their own weight [32,
36, 37]. Additionally, handling drilling cuttings and cleaning
the well presents additional problems [30, 38–42]. Accord-
ing to Agbaji [30, 43], in horizontal wells, cuttings tend to
accumulate on the lower side of the borehole, creating a
continuous and elongated bed of cuttings. The drilling fluid,
on the other hand, will flow above the drillpipe. Regardless
of the flow rate or viscosity of the mud, mechanical agi-
Fig. 1 ERD well trajectory profiles [30] tation is necessary to displace the cuttings and keep them
in motion. Conversely, a well path with higher torque like
undersection trajectory presents none of these issues [32].
2 Methodology The findings imply that alongside optimization, a flexible
selection approach that comprehensively assesses all the cri-
This approach is based on this idea that apart from an opti- teria is necessary to select the most suitable drilling path.
mization technique to identify the optimal path for drilling The steps of this method are as follows. The first step is
a well, it is essential to employ a selection method that to recognize all conceivable trajectories for drilling a well,
takes into account all the criteria needed to choose the best taking into account the geological features of the forma-
path. The design of the well trajectory is a critical factor in tions and the available resources and technologies. Then, a
the successful operation of ERD wells. The optimal trajec- multi-objective genetic algorithm is utilized to find the best
tory design aims to achieve long horizontal reaches with the well path for each trajectory by minimizing both torque and
least possible limitations, and there are several approaches length. At this step, the optimal parameters for the genetic
available to achieve this goal. As shown in Fig. 1, there are algorithm to solve the problem should be identified through
different trajectories that can be used for designing an ERD sensitivity analysis. Then, fuzzy decision-making method
well [30], each with its own set of advantages and disadvan- was used to choose the best solution. Finally, TOPSIS method
tages. However, it is important to note that the selection of the was used to choose the best trajectory.
well path based on a single criterion is not advisable, as this This approach involves identifying all potential drilling
approach does not take into account the multiple factors that trajectories based on geological features, equipment and
can impact the success of the well. Factors such as geologi- technologies. The risks associated with each trajectory, such
cal features, drilling technology, and operational limitations as the likelihood of keyseat formation, which depends on
and risks can all play a significant role in the selection of the tension at the build section, are evaluated. Additionally,
the optimal well trajectory. To ensure the success of ERD the required tools for each trajectory, including mud motors,
well operations, it is crucial to conduct a thorough evalua- MWDs, agitators, adjustable gauge stabilizers, near-bit subs,
tion of all relevant criteria before selecting the well path. This and rotary steerable system (RSS), are determined. Then,
evaluation should take into consideration the geological char- the optimal trajectory is selected using the TOPSIS method,
acteristics of the reservoir, the drilling technology available, which considers torque, length, risks, and required tools. A
and the operational constraints of the project. Only through a flowchart outlining the method is shown in Fig. 2. Results
comprehensive evaluation of all relevant criteria can the opti- demonstrate that this approach effectively identifies the opti-
mal well trajectory be selected, leading to successful ERD mal well path and assesses the advantages and disadvantages
well operations. of each trajectory compared to others. The comparison of
Previous research efforts have employed single-objective different trajectories provides useful insights for well design
and multi-objective optimization techniques to optimize a and predicting problems during drilling.
drilling trajectory of oil and gas wells [9, 11, 12, 18, 19, 21, This study focuses on the drilling of ERD wells with a
31]. However, in some trajectories drilling the trajectory that vertical depth of 5000 feet and a horizontal deviation of
is optimal may come with some challenges and difficulties, 12,000 feet. These types of wells are commonly used, but
including the need for some tools that inevitably increase drilling them can be very challenging, especially in the 8.5-
drilling costs [32]. For instance, this study demonstrated that inch hole, due to the high torque during rotary drilling [24,
drilling an ERD well with a multiple build profile results in 44]. Previous surveys conducted by El Sabeh et al. [44] and
the lowest torque. However, this profile also has the highest Huang et al. [24] have shown that ERD wells with a verti-
likelihood of drilling pipe buckling, mainly due to high fric- cal depth of less than 6,500 feet, and even as low as 4,000
tion levels in the horizontal section, which necessitates the
123
gas injection, and water injection wells. The Well Construc-

tion Team evaluates the drillability of ERD wells using 3D
geomechanical models. This study focuses on using ERD
wells with the same common dimensions as a case study.
The study results reveal the pros and cons of different drilling
path types for these wells, compare the various paths in terms
of their torque and length, and assess the potential of each
path to become the optimal path.
3 Mathematical Modeling
3.1 Wellbore Length
The radius of curvature technique is utilized to determine the

Fig. 2 Flowchart of the proposed method
well path trajectory, where the curvature and measured depth
between two points are computed using specific equations
[47]. Additionally, the different angles and elements of the
wellbore trajectory are defined using the terms outlined in
feet, are difficult to drill. They showed that 42% of the inves-
Fig. 4 [47, 48].
tigated wells had a vertical depth of less than 6,500 feet,
To determine the radius of curvature, the MD and the TVD,
and Huang et al. (2022) studied a well with a vertical depth
Eqs. (1)–(3) are used, respectively[19, 23, 47, 48].
of 4,000 feet, indicating that ERD wells with true vertical
depth (TVD) equal to 4000 feet are common but difficult to 18000
drill. Additionally, this study focuses on the Yuri Korchagin R (1)
π .DLS
field, an offshore oil and gas field in the north sector of the
Caspian Sea, located 240 km from Makhachkala and 180 km
(I1 + I2 )
from Astrakhan. The field was discovered in 2000, devel- MD R (θ2 − θ1 ) 2
(sin4 + (I2 − I1 )2 ) (2)
2
oped in 2004, and began production on April 28, 2010, with
an expected annual production of 1.2 billion cubic meters of 180 ∗ MD(sin(I2 ) − sin(I1 ))
TVD (3)
natural gas and 2.5 million tons of crude oil. The productive π ∗ (I2 − I1 )
formations are Volzhski and Neokomian, and the reservoir
consists of a 70-m gas cap and a 20-m oil rim [45]. Figure 3 where R is the radius of curvature, DLS is the dogleg sever-
illustrates the Yuri Korchagin field map and the drilled ERD ity, θ is the azimuth, and I is the inclination angle. The total
wells in the field. length of the wellbore is calculated by summing up the mea-
According to Sergey Bogdanov et al. [45, 46], there are 30 sured depths from different segments of the well, as shown
development ERD wells that have a step-out of up to 19,000 in Eq. (4).
feet and a vertical depth of 5,118 feet, including production,
Fig. 3 Yuri Korchagin field map [45]
123
Fig. 4 Calculation of the length

for a deviated section of the well
trajectory [20, 48]
MDtotal MDKOP + MDBuildSection + MDHoldSection (4)
3.2 Torque on the Drill String During Drilling
The soft string model is utilized to determine the torque on

a drill string and serves as the basis for the torque objective
function. The soft string model operates under the assump-
tion that the drill string resembles a heavy cable resting inside
the wellbore, and any stiffness impacts from the drill pipes
are disregarded. The calculations only account for the state
in which the drill string rotates and not any axial movement Fig. 5 Force balance in a pipe along a straight section [20, 49, 50]
up or down. Two general cases are addressed with formulas
provided [49].
section. In curved sections, the normal force between the
drill string and the borehole is heavily influenced by the axial
1. Straight section
loading within the pipe. Figure 6 illustrates the force balance
2. Curved section
for a pipe in a curved section.
Equations (7) and (8) are utilized to compute the torque
Equations (5) and (6) are used to calculate the torque and
and tension for the curved section of the well.
tension, respectively, for the straight section of the well.

sin(I2 ) − sin(I1 )
F2 F1 + BwLcos(I ) (5) F2 F1 + BwL (7)
I2 − I1
T μr BwLsin(I ) (6) T μr F1 β (8)
where F is the tension or drag force, T is the torque, B is cos(β) sin(I1 )sin(I2 )cos(θ1 − θ2 ) + cos(I1 )cos(I2 ) (9)
the buoyancy factor, w is the weight per unit length, L
is the length of the element, μ is the friction coefficient, Beta (β) is total directional change [49].
and r is the pipe radius. The normal force and friction on a Another objective function in well drilling is the torque on
drill pipe are not affected by the tension along it in a straight the drill string, denoted by T, which is calculated by summing
section. Straight sections are mainly dominated by weight, as up the torque in different parts of the wellbore as shown in
only the normal weight component creates friction. Figure 5 Eq. (10). The torque calculation begins at the bottom of the
demonstrates the force balance in a pipe along a straight drill string, specifically when it reaches the total depth (TD)
123
finally, the torque is calculated. The mathematical formulas

from Eqs. (11)–(14) provide the calculation for the measured
depth in different sections of ERD wells [48, 49, 51].
18000
R1 (11)
π .DLS1

(0 + I 1 )
MDBuildSection R1 (θ2 − θ1 )2 (sin4 + (I1 − 0)2 )
2
(12)
MDHoldSection (TVDAG − TVDAB

sin(I1 ) − sin(0)
− (MDBuildSection ))/cos(I1 ) (13)
I1 − 0
Fig. 6 Force balance for a pipe along a curved section [19, 23, 49]
MDtotal TVDAB + MDBuildSection + MDHoldSection (14)
Equations (15)–(18) provide a mathematical model for

calculating the true vertical depth of ERD wells.
TVDAB KOP (15)
MDBuildSection (sin(I1 ) − sin(0))

TVDAC−AB (16)
(I1 − 0)
TVDAG−AC MDHoldSection cos(I1 ) (17)
TVDtotal TVDAB + TVDAC−AB + TVDAG−AC (18)
Equations (19)–(22) provide mathematical models for

estimating the tension in different parts of ERD wells.
Fig. 7 Different parts of well profile [51] FD 0 (19)
of the wellbore, and then moves upward in increments toward FC FD + BwMDHoldSection Cos(I2 ) (20)
the wellhead, starting from the horizontal position.
sin(I2 ) − sin(0)
FB F C + BwMDBuildSection (21)
Ttotal TKOP + TBuildSection + THoldSection (10) I2 − 0
FA F B + BwTVDAB Cos(0) (22)

3.3 Formulation of the Objective Functions
for Optimization The torque on the drill string for various segments of ERD
wells can be determined using Eqs. (23)–(26).
The functions for calculating the length of a wellbore and the
TCD μr BwMDHoldSection Sin(I2 ) (23)
torque on a drill string in an extended reach drilling (ERD)
well are presented below. The well is divided into various
parts such as the Kick-Off Point (KOP), build section, and TBC μr FC β&β I2 (24)
hold section, as illustrated in Fig. 7.
To model the trajectory of a well, it is essential to follow TAB 0 (25)
a specific order of equations. The first step is to calculate
the measured depth, which is then used to calculate the true
vertical depth of the well. The tension force is then deter- Ttotal TAB + TBC + TCD (26)
mined based on the measured and true vertical depths, and
123
Fig. 8 Buckling of pipe,

a Sinusoidal buckling, b Helical
buckling [53]
3.4 Formulation of the Constraints for Optimization in the heel area. Therefore, it is essential to ensure that the
equivalent density remains lower than the fracture pressure
Constraints play a vital role in the field of optimization. of the well. This concern is particularly acute for ERD wells
During the optimization process, it is crucial to ensure that with limited vertical depth. Equation (29) is used to calculate
the variables, which are often referred to as chromosomes, ECD [30, 47].
adhere to the specified permissible limits. The evaluation
of these variables using objective functions is contingent P
ECDheel MW + (29)
upon their compliance with the permissible limits as well 0.052 ∗ Length
as satisfying all the imposed restrictions. For example, when
the crossover operator is applied to a given dataset, gener-
ating new variables, only those parameters that fall within 3.4.3 Buckling of Drill Pipes
the allowed limits and fulfill all the constraints are accepted.
Subsequently, these accepted variables are assessed using Buckling is a phenomenon that occurs when a drilling pipe
the objective functions. The MOGA verifies constraints at is subjected to an increase in load, causing it to bend or twist.
three different points: once the initial population is formed, The minimum amount of load required to cause buckling
after applying the crossover operation, and after applying the is known as the buckling load, which is dependent on the
mutation operation. drilling pipe’s weight, properties, and the geometry of the
drilling trajectory section. As shown in Fig. 8, there are two
3.4.1 Maximum Torque and Tension types of buckling load: critical buckling load (Fcr ) and heli-
cal buckling load (Fhel ). To determine whether sinusoidal or
During the drilling process, it is imperative to maintain the helical buckling is occurring, the buckling load is compared
torque and tension levels within permissible limits, while to other loads such as drag and torque at the point of the
also accounting for the requisite safety margins. In order to drilling trajectory section [53–55].
accomplish this, it is vital to follow the Eqs. (27)–(28) [47, The types of drilling pipe bending are determined by the
52], which present the relevant relationships. degree of deformation. Sinusoidal buckling and helical buck-
ling are two common types of deformation in drilling pipes.
Torsional Resistence Sinusoidal buckling can be seen as a wave-like pattern in
Ttotal < (27) the borehole, as shown in Fig. 8a. As the pressure on the
Safety Factor
drilling pipe increases, it may undergo severe deformation,
Yield Strength leading to helical buckling, as illustrated in Fig. 8b. Heli-
Ftotal < (28)
Safety Factor cal buckling occurs when the pipe spirals into a spiral shape
and makes contact with the borehole wall. While sinusoidal
buckling has little impact on drilling activities, helical buck-
3.4.2 Equivalent Circulation Density (ECD) ling makes it difficult to control the orientation of the drilling
pipe and to effectively transfer load to the drilling pipe. When
Extended reach drilling (ERD) wells, especially those with the degree of deflection is significant, the drilling pipe may
high inclination angles and long horizontal segments, are become immobile or lock-up may occur, resulting in drilling
susceptible to significant increases in equivalent circulating delays and increased costs. Therefore, it is critical to avoid
density (ECD), which can potentially cause formation failure helical buckling in directional drilling, such as horizontal
123
between the bore and the wall of the borehole (d), inclina-
tion angle (I ), and the effective weight within the drilling
trajectory (We ). Moment of inertia for drill pipes is calcu-
lated using Eq. (38) [31, 47].

π O D4 − I D4
Fig. 9 Three-dimensional schematic of sinusoidal and helical buckling
M (38)
64
[37]
where OD is outside diameter (in) and ID is inside diameter
drilling, as it can cause significant damage [53–55]. Figure 9 (in) of pipe.
shows a three-dimensional schematic of sinusoidal and heli- If the axial load on a structure is less than the critical
cal buckling of drill pipe in wellbore. buckling load, no buckling will occur. When the axial load
In this study, the model for buckling load is categorized is between the critical buckling load and the helical buckling
into two types: critical buckling load and helical buckling load, the structure will buckle sinusoidally. However, if the
load. This classification is determined based on the shape of axial load exceeds the helical buckling load, the structure will
the drilling trajectory section [33, 36]. buckle in a helical pattern [31].
Sinusoidal buckling force values of different sections are
given in Eqs. (30)–(33) [31, 53].
4 Multi-Objective Genetic Algorithm (MOGA)
Fcr (vertical Section) 2.55
3
EMW2e (30)
John Holland introduced genetic algorithm in the early 1970s
⎛ ⎞ [56, 57], and subsequent work by Holland (1975/1992),
4EM ⎝ d R 2 We sin(I ) ⎠ Goldberg, Rawlins, and Whitley laid the foundations for
Fcr (build Section) 1+ 1+
Rd 4EM Gas [58–60]. To begin, the GA generates an initial set of
random solutions that comply with defined limitations and
(31)
constraints, known as the population. In each iteration, the
EMWe sin(I ) GA creates a set of solutions, with the population referring to
Fcr (inclined Section) 2 (32) the set of solutions and each solution being called a chromo-
d
some [61]. These chromosomes evolve through generations,
EMWe with the creation of the next generation, known as offspring,
Fcr (horizontal Section) 2 (33)
d accomplished by either merging two chromosomes from the
current generation using a crossover operator or modifying
Helical buckling force values of different sections are as a chromosome using a mutation operator. Chromosomes are
shown in Eqs. (34)–(37) [31, 53]. evaluated during each generation using measures of fitness,
so that fitter chromosomes are more likely to be selected.
Fhel (vertical Section) 5.55 EMW2e
3
(34) After several iterations, the algorithm converges to the best
solution or chromosome, which represents the optimal or
⎛ ⎞ suboptimal solution to the problem. Murata et al. [62, 63]
12EM ⎝ d R 2 We sin(I ) ⎠ proposed MOGA, which comprises eight steps in each gener-
Fhel (build Section) 1+ 1+
Rd 8EM ation. This MOGA utilizes a weighted sum of multi-objective
(35) functions to amalgamate them into a scalar fitness function.
√ EMWe sin(I )
Fhel (inclined Section) 2 2 2 − 1 (36) 4.1 Step 1: Initialization
d
√ Generate an initial set of individuals that satisfy the given
EMWe
Fhel (horizontal Section) 2 2 2 − 1 (37) constraints.
d
Equations (30)–(37) provide a comprehensive overview of 4.2 Step 2: Evaluation

various parameters and their definitions used in the buckling
load model applied in this study. These parameters include The next step involves computing the objective functions for
the Young’s modulus (E), moment of inertia (M), radius each chromosome and normalizing them, as different objec-
of curvature of the drilling trajectory (R), spatial distance tive functions may have different units. To do so, Eq. (39) is
123
Fig. 11 Mutation mechanism of MOGA [62, 63]
4.5 Step 5: Mutation

Fig. 10 Crossover mechanism of MOGA [62, 63]
In this stage, the mutation operation is implemented, as
shown in Fig. 11, to introduce genetic diversity into the pop-
utilized to normalize each objective function. ulation.
f i − f imin 4.6 Step 6: Elitist Strategy

fi , i 1, 2, . . . , n (39)
fimax − f imin
The top-performing chromosome is conserved and carried
over to the subsequent generation.
4.3 Step 3: Selection
4.7 Step 7: Termination Time
The fitness value of each chromosome is determined by
applying Eq. (40). Steps 2 to 6 are iterated until the maximum number of itera-
tions is attained.
f (x) w1 f 1 (x) + w2 f 2 (x) + ... + wn f n (x) (40)
4.8 Step 8: Final Set of Pareto Optimal Solutions
The combined fitness function f(x) is obtained by mul-
tiplying each objective function fi(x) by a constant weight The MOGA provides a range of Pareto optimal solutions
w, summing up the weighted objective functions for a given from which the optimal one must be chosen.
chromosome x, and then dividing by the number of objective Several researchers have employed both single-objective
functions n, as shown in Eq. (41). and multi-objective meta-heuristic algorithms for different
purposes, such as finding the best neural network training,
randi identifying optimal coefficients for mathematical models,
wi n , i 1, 2, ..., n (41) and optimizing various objective functions [64–67].
j1 randj
where randi and randj are non-negative random numbers.

Two chromosomes are chosen using selection probability
5 TOPSIS
which is illustrated by Eq. (42).
TOPSIS, a multi-criteria decision analysis method, was ini-
f (x) − f min (ψ) tially introduced by Yoon [68] and further developed by
P(x) (42) Hwang and Yoon [69]. Subsequent enhancements were made
x∈ψ { f (x) − f min (ψ)} by Yoon [68] and Hwang, Lai, and Liu [70]. This approach
is commonly utilized in a variety of fields, including sup-
where f min (ψ) is as determined using Eq. (43). plier selection, financial performance evaluation, company
evaluation, tourism destination evaluation, selecting the most
f min (ψ) min{ f (x)|x ∈ ψ} (43) suitable machine, and location selection [71]. TOPSIS is
based on the idea that the selected option should have the
shortest distance from the positive ideal solution (PIS) and the
4.4 Step 4: Crossover longest distance from the negative ideal solution (NIS) [72].
To apply TOPSIS, a decision matrix (DM) is first constructed,
In this step, the genetic algorithm applies the crossover oper- and the criteria are identified. The DM is then normalized,
ation, which is illustrated in Fig. 10. and the weights for each criterion are determined. Next, a
123
weighted decision matrix is calculated, and the positive and 5.5 Step 5: Separation Distance
negative ideal solutions are determined. Finally, the distances
between each alternative and the ideal alternatives are calcu- The separation distances (S) are calculated using Eqs.
lated, and the alternative with the minimum distance to the (46)–(47).
PIS and the maximum distance to the NIS is selected as the
best alternative. In summary, TOPSIS is a robust and widely n
Si+ (PIS j − Vi j )2 (46)
used decision-making tool that can assist decision-makers in j1
evaluating alternatives in a multi-criterion setting. It provides n
a systematic and structured approach to decision-making, Si− (NIS j − Vi j )2 (47)
j1
which can lead to more informed and rational decisions [73,
74]. TOPSIS method involves the following steps:
5.6 Step 6: Relative Closeness

5.1 Step 1: Decision Matrix
In accordance with Eq. (48), the determination of the relative
In the TOPSIS method, the initial step involves creating closeness (CL) to the ideal solution is executed.
a decision matrix that includes all relevant factors along
with their respective quantities, which indicate their level Si+
of importance. It is important to note that for some factors, CLi (48)
Si+ + Si−
such as the length of a well in well design, a longer length is
considered to be unfavorable and therefore a negative param-
eter. Consequently, a shorter length is considered to be more
favorable. Negating the value or calculating the inverse of
6 Case Study
such parameters can help avoid errors in the decision matrix
This study involves designing an ERD well with a vertical
[74, 75].
depth of 5000 ft (1524 m) by considering four trajectories.
It is essential to take into account specific limitations while
5.2 Step 2: Normalized Decision Matrix designing each trajectory. The total vertical depth (TVD) is
constant in all trajectories. The limitations are based on the
The subsequent stage involves creating a normalized decision design characteristics of each well path, which are detailed
matrix (NDM), which is accomplished by utilizing Eq. (44). in the subsequent tables.
In this study, the following assumptions are considered:
Xi j
NDM Ri j (44)
m 2 1. The drill string has 0.167 ft radius.
i1 X i j
2. The drill string has 0.069 KN/ft weight.
3. The friction factor is 0.2
4. The buoyancy factor is 0.85
5.3 Step 3: Weighted Decision Matrix 5. Drilling fluid density is 9.8 ppg
The TOPSIS method assigns varying weights to each selec- The drilling equipment utilized in this study was Class
tion criterion based on its degree of importance. The weight N drill pipe, which conforms to the API RP 7G standard,
matrix is multiplied by the normal decision matrix in accor- 16th edition, August 1998, with a nominal weight of 15.7
dance with Eq. (45) to produce the weighted decision matrix. pounds force per foot. This drill pipe has a tensile yield
strength of 2593 kilo newtons and a torsional yield strength
V W j ∗ NDMi j (45) of 206.5 kN.ft.
The primary objective of this case study is to illustrate the
efficacy of the novel method for identifying the optimal well
path that minimizes torque on the drill string and wellbore
5.4 Step 4: Positive and Negative Ideal Solution length while taking into account several different criteria,
such as risk and the number of tools required. The outcomes
Positive ideal solution (PIS) and negative ideal solution (NIS) obtained via this novel method are heavily influenced by the
are defined as follows: weights allocated to each criterion. For instance, in one com-
PIS Minimum of V in each column. pany, the well’s length may be less important compared to
NIS Maximum V in each column. the torque and the likelihood of a keyseat, while in another
123
Table 1 Limitations of double build trajectory during the design of the Table 3 Undersection trajectory constraints
well path
Variable Minimum Maximum
HD GD 11,500 12,500
HD GF 11,500 12,500 TVD TVDAG 5000 5000
TVDAF 5000 5000 KOP TVDAB 2500 3000
KOP (1) TVDAB 1450 1550 EOB TVDAC 3650 3850
KOP (2) TVDAD 3450 3550 Inclination 20 85.5
EOB1 TVDAC 2150 2350 DLS 0.5 5
EOB2 TVDAE 3800 4000
Inclination (1) 10 40
Inclination (2) 50 85 Table 4 Multiple build trajectory constraints
DLS (1) 0.5 5
DLS (2) 0.5 10
HD GD 11,500 12,500
TVD TVDAG 5000 5000
KOP TVDAB 4000 4500
Table 2 Build and hold trajectory constraints
EOB TVDAC 4900 5000
Variable Minimum Maximum Inclination 89 90
DLS 0.5 10
HD GD 11,500 12,500
TVD TVDAG 5000 5000
KOP TVDAB 1450 1550
6.3 Undersection Trajectory Limitations
EOB TVDAC 2500 2700
Inclination 20 89 As shown in Fig. 14, the undersection trajectory profile is
DLS 0.5 5 presented below.
This study employed a set of constraints for the under-
section trajectory in designing the well path. The specific
constraints used are presented in Table 3.
company, both factors could be equally significant, leading to
varying outcomes. Furthermore, constraints on each trajec-
6.4 Multiple Build Trajectory Limitations
tory vary depending on the region, and an accurate geological
model is required to determine these constraints. In this study,
The multiple build trajectory profile is shown in Fig. 15.
the permissible limits for each trajectory have been consid-
The constraints applied in the well trajectory design of
ered hypothetically and according to the characteristics of
multiple build trajectory are presented in Table 4, as shown
that trajectory.
in Fig. 15.
6.5 Sensitivity Analysis

6.1 Double Build Trajectory Limitations
At the outset, a sensitivity analysis was carried out to iden-
The profile of the double build trajectory is shown in Fig. 12. tify the optimal parameters of the genetic algorithm that
The constraints for the double build trajectory used in this would yield accurate results in a shorter period. The anal-
study are presented in Table 1, which was taken into consid- ysis involved evaluating the impact of crossover probability,
eration during the design of the well trajectory. mutation rate, and population size on the algorithm’s perfor-
mance. The study revealed that increasing these parameters
may not necessarily lead to improved algorithm performance.
6.2 Build and Hold Trajectory Limitations
6.5.1 Crossover Probability
The profile of the build and hold trajectory is shown in Fig. 13.
The constraints applied to the well trajectory design for In general, setting a higher value for the crossover probability
the build and hold trajectory are presented in Table 2. (Pc) tends to lead to improved algorithm performance and a
123
Fig. 12 Double build profile [30]
Fig. 13 Build and hold profile

[30]
Fig. 14 Undersection profile [30]
shorter running time. However, setting it too high may restrict 6.5.2 Mutation Rate
the algorithm from exploring the entire solution space and
instead get stuck in a local optimum point. Figure 16 illus- In this study, another parameter that was examined through
trates that increasing the crossover probability enhances the sensitivity analysis is the mutation rate (Mu). Figure 17
algorithm’s performance, but increasing it beyond a certain demonstrates that an increase in the mutation rate leads to
threshold (0.9) does not result in optimal solutions. There- a wider exploration of the solution space. However, this also
fore, a value of 0.7 was selected for this study. results in longer processing time. Therefore, the mutation
rate was set to 0.2 to balance the search range and processing
time.
123
Fig. 15 Multiple build profile

[30]
Fig. 16 Genetic algorithm

performance in different
crossover probabilities
6.5.3 Population Size the drill string and the length of the wellbore, while also
taking into account unique constraints associated with each
The study also looked into the impact of population size on trajectory. To obtain the optimal parameters for this multi-
the genetic algorithm’s performance. While a larger popula- objective optimization problem, a sensitivity analysis was
tion can yield better results, it also increases the computa- carried out using a genetic algorithm. The analysis allowed
tional time. Therefore, a population size of 100 was chosen for the identification of optimal parameter settings that can
as a reasonable compromise. Figure 18 illustrates the perfor- deliver accurate results in a shorter time. The obtained opti-
mance of the genetic algorithm across different populations. mal parameters for the algorithm are summarized in Table
5.
Furthermore, the MOGA’s effectiveness under various
7 Results and Discussion
parameters was assessed using statistical measures, includ-
ing spacing metric (SM) and maximum spread (MS). The
As discussed earlier, the optimization of the well path
involves two objective functions, minimizing the torque on
123

mutation rates

populations
Table 5 Optimum parameters of the algorithm spacing metric, which is a parameter for measuring distribu-
tion based on distance, determines how closely spaced the
Algorithm parameter Optimal value
solutions are. A smaller value of the spacing metric indicates
Crossover probability 0.7 that the solutions are closely positioned, indicating a better
Mutation rate 0.2
distribution of non-dominated solutions. The spacing metric
can be computed using Eq. (49) [19].
Generations 100
123
The multi-objective function is minimized during the opti-

|P|
1 2 mization process. The MOGA starts with initialization. A
SM di − d (49)
|P − 1| random population of 100 solutions is generated, each of
i1
which meets the restrictions. By entering a loop, the objective
functions for each solution are calculated, and a fitness score
In this context, d represents the average of all di values,
is assigned to each of them depending on the value of these
while P denotes the total number of Pareto optimal solutions
functions. In each generation, the Pareto optimum solutions
obtained. The maximum spread (MS) is another metric used
are allocated to rank one. At least some of the parent solutions
to assess the diversity and coverage of the solution space. It
are chosen for the following generation. A crossover operator
represents the greatest distance between the boundary solu-
then adjusts the selected parent solutions and generates new
tions. The calculation for MS is as follows in Eq. (50) [19].
solutions. The crossover operator is intended to seek for new
solutions that may be more fit than some of the parents. The
P
new enhanced solutions then replace alternatives with low
MS max(d(ai , bi )) (50)
fitness scores that were previously ranked low. Once low-
i1
ranking solutions are eliminated, only the most promising
In this context, ai represents the maximum value, and bi options remain. Next, the population undergoes a mutation
represents the minimum value within the i th objective [19]. operator that generates a fresh set of potential solutions for
MOGA was executed multiple times using different the next generation. The mutation operator is applied to the
parameters, totaling 10 runs. The SM values of the Pareto parents, and the new solutions are accepted or rejected based
optimal solutions are presented in Fig. 19 as a box plot, while on their fitness relative to the parent fitness. The mutation
Fig. 20 displays the box plot of the MS values. Notably, the step marks the final stage of each iteration. In the analy-
configuration with a crossover probability of 0.7 and a muta- sis conducted for this study, the process sequence or loop
tion rate of 0.2 exhibited the lowest average SM. In terms was repeated two thousand times. The aim is to minimize
of the MS values, the setup featuring a crossover probability both torque and length objective functions concurrently, but
of 0.9 and a mutation rate of 0.2 achieved an MS of 581.5, it is acknowledged that finding a single solution that meets
whereas the configuration with a crossover probability of 0.7 both objectives is improbable. Therefore, a collection of solu-
and a mutation rate of 0.2 had an MS of 551.58, both of tions with equivalent mathematical values is produced. The
which surpassed the others. Figure 21 provides an illustra- Pareto set for any two objective functions can be presented
tive instance of the optimal solution for MOGA employing graphically as a Pareto frontier which is known as Pareto opti-
different parameters. It is evident that the solution generated mal solutions. Table 6 presents the Pareto frontier for double
by the configuration with a crossover probability of 0.7 and a build trajectory. It is typically observed that Pareto optimum
mutation rate of 0.2 is closer to the minimum values. This cor- solutions do not repeat across different iterations. However,
respondence between the parameter values and the minimum when the algorithm parameters are optimally chosen, the
values, as shown in Figs. 16 and 17, was also confirmed. boundaries representing the maximum, minimum, average
As shown in Figs. 16, 17, and 18, the algorithm con- values, median, as well as the solution identified through the
sistently progresses toward finding the minimum value. fuzzy decision-making method, exhibit a remarkable degree
Through the utilization of crossover and mutation opera- of proximity to one another.
tors, the algorithm is able to explore the entire search space Pareto frontier of double build trajectory is shown in
within the permitted limits over a span of 2000 iterations. Fig. 22. As can be seen, by reducing one of the objective
The outcomes exhibited in Fig. 21 demonstrate that even with functions, the other function increases, but there is no mean-
varying parameters, the algorithm yields values that exhibit ingful relationship between the answers; in other words, for
remarkably similar minimum, maximum, and average char- example, the answer that has the shortest length may not have
acteristics, with the fuzzy decision-making method’s chosen the highest torque.
solution being closely aligned with each other. Subsequently, The multi-objective optimization method generates a set
upon selecting the optimal parameters for the algorithm, the of non-dominant Pareto solutions, necessitating a decision-
solution with the lowest objective function values, represent- making methodology to determine the optimum Pareto
ing the optimal solution, is attained. solution. A fuzzy decision-making technique is used in this
Taking into account the objective functions and constraints work to select the best solution (tradeoff) from a set of non-
outlined earlier, the optimization of the well trajectory for dominant solutions on the Pareto front. Equation (51) is used
four ERD profiles was performed. to compute the fuzzy membership value of the jth objective
123
Fig. 19 Influence of crossover probability and mutation rate on the spacing metric of the MOGA
Fig. 20 Impact of crossover probability and mutation rate on the maximum spread metric of the MOGA
123
Fig. 21 Impact of varying the

crossover probability and
mutation rate on the optimal
solution produced by the MOGA
Table 6 MOGA Pareto optimum

solutions for the double build Pareto front Torque Length KOP1 KOP2 I1 I2 DLS1 DLS2
trajectory
1 25,008.96 14,406.87 1459.36 3455.12 36.04 83.95 3.78 4.29
2 25,157.20 14,375.76 1481.69 3467.71 37.97 83.82 4.06 4.22
3 25,906.24 14,291.49 1453.71 3455.93 39.11 83.26 4.86 4.82
4 25,639.21 14,290.92 1458.64 3453.64 39.93 83.44 4.83 4.22
5 25,048.83 14,550.48 1494.84 3457.86 29.87 84.13 3.34 5.33
6 25,044.58 14,577.44 1502.86 3457.41 29.67 84.19 3.34 5.27
7 25,128.10 14,387.26 1465.25 3460.66 37.25 83.87 3.99 4.22
Minimum 25,032.04 14,290.92 1453.71 3453.64 29.67 83.26 3.34 4.22
Maximum 25,906.24 14,577.44 1502.86 3467.71 39.93 84.19 4.86 5.33
Average 25,279.46 14,412.58 1473.76 3458.33 35.69 83.81 4.03 4.62
Median 25,128.10 14,387.26 1465.25 3457.41 37.25 83.87 3.99 4.29
Fig. 22 Pareto optimum solution

of double build trajectory. The
diamond-shaped point shows the
final solution (optimal trade-off)
obtained through the use of a
fuzzy decision-making system
123
Table 7 Normalized membership function for each solution
Pareto front (k) Torque Length χk
1 25,008.96 14,406.87 0.1782

2 25,157.20 14,375.76 0.1715
3 25,906.24 14,291.49 0.1096
4 25,639.21 14,290.92 0.1434
5 25,048.83 14,550.48 0.1181
6 25,044.58 14,577.44 0.1083
7 25,128.10 14,387.26 0.1707
Minimum 25,032.04 14,290.92 0.1083
Maximum 25,906.24 14,577.44 0.1782 Fig. 23 Optimum torque of different trajectories
function [27, 28, 76].

⎧
⎪
⎪ 1 for OF j < OFMin
⎪
⎨ j
j −OF j
OFMax
χj OFMin < OF j < OFMax (51)
⎪
⎪ j −OF j
OFMax Min j j
⎪
⎩ 0 for OF j > OFMax
j
where OFMin
j and OFMax
j are the objective function’s mini-
mum and maximum fitness values, respectively. The function
that measures the degree of membership, χ k , for each non-
dominant solution is mathematically expressed as Eq. (52)
[27, 77].
Fig. 24 Optimum KOP in different trajectories
Nobj
i1 χ j
k
χ k
Nobj
(52)
ND
k1 j1 χ kj had the highest torque and the lowest wellbore length. These
findings are presented in Table 9.
Equation 50 defines the normalized membership function Based on the results presented in Fig. 23, it can be con-
χ k for each non-dominant solution. The symbol ND refers cluded that the trajectory with multiple build has the lowest
to the number of non-dominated solutions, while Nobj refers torque, followed by the undersection, double build, and build
to the number of objective functions, which are torque and and hold trajectories, respectively.
length in this case. The optimal solution can be determined As shown in Fig. 24, the trajectory with multiple builds
by selecting the normalized membership function with the has the highest KOP followed by undersection, double build,
highest value of χ k in this process. Table 7 contains the for and build and hold, respectively. Therefore, increasing KOP
Pareto optimal solutions of double build trajectory. leads to a decrease in torque. Accordingly, it can be inferred
As explained earlier, all these answers have equal math- that it is more advantageous to deviate the well in the lower
ematical value. The solution in which both functions are at sections of the wellbore.
their lowest state on average was chosen. As shown in the Figure 25 illustrates that the multiple build trajectory has
table, answer number one has the maximum χ k ; therefore, the longest wellbore length compared to the undersection,
this solution was chosen as the optimal state [27, 78]. This double build, and build and hold trajectories, which have the
process was performed for all trajectories, and the optimal second, third, and fourth longest wellbore lengths, respec-
values of each well path were calculated. Results are pre- tively.
sented in Table 8. The outcomes of the optimization process Based on the findings presented in Fig. 26, it can be
are presented in Table 8, along with the optimization trends. inferred that as the KOP increases, the torque decreases
The optimization results for the four trajectories studied regardless of the well profile. However, it should be noted
here were compared based on the minimum length and torque that this increase in KOP leads to an increase in MD (mea-
obtained. Multiple build trajectory had the lowest torque and sured depth), resulting in longer drilling time and higher
the highest wellbore length, while build and hold trajectory costs, which should be taken into account during well design.
123
Table 8 Results of optimization

for four trajectories Trajectory Torque Length KOP1 KOP2 I1 I2 DLS1 DLS2
Build and hold 26,693.08 13,932.02 1462.16 77.64 4.53

Double build 25,008.96 14,406.87 1459.36 3455.12 36.04 83.95 3.78 4.29
Undersection 23,565.31 14,837.92 2541.67 83.57 4.4
Multiple build 21,000.32 16,034.1 4002.3 89.57 6.24
Table 9 Comparative results of

optimization on four trajectories Objective function Build and hold Double build Undersection Multiple build
Torque (N.ft) 26,693.08 25,008.96 23,565.31 21,000.32

Wellbore length (ft) 13,932.02 14,406.87 14,837.92 16,034.11
agitator and other special equipment in these wells, which

increases the cost. The hold section in the undersection tra-
jectory is not horizontal, which can make drilling easier. In
the double build trajectory, the possibility of key seat and
fatigue failure decreases compared to the build and hold tra-
jectory because of the low DLS in the first KOP. The results
indicate that the ultimate tension in all trajectories is equal,
which is demonstrated in Fig. 27. Therefore, drag amount
cannot be regarded as a design objective function in this
study. However, the tension in different parts of an extended
reach drilling (ERD) well is distinct and may impact the well
quality and drill pipe design during drilling.
Fig. 25 Optimum length in different trajectories
As previously discussed, it is crucial to note that the level
of tension varies across different sections of the trajectories,
particularly at the top of the build section in KOP. This is
significant as the likelihood of keyseat and fatigue failure is
correlated with the tension amount and DLS at KOP. The
tension in different parts of the four trajectories is compared
in Fig. 27. It can be observed from the figure that the build
and hold trajectory has the highest tension in different parts
compared to the other trajectories. The double build trajec-
tory, undersection trajectory, and multiple build trajectory
have progressively lower tension in different parts.
The crucial factor to consider is the tension level at the
KOP. Higher tension at KOP increases the likelihood of key-
seat and fatigue failure. Figure 28 demonstrates the tolerance
Fig. 26 Optimum torque, length and KOP in different trajectories com- level of each part of different trajectories to tension. The anal-
pared to each other ysis indicates that the build and hold profile has the highest
probability of keyseat and fatigue failure, while the multiple
build trajectory and undersection have the lowest probability.
The build and hold trajectory exhibits the highest level Figure 29 reveals that the build and hold trajectory has the
of torque and the lowest MD. Due to the high tension in highest tension at KOP compared to the other trajectories,
the build section, there is a high probability of keyseat and thereby making it the most susceptible to keyseat and fatigue
fatigue failure. Therefore, the build and hold trajectory is not failure. Although the tension at KOP in the double build tra-
recommended for drilling an ERD well. The multiple build jectory is similar to the build and hold trajectory, the double
trajectory, on the other hand, has the minimum torque and build trajectory has a lower inclination and Dogleg severity,
maximum MD with a long horizontal section. Due to forma- resulting in a lower probability of keyseat and fatigue failure.
tion tendency and gravity effect, it is difficult to keep the bit in
a horizontal state, and therefore, it is necessary to use RSS,
123
The multiple build trajectory, on the other hand, has the least account the well angle, as well as accurately determine the
amount of tension at KOP. duration and cost of tool usage. However, even in the absence
The graphical representation in Fig. 29 demonstrates that of such detailed information, the TOPSIS method can still be
unlike tension, torque varies in different parts of the ERD implemented by solely conducting qualitative comparisons
wells. Additionally, the graph indicates that the maximum among trajectories. In this particular case, the paths were
torque in ERD wells is associated with the hold section of the qualitatively assessed based on their requirement for special-
well, highlighting the crucial role of effective hole cleaning ized equipment. Deviation drilling, for instance, necessitates
in such wells. drilling equipment as well as survey tools like measurement
while drilling (MWD) and well deviation equipment such as
whipstock, stabilizer, or motor. Similarly, for multiple build
7.1 Selection of Best Trajectory Using TOPSIS
paths, equipment like agitator and RSS are essential, as high-
lighted by various sources [33–35]. In the case of double build
As previously stated, each trajectory has its own strengths
path, the complexity of the well path calls for the use of an
and weaknesses. Therefore, choosing a well path based on a
agitator to enhance drilling operations [79]. Consequently, a
single criterion is not a recommended practice. The selection
rating of 2 was assigned to the double build and undersec-
of the well path should depend on various criteria, such as
tion trajectories, a rating of 3 for the double build path, and
the torque on the drill string, the length of the wellbore, and
a rating of 4 for the multiple build trajectory to signify the
the risk, particularly the probability of keyseat failure, which
difficulty and challenges associated with each path. Huang
is determined by the tension at the KOP in each trajectory.
et al. (2022) have also presented a method to calculate the
The TOPSIS method exhibits a notable attribute of high flex-
difficulty of each ERD well path [24]. If the relevant infor-
ibility. For instance, when formation properties and drilling
mation is available, this difficulty calculation method can
speed data are accessible, it becomes feasible to accurately
be used with the TOPSIS method to analyze different paths
calculate the drilling speed for each trajectory by taking into
Fig. 27 Tension in different parts

of four trajectories
Fig. 28 Amount of tension

tolerated by each part of different
trajectories
123
Fig. 29 Amount of torque

tolerated by each part of different
trajectories
Fig. 30 Selection criteria of

TOPSIS
The results based on the TOPSIS method indicate the rel-

ative closeness of ERD well trajectories to the ideal state, as
shown in Fig. 28. The undersection trajectory has the high-
est relative closeness, followed by multiple build, whereas
the double build trajectory has the lowest relative closeness.
However, it is important to note that practical considerations
such as wellbore distortions and deviations must be taken
into account during drilling operations. Additionally, drilling
several undersection wells from the same platform may not
always be feasible due to well collision concerns. Therefore,
it is recommended to maintain the well as vertical as possible,
similar to the undersection trajectory, until the lower depths,
Fig. 31 Relative closeness of each well path and only then make slight deviations. Alternatively, drilling
a single mother well from the platform to the bottom vertical
in terms of their difficulty. In this study, the risk associated depth and branching off the remaining wells from it could
with each path was proportional to the tension present in the be a better option with the use of new technologies. It can
build section of the well, with the values corresponding to be concluded that choosing a trajectory with characteristics
the tension in the build section as shown in Fig. 27. The use similar to the undersection or multiple build profiles is the
of decision-making techniques, such as TOPSIS, can aid in best approach for drilling ERD wells.
the selection of the best trajectory. Figure 30 illustrates the
selection criteria utilized in TOPSIS for this study.
Table 10 presents the proximity of each wellbore trajectory 8 Conclusion
to the ideal state based on the TOPSIS method, which reflects
their relative closeness to each other. In summary, this study proposes a new approach for optimiz-
Figure 31 illustrates the relative closeness of different ing well trajectory design by combining a multi-objective
ERD well trajectories. genetic algorithm and TOPSIS method. The results of the
123
Table 10 Relative closeness of

each well path Multiple build Undersection Double build Build and hold Trajectory
0.5663 0.6040 0.2754 0.4338 Relative closeness
study demonstrated that the proposed method can lead to the 2. Li, Q.; Omeragic, D.; Chou, L.; Yang, L.; Duong, K. (eds.):
selection of a better well design by taking into account oper- New directional electromagnetic tool for proactive geosteering and
accurate formation evaluation while drilling. SPWLA 46th annual
ational conditions such as risk, cost, and drilling issues, in logging symposium: OnePetro (2005)
addition to optimizing torque and length. The undersection 3. Inglis, T.: Directional drilling: Springer Science & Business Media
trajectory was found to be the best option among different (2013)
trajectory profiles due to its relatively low torque and tension 4. Guan, Z.; Chen, T.; Liao, H.: Theory and technology of drilling
engineering: Springer (2021)
at KOP, making it less susceptible to keyseat and fatigue
5. Mohamed, A.; Salehi, S.; Ahmed, R.: Significance and complica-
failure. On the other hand, the multiple build trajectory had tions of drilling fluid rheology in geothermal drilling: a review.
the lowest torque on the drill string, but its inclination was Geothermics 93, 102066 (2021)
nearly 90 degrees, making it challenging to keep the well in 6. Gao, D.; Tan, C.; Tang, H.: Limit analysis of extended reach drilling
in South China Sea. Pet. Sci. 6, 166–171 (2009)
plan, leading to higher costs due to the need for RSS and agi-
7. Aadnoy, B.; Andersen, K. (eds): Friction analysis for long-reach
tator. The build and hold trajectory had the highest tension wells. IADC/SPE drilling conference: OnePetro (1998)
at KOP, making it more susceptible to keyseat and fatigue 8. Joshi, S. (ed): Cost/benefits of horizontal wells. SPE western
failure. The double build trajectory had tension levels sim- regional/AAPG Pacific section joint meeting: OnePetro (2003)
9. Mansouri, V.; Khosravanian, R.; Wood, D.A.; Aadnoy, B.S.: 3-D
ilar to the build and hold trajectory but a lower possibility well path design using a multi objective genetic algorithm. J. Nat.
of keyseat due to its low inclination at KOP. Therefore, the Gas Sci. Eng. 27, 219–235 (2015)
undersection trajectory can be a better option for ERD wells 10. Biswas, K.; Vasant, P.M.; Vintaned, J.A.G.; Watada, J.: A review
compared to other trajectories. Overall, this study provides of metaheuristic algorithms for optimizing 3D well-path designs.
Arch. Comput. Methods Eng. 28, 1775–1793 (2021)
valuable insights into the design of ERD wells and highlights 11. Shokir, E.E.M.; Emera, M.; Eid, S.; Wally, A.: A new optimization
the importance of considering multiple operational factors in model for 3D well design. Oil Gas Sci. Technol. 59(3):255–266
the optimization process. (2004)
12. Atashnezhad, A.; Wood, D.A.; Fereidounpour, A.; Khosrava-
Author Contributions HY involved in conceptualization, methodology, nian, R.: Designing and optimizing deviated wellbore trajectories
investigation, software, data curation, visualization, and writing—orig- using novel particle swarm algorithms. J. Nat. Gas Sci. Eng. 21,
inal draft, JQ involved in writing—review and editing and supervision, 1184–1204 (2014)
BSA involved in conceptualization, validation, and supervision, RK 13. Guria, C.; Goli, K.K.; Pathak, A.K.: Multi-objective optimization
involved in conceptualization, validation, and supervision. of oil well drilling using elitist non-dominated sorting genetic algo-
rithm. Pet. Sci. 11(1), 97–110 (2014)
14. Mansouri, V.; Khosravanian, R.; Wood, D.A.; Aadnøy, B.S.: Opti-
Declarations mizing the separation factor along a directional well trajectory
to minimize collision risk. J. Petrol. Explor. Prod. Technol. 10,
Conflict of interest The authors declare no competing interests. 2113–2125 (2020)
15. Al-Mudhafar, W.J.; Wood, D.A.; Al-Obaidi, D.A.; Wojtanowicz,
Open Access This article is licensed under a Creative Commons A.K.: Well placement optimization through the triple-completion
Attribution 4.0 International License, which permits use, sharing, adap- gas and downhole water sink-assisted gravity drainage (TC-
tation, distribution and reproduction in any medium or format, as GDWS-AGD) EOR process. Energies 16(4), 1790 (2023)
long as you give appropriate credit to the original author(s) and the 16. Yasari, E.; Pishvaie, M.R.; Khorasheh, F.; Salahshoor, K.; Kharrat,
source, provide a link to the Creative Commons licence, and indi- R.: Application of multi-criterion robust optimization in water-
cate if changes were made. The images or other third party material flooding of oil reservoir. J. Petrol. Sci. Eng. 109, 1–11 (2013)
in this article are included in the article’s Creative Commons licence, 17. Khosravanian, R.; Aadnoy, B.S.: Optimization of casing string
unless indicated otherwise in a credit line to the material. If material placement in the presence of geological uncertainty in oil wells:
is not included in the article’s Creative Commons licence and your offshore oilfield case studies. J. Petrol. Sci. Eng. 142, 141–151
intended use is not permitted by statutory regulation or exceeds the (2016)
permitted use, you will need to obtain permission directly from the copy- 18. Khosravanian, R.; Mansouri, V.; Wood, D.A.; Alipour, M.R.: A
right holder. To view a copy of this licence, visit http://creativecomm comparative study of several metaheuristic algorithms for opti-
ons.org/licenses/by/4.0/. mizing complex 3-D well-path designs. J. Petrol. Explor. Prod.
Technol. 8, 1487–1503 (2018)
19. Biswas, K.; Vasant, P.M.; Vintaned, J.A.G.; Watada, J.: Cellular
automata-based multi-objective hybrid Grey Wolf Optimization
and particle swarm optimization algorithm for wellbore trajectory
References optimization. J. Nat. Gas Sci. Eng. 85, 103695 (2021)
1. Kaiser, M.J.: Modeling the time and cost to drill an offshore well.
Energy 34(9), 1097–1112 (2009)
123
20. Huang, W.; Wu, M.; Chen, L.; She, J.; Hashimoto, H.; Kawata, S.: wells. In: SPE Asia Pacific Oil and Gas Conference and Exhibition:
Multiobjective drilling trajectory optimization considering param- OnePetro (2011)
eter uncertainties. IEEE Trans. Syst. Man Cyber. Syst. 52(2), 39. Baumert, M.E.; Allouche, E.N.; Moore, I.D.: Drilling fluid con-
1224–1233 (2020) siderations in design of engineered horizontal directional drilling
21. Huang, W.; Wu, M.; Chen, L.; Chen, X.; Cao, W.: Multi-objective installations. Int. J. Geomech. 5(4), 339–349 (2005)
drilling trajectory optimization using decomposition method with 40. Zakerian, A.; Sarafraz, S.; Tabzar, A.; Hemmati, N.; Shadizadeh,
minimum fuzzy entropy-based comprehensive evaluation. Appl. S.R.: Numerical modeling and simulation of drilling cutting trans-
Soft Comput. 107, 107392 (2021) port in horizontal wells. J. Petrol. Explor. Prod. Technol. 8, 455–474
22. Wang, G.; Zhang, H.; Sun, J.; Yang, Y.; Dou, T.; Zhang, W. (2018)
et al. (eds). Optimization technology of well trajectory of shale 41. Khan, M.S.; Barooah, A.; Rahman, M.A.; Hassan, I.; Hasan, R.;
oil horizontal well group in cangdong sag. In: Proceedings of Maheshwari, P.: Application of the electric resistance tomographic
the International Field Exploration and Development Conference technique to investigate its efficacy in cuttings transport in hori-
2021: Springer (2022) zontal drilling scenarios. J. Nat. Gas Sci. Eng. 95, 104119 (2021)
23. Biswas, K.; Rahman, M.T.; Almulihi, A.H.; Alassery, F.; Al Askary, 42. Larsen, T.; Pilehvari, A.; Azar, J.: Development of a new cuttings-
M.A.H.; Hai, T.B.; et al.: Uncertainty handling in wellbore tra- transport model for high-angle wellbores including horizontal
jectory design: a modified cellular spotted hyena optimizer-based wells. SPE Drill. Complet. 12(02), 129–135 (1997)
approach. J. Petrol. Explor. Prod. Technol. 12(10), 2643–2661 43. Sanchez, R.A.; Azar, J.; Bassal, A.; Martins, A. (eds.): The effect of
(2022) drillpipe rotation on hole cleaning during directional well drilling.
24. Huang, W.-J.; Gao, D.-L.: Analysis of drilling difficulty of In: SPE/IADC Drilling Conference: OnePetro (1997)
extended-reach wells based on drilling limit theory. Pet. Sci. 19(3), 44. -
El Sabeh, K.; Gaurina-Medimurec, N.; Mijić, P.; Medved, I.; Pašić,
1099–1109 (2022) B.: Extended-reach drilling (ERD)—the main problems and current
25. Wood, D.A.: Constrained optimization assists deviated wellbore achievements. Appl. Sci. 13(7), 4112 (2023)
trajectory selection from families of quadratic and cubic Bezier 45. Bogdanov, S.; Deliya, S.; Latsin, D.; Akhmetov, M.; Gagaev, Y.;
curves. Gas Sci. Eng. 110, 204869 (2023) Udodov, A. (eds.) Drilling world-class ERD wells in the North
26. Mirhaj, S.A.; Kaarstad, E.; Aadnoy, B.S.: Improvement of torque- Caspian Sea. In: SPE Russian Oil and Gas Exploration and Pro-
and-drag modeling in long-reach wells. Mod. Appl. Sci. 5(5), 10 duction Technical Conference and Exhibition: OnePetro (2012)
(2011) 46. Krepp, T.; Kn, M.: ERD campaign feasibility study on Korchagina
27. Das, B.K.; Hassan, R.; Tushar, M.S.H.; Zaman, F.; Hasan, M.; Das, field (2011)
P.: Techno-economic and environmental assessment of a hybrid 47. Bourgoyne, A.T.; Millheim, K.K.; Chenevert, M.E.; Young, F.S.:
renewable energy system using multi-objective genetic algorithm: Applied drilling engineering (1986)
A case study for remote Island in Bangladesh. Energy Convers. 48. Farah, F.O.: Directional well design, Trajectory and survey calcula-
Manage. 230, 113823 (2021) tions, with a case study in Fiale, Asal rift, Djibouti. In: Geothermal
28. Mondal, S.; Bhattacharya, A.; Nee Dey S.H.: Multi-objective eco- Training Programme, p. 27 (2013)
nomic emission load dispatch solution using gravitational search 49. Aadnoy, B.S.; Fazaelizadeh, M.; Hareland, G.: A 3D analytical
algorithm and considering wind power penetration. Int. J. Electrical model for wellbore friction. J. Can. Pet. Technol. 49(10), 25–36
Power Energy Syst. 44(1):282–292 (2013) (2010)
29. Shi, X.; Huang, W.; Gao, D.; Zhu, N.: Optimal design of drag reduc- 50. Mansouri, H.: Stress–Based Torque and Drag Model (2017)
tion oscillators by considering drillstring fatigue and hydraulic loss 51. Aadnøy, B.S.; Andersen, K.: Design of oil wells using analytical
in sliding drilling. J. Petrol. Sci. Eng. 208, 109572 (2022) friction models. J. Petrol. Sci. Eng. 32(1), 53–71 (2001)
30. Agbaji, A.L.: Development of an algorithm to analyze the inter- 52. Aadnoy, B.S.: Modern Well Design. CRC press (2010)
relationship among five elements involved in the planning, design 53. Wu, J.; Juvkam-Wold, H. (eds.): Study of helical buckling of pipes
and drilling of extended reach and complex wells. (2009) in horizontal wells. In: SPE Production Operations Symposium:
31. Jeong, J.; Lim, C.; Park, B.-C.; Bae, J.; Shin, S.-C.: Multi-objective OnePetro (1993)
optimization of drilling trajectory considering buckling risk. Appl. 54. Sun, P.; Luo, T.; Wang, B.; Yang, W.: Sinusoidal buckling behaviour
Sci. 12(4):1829 (2022) of surface casing with negative friction in thawing permafrost. J.
32. Richard, S.; Carden, R.D.G.: Horizontal and directional drilling. Petrol. Sci. Eng. 208, 109616 (2022)
Printed in USA: by Petroskills, LLC. An Ogci Company., Tulsa, 55. Hajianmaleki, M.; Daily, J.S.: Advances in critical buckling load
Oklahoma (2007) assessment for tubulars inside wellbores. J. Petrol. Sci. Eng. 116,
33. Barton, S.; Baez, F.; Alali, A. (eds.): Drilling Performance Improve- 136–144 (2014)
ments in Gas Shale Plays Using a Novel Drilling Agitator Device. 56. Hines, J.W.: Fuzzy and neural approaches in engineering. A Wil-
North American Unconventional Gas Conference and Exhibition ley–Interscience Publication, New york (1997).
OnePetro (2011) 57. Hines, J.W.; Tsoukalas, L.H.; Uhrig, R.E.: MATLAB supplement
34. Zhang, L.G.; Liu, G.; Li, W.; Li, S.B.: Analysis and optimization to fuzzy and neural approaches in engineering. John Wiley & Sons,
of control algorithms for RSS TSP for horizontal well drilling. J. Inc. (1997)
Petrol. Explor. Prod. Technol. 8, 1069–1078 (2018) 58. Jh, H.: Adaptation in natural and artificial systems. Ann Arbor
35. Peach, S.; Kloss, P. (eds): A new generation of instrumented steer- (1975)
able motors improves geosteering in North Sea Horizontal Wells. 59. Goldberg, D.E.: Genetic algorithms in search. Optimization and
In: IADC/SPE Drilling Conference: OnePetro (1994) machine learning. (1989)
36. Cai, L.; Xu, G.; Polak, M.A.; Knight, M.: Horizontal directional 60. Whitley, L.: Foundations of Genetic Algorithms. Morgan Kauf-
drilling pulling forces prediction methods–A critical review. Tunn. mann, San Mateo, GA (1993)
Undergr. Space Technol. 69, 85–93 (2017) 61. Lee, W.; Kim, H.-Y. (eds.) Genetic algorithm implementation in
37. Feng, T.; Bakshi, S.; Gu, Q.; Yan, Z.; Chen, D.: Design optimization Python. In: Fourth annual ACIS international conference on com-
of bottom-hole assembly to reduce drilling vibration. J. Petrol. Sci. puter and information science (ICIS’05), IEEE (2005)
Eng. 179, 921–929 (2019) 62. Murata, T.; Ishibuchi, H.; Tanaka, H.: Multi-objective genetic algo-
38. Mohammadsalehi, M.; Malekzadeh, N. (eds.): Optimization of hole rithm and its applications to flowshop scheduling. Comput. Ind.
cleaning and cutting removal in vertical, deviated and horizontal Eng. 30(4), 957–968 (1996)
123
63. Yeh, W.-C.; Chuang, M.-C.: Using multi-objective genetic algo- 72. Onder, E.; Sundus, D.: Combining analytical hierarchy process and
rithm for partner selection in green supply chain problems. Expert TOPSIS approaches for supplier selection in a cable company. J.
Syst. Appl. 38(4), 4244–4253 (2011) Bus. Econ. Finance 2(2), 56–74 (2013)
64. Sakhaei, Z.; Nikooee, E.; Riazi, M.: A new formulation for non- 73. Assari, A.; Mahesh, T.; Assari, E.: Role of public participation in
equilibrium capillarity effect using multi-gene genetic program- sustainability of historical city: usage of TOPSIS method. Indian
ming (MGGP): accounting for fluid and porous media properties. J. Sci. Technol. 5(3), 2289–2294 (2012)
Eng Comput 38(2), 1697–1709 (2022) 74. Lai, Y.-J.; Liu, T.-Y.; Hwang, C.-L.: Topsis for MODM. Eur. J.
65. Yavari, H.; Sabah, M.; Khosravanian, R.; Wood, D.: Application Oper. Res. 76(3), 486–500 (1994)
of an adaptive neuro-fuzzy inference system and mathematical rate 75. Wang, T.-H.; Wu, H.-C.; Meng, J.-H.; Yan, W.-M.: Optimization
of penetration models to predicting drilling rate. Iran J Oil Gas Sci of a double-layered microchannel heat sink with semi-porous-ribs
Technol. 7(3), 73–100 (2018) by multi-objective genetic algorithm. Int. J. Heat Mass Transf. 149,
66. Yavari, H.; Khosravanian, R.; Wood, D.A.; Aadnoy, B.S.: Appli- 119217 (2020)
cation of mathematical and machine learning models to predict 76. Biswas, P.P.; Suganthan, P.N.; Qu, B.Y.; Amaratunga, G.A.: Multi-
differential pressure of autonomous downhole inflow control objective economic-environmental power dispatch with stochastic
devices. Adv Geo Energy Res. 5(4), 386–406 (2021) wind-solar-small hydro power. Energy 150, 1039–1057 (2018)
67. Rabiei, A.; Sayyad, H.; Riazi, M.; Hashemi, A.: Determination of 77. Brka, A.; Al-Abdeli, Y.M.; Kothapalli, G.: The interplay between
dew point pressure in gas condensate reservoirs based on a hybrid renewables penetration, costing and emissions in the sizing of
neural genetic algorithm. Fluid Phase Equilib. 387, 38–49 (2015) stand-alone hydrogen systems. Int. J. Hydrogen Energy 40(1),
68. Yoon, K.: Systems selection by multiple attribute decision making: 125–135 (2015)
Kansas State University (1980) 78. Qu, B.-Y.; Liang, J.J.; Zhu, Y.; Wang, Z.; Suganthan, P.N.: Eco-
69. Hwang, C.-L.; Yoon, K.: Methods for multiple attribute decision nomic emission dispatch problems with stochastic wind power
making. w multiple attribute decision making: methods and appli- using summation based multi-objective evolutionary algorithm.
cations a state-of-the-art survey. Springer-Verlag, Berlin (1981) Inf. Sci. 351, 48–66 (2016)
70. Hwang, C.-L.; Lai, Y.-J.; Liu, T.-Y.: A new approach for multiple 79. Zhang, H.; Ashok, P.; van Oort, E.; Shor, R.: Investigation of pipe
objective decision making. Comput. Oper. Res. 20(8), 889–899 rocking and agitator effectiveness on friction reduction during slide
(1993) drilling. J. Petrol. Sci. Eng. 204, 108720 (2021)
71. Behzadian, M.; Otaghsara, S.K.; Yazdani, M.; Ignatius, J.: A
state-of the-art survey of TOPSIS applications. Expert Syst. Appl.
39(17), 13051–13069 (2012)
123

Selection of Optimal Well Trajectory Using Algorithm

Uploaded by

Copyright:

Available Formats

Selection of Optimal Well Trajectory Using Algorithm

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Selection of Optimal Well Trajectory Using Algorithm

Uploaded by

Copyright:

Available Formats

Arabian Journal for Science and Engineering (2023) 48:16831–16855

RESEARCH ARTICLE-PETROLEUM ENGINEERING

Selection of Optimal Well Trajectory Using Multi-Objective Genetic

List of Symbols MOGA Multi-objective genetic algorithm

use of an agitator and rotary steerable system (RSS) [33–35].

gas injection, and water injection wells. The Well Construc-

3.1 Wellbore Length

The radius of curvature technique is utilized to determine the

Fig. 3 Yuri Korchagin field map [45]

Fig. 4 Calculation of the length

MDtotal MDKOP + MDBuildSection + MDHoldSection (4)

3.2 Torque on the Drill String During Drilling

The soft string model is utilized to determine the torque on

T μr BwLsin(I ) (6) T μr F1 β (8)

finally, the torque is calculated. The mathematical formulas

MDHoldSection (TVDAG − TVDAB

Equations (15)–(18) provide a mathematical model for

TVDAB KOP (15)

MDBuildSection (sin(I1 ) − sin(0))

TVDAG−AC MDHoldSection cos(I1 ) (17)

TVDtotal TVDAB + TVDAC−AB + TVDAG−AC (18)

Equations (19)–(22) provide mathematical models for

Fig. 7 Different parts of well profile [51] FD 0 (19)

FA F B + BwTVDAB Cos(0) (22)

Fig. 8 Buckling of pipe,

Equations (30)–(37) provide a comprehensive overview of 4.2 Step 2: Evaluation

Fig. 11 Mutation mechanism of MOGA [62, 63]

4.5 Step 5: Mutation

f i − f imin 4.6 Step 6: Elitist Strategy

where randi and randj are non-negative random numbers.

5.6 Step 6: Relative Closeness

6.5 Sensitivity Analysis

Fig. 12 Double build profile [30]

Fig. 13 Build and hold profile

Fig. 14 Undersection profile [30]

Fig. 15 Multiple build profile

Fig. 16 Genetic algorithm

Fig. 17 Genetic algorithm

Fig. 18 Genetic algorithm

 The multi-objective function is minimized during the opti-

Fig. 21 Impact of varying the

Table 6 MOGA Pareto optimum

Fig. 22 Pareto optimum solution

Table 7 Normalized membership function for each solution

Pareto front (k) Torque Length χk

1 25,008.96 14,406.87 0.1782

function [27, 28, 76].

Table 8 Results of optimization

Build and hold 26,693.08 13,932.02 1462.16 77.64 4.53

Table 9 Comparative results of

Torque (N.ft) 26,693.08 25,008.96 23,565.31 21,000.32

agitator and other special equipment in these wells, which

Fig. 27 Tension in different parts

Fig. 28 Amount of tension

Fig. 29 Amount of torque

Fig. 30 Selection criteria of

The results based on the TOPSIS method indicate the rel-

Table 10 Relative closeness of

0.5663 0.6040 0.2754 0.4338 Relative closeness

You might also like

The multi-objective function is minimized during the opti-