You are on page 1of 9

Journal of Petroleum Science and Engineering 196 (2021) 107668

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering


journal homepage: http://www.elsevier.com/locate/petrol

Error reduction of tracking planned trajectory in a thin oil layer drilling


using smart rotary steerable system
Habiballah Zafarian, Mohammadjavad Ameri *, Younes Alizadeh Vaghasloo,
javad soleymanpour
Amirkabir University of Technology, Tehran, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: Directional drilling which is a controlled drilling process, has been efficiently utilized to reach oil/gas reservoirs.
RSS technology Shale oil/gas development and also the need for extraction from wells that are hardly accessible, have recently
Planned trajectory introduced the directional drilling process as an interesting method in the oil/gas industry. However, error
Thin oil layer
reduction during tracking the planned path is a challenging issue, especially in the cases of drilling thin oil/gas
Smart well
Delayed directional drilling system
layers, as they may result in significant errors. In these conditions, it is inevitable to use smart systems (e.g. RSS)
Tracking for directional drilling. Hence, it is essential to develop intelligent and automated systems for directional drilling.
This study aims to introduce an operational control approach for intelligentization of the push-the-bit RSS system
that can improve the tracking performance of directional drilling. To describe the borehole evolution, we utilized
the Perneder-Detournay’s model which has been derived via combining the BHA modeling, rock/bit interaction
law and kinematic relationships between them. Considering the existence of delay and uncertainty in this model,
the controllers were designed in two cases (system with and without reference inputs) to prove the stability of the
system. Afterward, a relatively complicated path including horizontal and sinusoidal section in a thin reservoir
layer is designed and then the suggested controller is applied to this trajectory. According to the simulation
outcomes, the system was well able to follow the planned trajectory with a too small error and operational RSS
force.

1. Introduction and drilling multi-lateral wells have been among the effective methods
(Panchal et al., 2010). The most notable example is the development of
Directional (deviated) drilling is defined as controlled drilling along shale oil and shale gas reservoir, especially in the United States, where
a specific planned trajectory until reaching the target location. The first application of the horizontal drilling technology along with hydrauli­
controlled directional well was drilled in California in 1930, which c/acid fracturing has led to the development of these unconventional
encountered legal problems because of entering a region belonging to reservoirs. Therefore, development of directional drilling systems and
another reservoir. In the 1930s, directional wells were also drilled from enhancing their efficiency is currently an important topic in the oil and
inaccessible locations such as cemeteries. However, prior to drilling a gas industry (Li et al., 2016; Xinjuna and Jinga, 2014). In this regard,
directional relief well in Texas, little attention was paid to the direc­ various tools have been developed to perform more efficient directional
tional drilling process. After that, since vertical drilling to access oil and drilling, among which the rotary steerable system (RSS) can be
gas reservoirs is associated with various limitations, directional drilling mentioned as the most important tools. For example, one of the longest
has been employed and developed as a useful method in oil and gas horizontal wells in Al-Shaheen field in Qatar with a length of 40,320 ft
industry (Carden and Grace, 2007). (about 13 km) and a horizontal length of 35,770 ft has been drilled
On the other hand, in recent years, the global drilling industry has (about 11 km) by Schlumberger’s PowerDrive Xceed Tool (Panchal
sought to develop fields in which drilling well using conventional et al., 2010).
methods has not been economically justified. In this regard, the use of One of the challenges encountered during directional drilling,
directional drilling in these wells, increasing the length of drilled wells especially when the drilling length is increased, is to guide the drill bit in

* Corresponding author.
E-mail address: ameri@aut.ac.ir (M. Ameri).

https://doi.org/10.1016/j.petrol.2020.107668
Received 6 March 2020; Received in revised form 14 July 2020; Accepted 17 July 2020
Available online 29 August 2020
0920-4105/© 2020 Published by Elsevier B.V.
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

a direction that can track the planned path with the least error (Panchal
et al., 2010). Currently, the conventional method used for directional
drilling involves application of downhole motors that are guided by the
human agent (directional driller). In directional wells, the acceptable
error in the actual path and the target location error depends on the well
design conditions and are generally determined by the owner. The
acceptable error is usually presented as an ellipse of uncertainty, as
Fig. 2. The importance of tracking planned trajectory, case study from Saudi
shown in Fig. 1. Hence, in the current conventional drilling, error
Aramco well: Undulating 3- to 9-ft-thick sand layer (Schlumberger)
reduction and omission are entirely based on the driller experience, (Limited, 2020).
which is clearly associated with some errors. Furthermore, it can only
reduce the error to a certain extent. The increase in error becomes
especially important when the layer under drilling is thin. If the drilling
accuracy is low, it can lead to the entrance of well into the adjacent (gas
and water) layers and subsequent problems (Fig. 2). Furthermore, eco­
nomic considerations or geological and operational restrictions may
sometimes necessitate drilling directional wells with complex
three-dimensional paths. In most of the cases, it is not possible to steer
the directional well path using the current tools that are based on the Fig. 3. Schematic of push-the-bit rotary steerable system (RSS)
experience of the driller (human agent) and there is a need for smart (Limited, 2020).
automated system to achieve this goal (Denney, 2011; G. Downton,
Hendricks, Klausen and Pafitis, 2000; Kremers et al., 2015; Panchal • Increased accuracy in tracking the designed path and consequently,
et al., 2010). reduced path error (due to intelligentization and human factor
Also, intelligentization of different processes in the oil and gas in­ omission).
dustry, generally known as “smart well, reservoir and field”, has been • Continuous rotation of drill string during drilling (which improves
extensively developed in recent years. In the existing literature, a smart well quality, rate of penetration (ROP) and well cleaning, reduces
well is defined as a well that has been equipped with special tools to drilling problems and reduces torque and drag).
allow monitoring the characteristics of the region surrounding the well
and also collecting, transmitting and analyzing the well and tank data Given that intelligentization of directional drilling (especially in the
(Luu, 2015; Peymani, 2007). Intelligentization processes are based on case of RSS technology) are at the beginning of their development
two key principles, i.e. monitoring (by sensors) and control. Since pathway, improving the performance of such systems is theoretically
applying these two process is possible in drilling, directional drilling can and practically important. From the theoretical point of view, improving
be seriously considered as smart process in the oil/gas industry. Smart the directional drilling systems requires improvement of borehole evo­
directional drilling can be used especially for complex lution modeling and utilization of more precise control methods to track
three-dimensional paths and where a high accuracy is needed. Based on the planned path.
these considerations and the need for drilling along long paths, rotary Intelligentization of directional drilling has been performed based on
steerable system (RSS) technology has been recently developed by major different models of directional drilling system description, which can be
oil companies. RSS technology generally uses two mechanisms to direct generally divided into two categories of tool-independent and tool-
the drill bit. In the first mechanism i.e. “Push-the-bit”, a lateral force that dependent models. A tool-independent model is entirely based on ki­
is created by pads, pushes borehole wall and results in the modification nematic (geometric) relationships and models the system regardless of
of path and movement direction (Fig. 3). A successful example of this the type, shape, and applied forces on the actuator (drilling tool). The
mechanism is Schlumberger’s PowerDrive. In the second mechanism, advantage of this type of modeling is its simplicity and also applicability
“Point-the-bit”, change in position of eccentric ring results in deflection for a variety of directional drilling tools because of the generality of the
of shaft and then change in bit direction. In this model two different model. However, the accuracy of this type of modeling is very low, as the
technologies have been used, namely the Schlumberger’s Xccede and actual shape of drill string, bottom hole assembly (BHA) and drilling
Wotherford’s Revoloution (Limited, 2020). environment (porous rock) are not taken into account (Panchal et al.,
This technology has generally two main advantages, as the 2010; Wen and Kreutz-Delgado, 1991).
following: In order to control directional drilling system, several studies have
been conducted based on these models especially since 2000. Panchal
(Panchal et al., 2010) developed a tool-independent model and used a PI
controller to control this model. This is a two-input i.e. azimuth and
inclination and two-output i.e. tool face and sliding to rotation ratio
(steering ratio) model. Furthermore, in this model, the independent
variable is time. According to Panchal’s point of view, directional dril­
ling is indeed an attitude control that basically controls azimuth and
inclination. In a subsequent research (Bayliss et al., 2014), an optimal
controller, H∞, was used to control Panchal’s model, but the results
were almost identical. Since in Panchal’s model, only the position was
controlled, in another study (Matheus et al., 2014), an external ring was
added to allow controlling the directional drilling path. However, these
studies are far from the actual drilling conditions and then unreliable,
due to utilization of tool-independent models.
In another set of studies, it has been attempted to model borehole
evolution through modeling BHA and also rock/bit (R/B) interaction. In
initial studies, such as those performed by Lubinski and Woods (Lubinski
Fig. 1. Schematic of ellipse of uncertainty (right) and actual drilling and Woods, 1953), analytical models had been extracted for BHA
result (left).

2
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

modeling and R/B interaction, though no general formulation of system and interpretation of the system. First step is deriving BHA deformation
kinematics had been presented. In the 1970s and 1980s, the majority of equation in order to calculate the forces and torques (axial, lateral, and
studies were focused on the mechanical models of BHA, as the design of angular) applied on the drill bit. Here, BHA is modeled as an Euler-
BHA was believed to be the most important factor in directional drilling Bernoulli linear beam with constant EI (bending stiffness) and uniform
model. Thus, the R/B interaction did not receive a great deal of attention weight distribution ω (the weight adjusted due to drilling fluid buoy­
until the mid-1980s. Most of studies on the prediction of directional well ancy). It is noted that the contact of BHA with the borehole wall is
evolution have been carried out through numerical methods. Only a few limited by some stabilizers, thus the well applies forces and torques on
studies have sought to analytically derive the governing equations, such the drill bit in a spatially delayed manner. (Kremers et al., 2015; Per­
as Downton (2007) (Kremers et al., 2015). neder, 2013).
In (G. C. Downton, 2007) he has proposed three models to describe ( / )
borehole evolution. In Downton’s models, the independent variable in EI ∂3 θ ∂s3 = ω.sinΘ1 (1)

( / ) ∑
n− 1
̂2 = − EI ∂2 θ0 ∂s2
F s=0
F2 / F* = F b ( ̂
; ̂ θ − Θ1 ) + F w γ sinΘ1 + F r Γ + F i (Θi − Θi+1 ) (2)
i=1


n− 1
M ̂ / F* λ1 = M b (̂
̂ = − EI (∂θ0 /∂s)s=0 ; M θ − Θ1 ) + M w γ sinΘ1 + M r Γ + M i (Θi − Θi+1 ) (3)
i=1

the transform function is not time, but is the drilled length, which
simplifies data extraction and interpretation. In another research (Sun
et al., 2012), Downton’s model has been controlled by the L1 controller. Where:
Among these researches, the most complete borehole evolution model / / /
has been developed by Perneder and Detournay (Perneder, 2013). This F* = 3.EI λ21 , Γ = FRSS F* , γ = ω.λ1 F* (4)
model, which has been used as the base model in the present study,
In these equations, the terms on the left side of equations are the
models the directional drilling system as a delayed nonlinear system. In
lateral force and torque applied to the drill bit, respectively, λ is the
this study, the robustness of the proposed controller for system un­
distance of each component of the BHA, and the italic variables, F and
certainties have generally investigated, but the stability has not been
M, are impact coefficients that solely depend on the arrangement of
proved for system uncertainty. This study also shows that the weight
BHA. Θ and θ are inclination angles of the borehole and BHA, respec­
effect of BHA is very small and does not affect the final solution.
tively. A hat mark on the variable denotes “at the bit”. It should be noted
The aim of the present paper is to design an intelligent system so as to
that the indices w and r indicate the effects of string weight and RSS
automatically perform the directional drilling process. We have indeed
force, respectively. The indices b and i denote the relative direction of
designed the software available in the RSS system. Here, it is mainly
drill bit with respect to the first part and the constraints imposed by the
attempted to deal with the actual and practical issues in directional
geometry of the hole, respectively (Perneder, 2013).
drilling and avoid unnecessary theoretical complexities. Three issues are
The kinematic expression has been presented as below. This relation
highlighted here. The first issue is related to the uncertainties of direc­
expresses that the difference between the borehole and BHA inclination
tional drilling system, which are mainly due to the alterations in the
at the bit (bit tilt) is equal to proportion of lateral penetration of bit in
material of the rock under drilling and WOB1 variations. Another
the rock divided by axial one:
important operational issue is the maximum level of RSS force exerted
by the system. Existing RSS systems can apply the determined level of ψ = ̂θ − Θ
̂ = atan( − d2 / d1 ) ≈ − d2 /d1 (5)
lateral force and change the direction of drill bit to a certain extent.
Therefore, all designs must be performed such that they fall within this Moreover, the R/B interaction law indicates how the forces and
operational range. Finally, with respect to common paths in directional moments (F1, F2, M) acting on the drill bit are related to the drill bit
drilling, the system should be able to solve the problem of tracking for penetration level in the rock (in two dimensions d1, d2, ϕ1). In Perneder-
complex paths. We use a delayed Perneder-Detournay’s model. In all Detournay’s model, the R/B interaction law has been derived using the
stages, stability is mathematically proved, with respect to operational theory of cutter/rock interaction. As shown in Fig. 5, the interaction of
aspects. blunt cutter with rock is modeled by bilinear relation between the depth
of cut and the force per unit width of the cutter:
Based on this theory, after some algebra, the linear relationship for
2. Directional drilling system model
rock/bit interaction law is derived:
This section presents modeling of directional drilling system based ̂
F1 = − G1 − H1 d1 (6.1)
on the Perneder-Detournay’s model, which is the most complete
modeling available for borehole evolution (Fig. 4). Perneder has derived ̂
F2 = − H2 d2 (6.2)
this model using a three-component model, consisting of BHA modeling,
rock/bit interaction law, and a geometrical (kinematics) relationship ̂ = − H3 ϕ
M (6.3)
(Perneder, 2013).
Before deriving equations, it is necessary to introduce the dimen­ Here, H3, H1, H2 depend on the rock and drill bit properties and are
sionless length drilled (ξ = L/λ ). As mentioned above, using drilled experimentally determined. Furthermore, d1, d2, ϕ are the axial, lateral
1
and angular penetration levels of the drill bit in rock, respectively. In the
length as independent variable (instead of time) can simplify extraction
first expression, G1 is a part of axial force that is not influential on
penetration into the rock.
1 Combining the three components leads to the following equation for
wt on the bit: the amount of downward force exerted on the drill bit.

3
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

Fig. 4. Modeling of borehole evolution (left) and components of the model (right) in Perneder-Detournay’s model (Perneder, 2013).

Table 1
Main input and output parameters.
ξ Dimensionless length drilled
xr Desired bit inclination (reference input), rad.
x1 Borehole inclination (system output, state variable), rad.
e1 Error of borehole inclination (the difference between xr and x1), rad.
Γ Input (control effort), scaled force exerted by RSS

Θ́i = (1 / xi )[Θ (ξi− 1 ) − Θ (ξi )] (9)


Based on the equations presented for the three states of borehole
evolution system, the system state-space form can be formulated as
Fig. 5. Bilinear rock/blunt cutter interaction (Perneder, 2013).
below (Kremers et al., 2015):

(10)
′ ′
x (ξ) = A0 x(ξ) + A1 x(ξ1 ) + A2 x(ξ2 ) + B0 Γ + B1 Γ + BW

A0 =[a b c;d 0 0;0 0 0], A1 = [e 0 0;f 0 0;g 0 0], A2 = [h 0 0;0 0 0;i 0 0]


As it is clear, this is a time delay system of input-delay type and in
terms of delay characteristics, it has two constant delays (with values of
1 and approximately 2.66 according to geometry of BHA). In this system,
the control input is the amount of lateral force exerted by the RSS (Γ)
Fig. 6. System states in the delayed directional drilling model (Per­ and the control parameter is related to all three states of the system,
neder, 2013). especially the inclination of the borehole. In fact, the borehole inclina­
tion error measured by the sensors above the actuator, is compared to
the variations of the drill bit inclination angle: the inclination reference, and based on this error, the controller com­



2 ∑
2
χ .Π.Θ = M b .(Θ1 − Θ) + (χ / η).F b .(Θ − Θ1 ) + (F b M i − F i M b − M i ηΠ / ηΠ).(Θi − Θi+1 ) − (χ / η) F i .(((Θi− 1 − Θi ) / xi ) − ((Θi − Θi+1 ) / xi+1 ))
i=1 i=1

(7)

− (χ / η).F w .γ.(Θ − Θ1 ).cos ​ Θ1 + ((F b M r − F r M b − M r ηΠ) / ηΠ).γ.sin ​ Θ1 ± (χ / η).F r .Γ + ((F b M r − F r M b − M r ηΠ) / ηΠ).Γ

mands the RSS actuator to apply force on the wellbore. It is notable that
in this system, two parameters, i.e. η (the ratio of lateral strength to axial
Considering Θ (borehole inclination) as one of the system states, the strength of rock against the bit penetration) and Π (effective WOB) are
mean inclination in the first and second parts of BHA can be regarded as considered as system uncertainties (Kremers et al., 2015; Perneder,
two other states, which can be obtained based on the system geometry, 2013).
according to Fig. 6:

∫ξi 2.1. New formulation of model


Θi = (1 / xi ) Θ (σ)dσ (8)
ξi− 1
The directional drilling system (equation (10)) has two nonlinear
terms due to the weight of BHA. The order of these two terms related to
the other system terms is negligible. For instance, if uncertain parameter

4
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

of system ηΠ = 0.1, despite that the order of the terms of the system is to Γ = 0.0062). It is noted that the maximum lateral force applied by RSS
about 103 , the coefficient of the terms sin ​ 〈Θ〉1 and cos ​ 〈Θ〉1 are about is approximately equal to 20,000 N. As it is evident, the system (with ηΠ
10− 3 and 3 × 10− 7 ,respectively. This difference can be also predicted = 0.06 and ηΠ = 0.279) is unstable and tends to infinity in an oscillatory
from system modeling, as the effect of BHA weight is very weak, manner. In fact, with applying the RSS lateral force to the borehole wall,
compared to other applied forces. According to practical approach, it is the bit inclination tends to infinity, due to system instability. As illus­
better to eliminate these nonlinear terms. In fact, a reasonable model trated in Fig. 7, depending on the system uncertainty (which is equal to
reduction takes place here and a simplified system is controlled. ηΠ), the oscillation level of the response would be different, such that a
Moreover, the system presented in Equation (10) has three states lower uncertainty is associated with a higher oscillation and vice versa.
such that the second and third states are functions of the first one, ac­ This can be justified based on the physics of the drilling problem. In fact,
cording to Equation (8). On the other hand, the matrices of the a lower effective WOB (П) would be associated with the more freedom of
mentioned system are sparse matrices that makes their analysis drill bit to deviate from the drilling axis and oscillate in the lateral di­
complicated. Therefore, we derivate the existing equations and with rection. In addition, a lower lateral drilling resistance compared to the
substituting the derivatives of the second and third equations into the axial resistance (i.e. a smaller η level) would provide more freedom for
first one, equation (11) is obtained: the drill bit to move in the lateral direction and swing in the borehole. As
stated above, a smaller uncertainty results in a greater degree of freedom
x́′ 1 (ξ) = ax´1 + ex´1 (t − 1) + hx´1 (t − 2.6) + bdx1 + (bf + cg)x1 (t − 1) for the drill bit. Therefore, compared to larger uncertainties, more de­
(11)

+ cix1 (t − 2.6) + bu; bu = b0 Γ + b1 Γ´ ​
´ viations of the drill bit are allowed at smaller uncertainties.
Another important point is that the system becomes non-minimum
Therefore, with selecting the following states for the system, we will phase at large uncertainties (Fig. 8) (for ηΠ levels above 8.3). The
have: physical interpretation of this problem is that at a certain critical un­
certainty, due to the high weight on the drill bit or the high lateral
z1 = x1 , z2 = x´1 resistance of the rock, the shaft connected to the drill bit is under
[ ] [ ][ ] [ ][ ] [ ][ ] [ ] pressure and the applied lateral force acts in the opposite direction.
´ 0 1 z1 0 0 z1 (t− 1) 0 0 z1 (t− 2.6) 0
Z = z´1 = Therefore, this range is unacceptable for the directional drilling prob­

+ + + u
z2 bd a z2 bf +cg e z2 (t− 1) ci h z2 (t− 2.6) 1
lem, and the present paper examines and analyzes the system below this
(12) critical value.
Table 1 shows the main input and output parameters used for According to the open loop response of the system, the first step is to
simulation: stabilize the uncertain system using an appropriate controller design.
Based on the nature of directional drilling, the next step is to prove
3. Open loop system analysis stability of the tracking problem for planned path. It should be
emphasized that all these steps must be carried out under operational
As indicated in the previous section, modeling the directional drilling conditions in terms of the force exerted by RSS (control effort).
system results in a delayed model for the system. Generally speaking, all
systems are delayed in reality, but when modeling, some systems are 4. Designing controller for smart directional drilling system
modeled as a delayed system and others are modeled without delay.
Delayed systems can be categorized from two general perspectives. Similar to non-delayed systems, the stability of delayed systems can
From the viewpoint of delay location, a delay can occur in the input, be analyzed based on two general logics, including frequency domain
output, state and in the state derivative. Delayed-state systems are called and time domain methods. The main advantages of frequency based
retarded TDSs and those with delay in the state derivative are referred to
as neutral systems. In terms of delay characteristics, delayed systems are
divided into two categories of discrete and distributed delayed systems
(Lubinski and Woods, 1953; Sun et al., 2012). The system modeled in
this paper is a system with discrete delay in the state. Before designing
the relevant control law to control the system, the behavior of the open
loop system is first evaluated.
To stimulate the open loop of the system, the RSS force is applied to
the system as a step input at location 1 from zero to 10,000 N (equivalent

Fig. 7. Response of open loop system against the step input (with uncertain Fig. 8. The effect of non-minimum phase response on the direction of bit
parameter ηΠ = 0.06 and ηΠ = 0.279). (Perneder, 2013).

5
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

methods include better intuitive understanding and development of


k0 = [7513.4 3168.7], k1 = [− 13917 ​ − 22], k2 = [6401.1 3.5] (17)
conventional methods for delayed systems, such as Nyquist diagram. On
the other hand, their disadvantages are related to the fact that they Now simulating the system in the presence of this controller will lead
cannot be extended to multi-input and multi-output systems or to time to the following results. It should be noted that for a more accurate and
dependent delays, and that they require highly complicated calcula­ realistic examination of the problem, the initial value (initial error) of
tions, especially for multiple delays. In contrast, time based methods the system has been assumed to be 2◦ .
(using the Lyapunov function) have advantages such as the ease of As can be seen in Fig. 9, the system has been stabilized at a spatial
extension to multi-input multi-output systems and time dependent de­ distance about 8 λ1 (the distance from drill bit to the RSS force). From
lays and the ease of extension to uncertain, random and perturbed sys­ the operational point of view, the control effort required for system
tems. As with non-delayed systems, the major disadvantage of these stabilization is about 0.0082 that comfortably falls within the opera­
methods is that they mainly present necessary conditions. The Kra­ tional range.
sovskii’s method for Lyapunov functionals (1956) and the Razumikhin’s
method for Lyapunov functions (1956) are two methods of time analysis 4.2. Designing controller for the uncertain system
using the Lyapunov direct method for delayed systems (Boukas and Liu,
2002; Fridman, 2014). For a simple delayed system, equation (13), se­ In the section, the parametric uncertainty of the directional drilling
lection of Lyapunov function represented by equation (14) will lead to system is considered. As mentioned, the parametric uncertainty of the
LMI condition represented in equation (15) (Boukas and Liu, 2002): model is due to the changes in the rock properties and changes of WOB

m that both of them are real in drilling. Considering the system matrices
( )
ẋ(t) = Ax(t) + Aj x t − hj (t) + Bu (13) consisting of the certain fixed nominal matrices and uncertain matrices
j=1 and using the state feedback controller, the system will be stable if the
following LMI holds, assuming positive X, V and ε (Boukas and Liu,
m ∫
t
∑ 2002):
T
V(x(t)) = x (t)Px(t) + xT (s)Qj x(s)ds (14) ⎛ ⎞
j=1
t− tj
⎜ M11 M1 … Mm β ⎟
⎛ ⎞ ⎜ ⎟
⎜ T ⎟

m ⎜ M1 − V1 … 0 β1 ⎟
⎜ AT P + PA + ⎜ ⎟
Qj PA1 … PAm ⎟ ⎜ ⎟
⎜ ⎟ L=⎜ ⋮ ⋮ ⋱ ⋮ ⋮ ⎟ <0 (18)
⎜ j=1 ⎟ ⎜ T ⎟
⎜ ⎟ ⎜ Mm 0 … − V1 βm ⎟
⎜ ⎟
L=⎜ AT1 P − Q1 0 ⎟<0 (15) ⎜
⎜ T


⎜ ⎟ ⎝ β βT1 ⋯ βTm − εI ⎠
⎜ ⋮ ⎟
⎜ ⎟
⎝ 0 − Qm ⎠
ATm P

Thus, in case the positive matrices P and Q exist and the LMI equa­
tion holds, it can be proved that the Lyapunov derivative will be nega­
tive and stability is proved:

4.1. Designing controller for the system without uncertainty

The controller of linear delayed systems is mainly in the form of the


feedback of system states (assuming accessibility of the system states).
u(t) = k0 x(t) + k1 x(t − h1 ) + k2 x(t − h2 ) (16)
Here h1 and h2 are system delays. Considering the directional drilling
system without uncertainty; i.e. with nominal value of ηП (ηП = 0.1),
using yalmip in the Matlab environment, the LMI derived in the previous
section is solved by the Mosek solver and the controller gains are
obtained.

Fig. 9. Inclination error of nominal system under controller.


Fig. 10. Inclination error of uncertain system (ηΠ = 0.03) under controller and
its RSS force.

6
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668


m either one of these models or a combination of them. These models are
M11 = AX + XAT + BY + Y T BT + εDDT + Vj ; Mi = Ai X + XAT + BYi ; 1 type I: build and hold, type II: S curve, type III: continues build and type
IV: horizontal or ERD2 (Carden and Grace, 2007). Thin oil layers are
j=1

<i<m usually drilled using the fourth type of directional drilling models. Thus,
here, the designed path is based on this model, as well. In this trajectory,
β = XEAT + Y T EBT ; 1 < i < m; βi = XEiT + YiT EBT ; 1 < i < m the well is initially vertical. Therefore, the inclination angle of the drill
bit equals zero. Afterward, at the KOP3 point at a depth of 2000 m
According to the obtained LMI, the system is stable for all values of
(equivalent to ξ = 546), the drill bit begins to “build” until it reaches a
ηΠ. This is demonstrated by system simulation in the presence of the 45◦ angle. This about 730 m section is associated with a linear increase
great uncertainty that can be simulated with the initial condition of 1◦ .
in the inclination angle (ramp reference input). Then, the inclination is
As can be seen Fig. 10, due to the need for covering uncertainty in the
preserved at this angle and “hold” section is produced. This section,
system, in comparison with certain system, a greater force has to be
where the angle of inclination is constant, is similar to the step reference
applied by the RSS to guarantee the stability. As shown in this figure, the
input. In the next stage, with build section, the inclination of the drill bit
required RSS force is approximately maximum of the operational force.
is increased to 90◦ . Finally, the drill bit enters the thin reservoir layer at
Therefore, considering uncertainty in the range of 0.03 <ηΠ <8.3, for
a depth of about 3500 m and must maintain its path within this layer. It
initial error of 0.01 rad the RSS force level falls within the operational
is notable that in the simplifying assumption, the reservoir layer is
range. As it is clear, the uncertainty interval has been only limited by the
considered to be horizontal (with no change in inclination). Therefore,
lower bound, and the upper bound of uncertainty has no operational
the aim in the horizontal section is only to maintain a 90◦ angle of
constraint problem (the upper limit is determined according to non-
inclination (step input). But in a more complex and probably realistic
minimum phase system). This conclusion is also consistent with the
case, it can be assumed that the layer is sinusoidal (according to Fig. 11)
physics of the problem, because the lower bound of uncertainty causes
and the drill bit should track this path. Therefore, the overall path of the
severe oscillations in the borehole wall that require more RSS force to
system reference input in terms of inclination angle is as the following:
control them.

5. Simulation of smart directional drilling in a thin reservoir


layer

As mentioned above, one of the most important applications of


intelligentization of the directional drilling is to drill a thin layer of
reservoir, as error occurrence in drilling such layers can cause subse­
quent problems such as water and gas production. Therefore in some
cases, the only way to drill such layers is to use smart technologies such
as RSS. All the solutions presented so far for different cases only solve the
problem of regulation, in which the reference input is always considered
to be zero. Although this solution justifies some directional drilling
patterns such as “hold”, the problem of tracking for the system should be
solved as other patterns exist in directional drilling, such as “Continuous
Build”. In this section, the trajectory of directional drilling from the
surface to the target of thin reservoir layer is planned and then the
tracking problem is solved using PI controller.

Fig. 11. Planned complex reference input.


Fig. 12. Simulation results of tracking complex reference input and associated
RSS force.

5.1. Planning trajectory for drilling of thin layer

There are several models for well trajectory design in directional


drilling. Four models of such paths are known as the main models, each 2
extended-reach drilling: deep wells with horizontal distance-to-depth, or H:
with its own applications. The majority of directional wells are related to V, ratios less than two.
3
Kick of point: The point at which a vertical well is intentionally deviated.

7
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

Fig. 13. Error of tracking the complex trajectory in two resolution.

5.2. Simulation results of planned trajectory to track our designed path.

Using method presented in (Ogata and Yang, 2002), feedback of the Credit author statement
error (the difference between input and output) to the reference input
and then eliminating this error by means of controller like PID,4 can Habiballah Zafarian: Conceptualization, Methodology, investiga­
guarantee the stability of the tracking problem: tion, software Mohammadjavad Ameri: Conceptualization, investiga­
tion, Supervision Younes Alizadeh Vaghasloo: Supervision javad
u(t) = k0 x(t) + k1 x(t − h1 ) + k2 x(t − h2 ) + kr r(t) (19)
soleymanpour: software, Writing- Reviewing and Editing.
Based on PI controller design (gain as bellow), the simulation of
tracking the trajectory for nominal system is presented in Fig. 12: Declaration of competing interest
kP = [− 100 ​ − 100], kI = [− 2 ​ − 1] (20)
The authors declare that they have no known competing financial
To demonstrate the quality of the response, the error of tracking interests or personal relationships that could have appeared to influence
complex reference has been presented in Fig. 13. As illustrated, while the work reported in this paper.
the order of reference is about 10− 1 rad., the order of the error is about
10− 3 .This result is quite acceptable for directional drilling. Appendix A. Supplementary data

6. Conclusion Supplementary data to this article can be found online at https://doi.


org/10.1016/j.petrol.2020.107668.
In this paper, based on Perneder-Detournay’s analytical model, we
proposed a robust operational controller that is capable of tracking References
complex planned trajectory of directional drilling with a high precision.
Bayliss, M.T., Whidborne, J.F., Panchal, N., 2014. Structured uncertainty analysis of pole
The proposed controller has several advantages:
placement and H∞ controllers for directional drilling attitude tracking. IFAC Proc.
Vol. 47 (3), 9283–9288.
• Operational perspective: The nonlinear terms of the model, which Boukas, E.-K., Liu, Z.-K., 2002. Stochastic time delay systems. In: Deterministic and
Stochastic Time-Delay Systems. Springer, pp. 177–183.
have negligible effects on the system, were eliminated to obtain a
Carden, R.S., Grace, R.D., 2007. Horizontal and Directional Drilling (Tulsa, Oklahoma).
linear model, which was quite reasonable from the operational point Denney, D., 2011. Drilling automation: an automatic trajectory-control system. J. Petrol.
of view. However, the uncertainty parameter of the rock and the Technol. 63 (12), 84–87.
amount of WOB have also been taken into account, because of the Downton, G., Hendricks, A., Klausen, T.S., Pafitis, D., 2000. New directions in rotary
steerable drilling. Oilfield Rev. 12 (1), 18–29.
importance of uncertainty in drilling. Also, the key operational Downton, G.C., 2007. Directional drilling system response and stability. In: Paper
constraint of this system, the force exerted by RSS, falls within the Presented at the 2007 IEEE International Conference on Control Applications.
acceptable operational range in all designing stages. Fridman, E., 2014. Introduction to Time-Delay Systems: Analysis and Control. Springer.
Kremers, N.A.H., Detournay, E., Van De Wouw, N., 2015. Model-based robust control of
• Proving the stability: Using the Lyapunov function for delayed directional drilling systems. IEEE Trans. Contr. Syst. Technol. 24 (1), 226–239.
systems and its conversion into LMI conditions, the stability of the Li, X., Gao, D., Zhou, Y., Zhang, H., 2016. Study on open-hole extended-reach limit
controlled system was guaranteed for the uncertain system in two model analysis for horizontal drilling in shales. J. Nat. Gas Sci. Eng. 34, 520–533.
Limited, S., 2020. PowerDrive X6 Rotary Steerable System. Retrieved from. https://www
cases: i.e. with and without reference inputs. .slb.com/drilling/bottomhole-assemblies/directional-drilling/powerdrive-x6-rota
• High accuracy: Using the proposed model, the error magnitude of ry-steerable-system.
tracking planned drilling path is approximately 10− 2 times the Lubinski, A., Woods, H., 1953. Factors affecting the angle of inclination and dog-legging
in rotary bore holes. In: Paper Presented at the Drilling and Production Practice.
reference input, which outperforms previous works in terms of
Luu, H.V., 2015. SPE Rotary Steerable System.
accuracy. Matheus, J., Ignova, M., Hornblower, P., 2014. A hybrid approach to closed-loop
directional drilling control using rotary steerable systems. In: Paper Presented at the
SPE Latin America and Caribbean Petroleum Engineering Conference.
A relatively complicated path was designed for directional drilling in
Ogata, K., Yang, Y., 2002. Modern Control Engineering, vol. 4. Prentice hall India.
a thin oil layer based on the conventional models that are available for Panchal, N., Bayliss, M.T., Whidborne, J.F., 2010. Robust linear feedback control of
directional wells. Simulation results show that the system was well able attitude for directional drilling tools (13th IFAC symposium on automation in
mining, mineral and metal processing). IFAC Proc. Vol. 43 (9), 92–97.
Perneder, L., 2013. A Three-Dimensional Mathematical Model of Directional Drilling.
Peymani, Y., 2007. Smart wells concept, elements and applications. In: Paper Presented
4
proportional–integral–derivative controller. at the 2nd Iranian Petroleum and Energy Club Congress, Tehran, Iran.

8
H. Zafarian et al. Journal of Petroleum Science and Engineering 196 (2021) 107668

Sun, H., Li, Z., Hovakimyan, N., Başsar, T., Downton, G., 2012. L 1 adaptive control for Xinjuna, G., Jinga, L., 2014. Research on application of steering drilling technologies in
directional drilling systems. IFAC Proc. Vol. 45 (8), 72–77. shale gas development. Procedia Eng. 73, 269–275.
Wen, J.-Y., Kreutz-Delgado, K., 1991. The attitude control problem. IEEE Trans.
Automat. Contr. 36 (10), 1148–1162.

You might also like