Vous êtes sur la page 1sur 26

IADC/SPE 151416

Automated Drilling Systems for MPDThe Reality


Saad Saeed, Randy Lovorn, and Kjetil Arne Knudsen, Halliburton
Copyright 2012, IADC/SPE Drilling Conference and Exhibition

This paper was prepared for presentation at the 2012 IADC/SPE Drilling Conference and Exhibition held in San Diego, California, USA, 68 March 2012.

This paper was selected for presentation by an IADC/SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not
been reviewed by the International Association of Drilling Contractors or the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily
reflect any position of the International Association of Drilling Contractors or the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any
part of this paper without the written consent of the International Association of Drilling Contractors or the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is
restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of IADC/SPE copyright.


Abstract
Managed pressure drilling (MPD) has heralded an era of accurate and precise downhole pressure control. Not only can
todays MPD systems operate in tight operational envelopes, but they, more importantly, provide dynamic real-time well
control while drilling. This dynamic precision has directly enabled access to assets that were previously considered virtually
undrillable. But, how can MPD boast such success? One of the major reasons is automation. Automation can provide levels
of functional control that are difficult, if not impossible, for human operators to achieve and maintain. MPDs inherent
closed-loop setup, coupled with conventional methodology, naturally lends to automated applications.
Furthermore, despite all the advancements in MPD automation within the past few years, there is a common
misconception that current MPD systems can provide fully automatic control (cruise control) of the entire drilling process.
This is far from reality, as MPD systems today provide supervised automation.
Though there is a lot of material pertaining to the benefits and functionality of MPD automated drilling systems, there is
very little technical information on how the systems actually function. The primary focus of this paper is to fill this disparate
gap by examining the internals of such a system and, in turn, detailing how it actually works.
The paper will start by developing a generic framework, which is common to all MPD automated drilling systems
(independent of the company). The components, technology, and architecture will all be presented in detail, which will be
augmented by tracing the information flow through the system to clearly illustrate how the MPD process is actually carried
out. Having established the technologies, capabilities, and limitations will become apparent. This will then be followed by
examples of concrete implementations.

Introduction
Managed pressure drilling (MPD) has not only facilitated the introduction of innovative control solutions, but spurred the
development of real-time automated systems in the drilling industry, heralding an era of accurate and precise downhole
pressure control. However, these advancements are not universal. It is important to note that MPD comes in a number of
different forms with varying automation influence. Interestingly, the term MPD, is unique in that it not only relates to a
specific technique, but is also commonly used as an umbrella term to describe an entire technology. The industry has tried to
clarify this situation by taking one of two approaches. The first path is by defining MPD in broad, generic terms, which cover
a variety of divergent techniques. The second path takes the opposite approach and counters generality by distilling and
defining each technique into distinct categorical delineations based primarily on physical phenomena. As the primary focus
of this paper is the reality of the automated drilling system in MPD, it is important to note that all MPD is not equal when it
comes to automation. Actually, many are surprised that automation does not apply equally to all forms of MPD, irrespective
of the definition that is used. Rather, automation (as we know it) has primarily been developed and applied in primarily one
segment of MPD. This segment is focused on the precise prediction and control of bottomhole pressure (BHP) within
predefined limits. The most well-known of these techniques is constant bottomhole pressure (CBHP) drilling, where applied
surface (back) pressure is manipulated by a choke device to indirectly control the BHP profile. This workflow naturally lends
very well to an automated philosophy and is not only the primary focus of MPD automation today, but the core focus of this
paper.
There is also a large misconception and misunderstanding as to the level of automation in MPD. Many perceive systems
to be fully automatic, whereby they can be turned on and forgottenhandling all operational phases, events and scenarios in
a fully autonomous manner. Reality is far from this, as todays systems provide supervised automation at best, where
2 IADC/SPE 151416
specific phases are handled by a fully automated workflow, but certain events and scenarios demote systems into a state
requiring user guidance and intervention.
The aim of this paper is to clarify the state of MPD automated drilling systems. In particular, current philosophies will be
examined, augmented by a discussion on systems architecture. The emphasis will then shift to core components of the
technology and how the system actually works. But, before starting an in-depth technical discussion, it is important to review
some background fundamentals on which all the systems are firmly based.

Background (What is being Controlled?)
Automated drilling systems for MPD have one primary focus and that is to control the BHP (and associated profile) to a
desired value. As there is (currently) no mechanism or tool to directly control BHP, what are we conceptually controlling?
We are all well aware of the importance of accurate BHP control and regulation. It not only enables us to minimize fluid loss
and prevent undesired fracturing, but also assists in preventing undesired influx of formation fluids into the wellbore. In
typical MPD, the BHP pressure is maintained slightly above the pore pressure of the formation, without exceeding the
associated fracture pressure. In a conventional drilling setup, this could only be attempted purely by manipulation of the fluid
system and mud weights. However, owing to the narrower operating envelope in typical MPD applications and subsequent
loss of an operational pressure cushion, additional techniques are required to achieve the desired control. The solution to this
challenge presents itself in the form of converting the conventional open system to a managed closed system.


Fig. 1(a) Conventional Open System, (b) Managed Closed System.

Whereas, in an open system (Fig. 1a), the annulus is completely open to the atmosphere with all returning fluids flowing
unregulated to fluid treatment facilities, a closed system (Fig. 1b) is obtained by closing or sealing off the annulus (which not
only isolates it from the atmosphere but enables pressurization at or near the surface) using a rotating control device (RCD).
The RCD seals about the drill string above the wellhead and redirects returning well fluid (via a flow spool) to a choke
manifold. Backpressure can then be applied to the annulus (to influence BHP) by variably restricting flow of the fluid
through the operative chokes. The greater the restriction of the flow, the greater the backpressure applied to the annulus.
Thus, the BHP can be conveniently regulated by varying surface (back) pressure applied to the annulus. This can more
clearly be seen if we review the BHP equation, which can be simply expressed as follows:
P
bhp
= P
h
+P
]
+P
s
(1)

In its simplest form, BHP is a direct function of three major components:
P
h
: Hydrostatic pressure. This term is very well-known and understood in the industry and forms the primary form of
control in conventional drilling setups. The calculation of hydrostatic pressure is also quite straightforward, with well-
established equations requiring minimal corrections to provide accurate results. However, one of the major disadvantages in
P
h
is that it can only be manipulated indirectly by altering the mud weight, which can be a lengthy (if not expensive) process,
with the intended pressure effect requiring a large transient time to take effect.
P
f
: The Frictional (or dynamic) pressure term is the most complex of the three components. It is not only affected by well
geometries and fluid rheologies but is a function of fluid and pipe movement. This is also the most difficult term to calculate
accurately, requiring solutions to complex equations and workflows, usually provided by hydraulic models. P
f
is also the
component that we have the least direct influence or control over. It is true that drilling programs attempt to model and
account for P
f
, but this influence is estimation at best.
P
s
is not usually used in conventional drilling setups, as this is the surface or applied (back) pressure component. As this
component is applied, minimal calculations are required to work out its contribution. P
s
is usually implemented by controlling
a pressure source on the annular side, thereby directly applying pressure to the annulus and, in turn, indirectly the BHP. The
main benefit of using P
s
is that it is not only dynamic in its application, but with incompressible (one phase) systems, the
impact on BHP is seen almost instantaneously. The manipulation of P
s
, thereby, gives us a method to indirectly control BHP
in a dynamic, timely, real-time, and accurate manner. This is the component that all MPD automated drilling systems
manipulate to control BHP.
IADC/SPE 151416 3

Fig. 2Components of BHP.

Now that we have identified what automated systems are actually controlling, let us introduce some important, related
concepts that all systems functionally rely on to achieve this control. These concepts are as follows:
Controlled quantities: Those quantities that we are trying to control or keep at a desired level (i.e., wellhead pressure
(WHP) and BHP). For each of these variables, a desired value (or set point) is established.
Manipulated quantities: For each of the controlled quantities, there are a set of manipulated quantities that when
altered, effect the controlled variable (i.e., choke position, flow rate).
Disturbances: These quantities may be known or unknown, and as they enter the system, tend to drive the controlled
quantities away from their desired (set point) values.

Therefore, it is the function of the automatic system to adjust the manipulated variables to keep the controlled variables
near their set point values, despite the influence of disturbances. This process is conceptually illustrated in Fig. 3.



Fig. 3Control Process Overview.









4 IADC/SPE 151416
Types of Control and Automation
Having established the basis for automated control (i.e., the indirect control of BHP by the direct manipulation of applied
surface pressure, P
s
), the next question that arises is: what type of control can the systems actually provide?
Manual control has traditionally been the prevailing paradigm. This setup is illustrated in Fig. 4 below. Indicators
(sensors) are present, monitoring the control variables (typically WHP and BHP). This real-time feed is continuously
monitored by operators, and appropriate adjustments are made to achieve the desired affect or set point of the controlled
variable. The problems in this approach are inherently obvious and, for tight drilling margins, possess a further challenge,
especially because the set point determination and all the control decisions are relegated to operators in a non-deterministic
time domain.



Fig. 4Manual Control.

Though such control is still common and an important part of contingency and emergency management, a better approach
is required for an automated control workflow. This is where the feedback control paradigm has gained wide acceptance and
is the implementation of choice for all MPD system providers. In this approach, sensors are installed to monitor controlled
variables. These actual values are transmitted to a feedback control system, where a specialized controller makes automatic
determinations of divergence between desired values and actual values. This controller then calculates appropriate signals
that reflect the needed values of the manipulated parameters, which are automatically transmitted to control devices (choke).
One of the great advantages of this approach is that the system does not need to know in advance what types of disturbances
may affect the process. In fact, the system does not even need to know the ultimate effect on the controlled variable or
quantitative relationship between the disturbances, which makes it ideally suited to the ever-changing and unpredictable
drilling environment.



Fig. 5Feedback Control.

IADC/SPE 151416 5
As shown in Fig. 5, the effectiveness of feedback control is heavily reliant on how well the system manages and
influences the manipulated variable to achieve the desired controlled variable. This effectiveness is, in turn, a direct result of
the type of controller used. The controller is a special calculator that determines the changes that are needed in the
manipulated variable to compensate for disturbances that upset the process or for changes in set point. There are a number of
standard controllers available in the industry today; the following are the most popular continuous-mode controllers:
Proportional (P): This is the simplest of controllers, where the output is algebraically proportional to the input error
signal.
Integral (I): Also known as Reset control. As the name applies, integration is performed on the input error signal, and
the manipulated variable is adjusted at a rate proportional to the error. It can also be thought of as a historical count
(accumulation) of error. Any positive error will increase the count; any negative will reduce the count. The net count
at any time becomes the reset contribution to the manipulated variable.
Derivative (D): Also commonly referred to as Rate control. The output of this controller is solely based on the rate of
change of the input error signal.

In combination these controllers are referred to as PID. The controllers above are rarely used in isolation, rather particular
combinations have been found to exhibit excellent control ability. Therefore, variants of the above have been universally
adopted as de facto standards in the MPD automation industry, and each vendor uses a combination of the above in their
systems. The most common are either PI or a more sophisticated PID controller.
This control framework imparts incredible functionality and flexibility to systems, but at the same time also establishes a
foundation that is common to all systems. So, how does this map to what kind of control is actually possible? No matter the
provider, MPD automated systems provide the following basic control functionality or modes (or a variant thereof):
Stand-by: This is the fully manual mode in which the system continues to actively acquire data but does not perform
any automatic control actions (Fig. 6a).
Position: In this mode, the system simply adjusts the position of the choke to a desired position. This position
movement is automatic. No consideration is provided to resultant well conditions. All other actions are manual and
must be taken by an operator (Fig. 6b).
Surface Pressure: In this mode, the system adjusts the position of the choke to achieve a desired choke or surface
pressure. Convergence to the desired surface pressure is an automatic process, which usually relies on a feedback
control scheme. No consideration is provided to resultant well conditions. All other actions are manual and must be
taken by an operator (Fig. 7a).
BHP: This is the most complex mode of all. In this mode, a surface-pressure set point is calculated by a hydraulic
model, which is receiving real-time well-based information and data. The hydraulic model has a range of desired
BHP pre-configured, and using this information generates a value that should be held at surface to (theoretically)
produce the desired value at a particular depth. The system takes this pressure set point and then holds it at the
choke(s). Convergence to the desired surface pressure is an automatic process, which usually relies on a feedback
control scheme. In this case, resultant well conditions are taken into consideration as the hydraulic model is
continuously monitoring the effects of change and recalculating the set point (Fig. 7b).



Fig. 6(a) Stand-by, (b) Position.

6 IADC/SPE 151416


Fig. 7(a) Surface Pressure, (b) BHP.

The modes in Fig. 7 are all closely related and actually form a set of states that combine to form a state machine. The
system transitions through each step prior to moving into a state, which is at a higher level of automation. Such a transitional
approach not only provides efficient operation, but incorporates inherent safety by having checks prior to increasing
automation. Furthermore, if unforeseen problems arise, systems can safely demote to a previous state to regain control.
Though this is an extreme simplification, it provides an overview of how the system works.



Fig. 8Control Mode State Machine.

How Does It Work? (Control Overview)
Now that we have described what type of control can be done on MPD projects, let us shift focus by describing how this
control actually works as part of an automatic system. The best way to approach this is to first examine a generic architecture
and then trace the information as it flows through this setup.



Fig. 9Generic MPD Automatic Control System.
IADC/SPE 151416 7

Applying the feedback control philosophy introduced in the previous section, all MPD systems can be set up as illustrated
in Fig. 9. By monitoring the BHP pressure, we can make appropriate changes to the WHP to achieve accurate control. At first
glance, this setup may look reasonable, but investigating the details further shows that the time delay through the well makes
this setup not only impractical, but impossible for a direct PID type control scheme on BHP (even with the advent of
Intelliserv Intellipipe collars). One possible solution to this is cascade control, where one feedback control loop is nested
inside another, as shown in Fig. 10 below:



Fig. 10Inner and Outer Control Loops.

Measuring WHP pressure, comparing set point and measured values, calculating PID output, and signaling to the
hydraulic valve (used to control the choke) is called the inner controller loop. This control loop calculation is performed
several times a second. The outer control loop, on the other hand, includes the desired BHP, hydraulic model, and feedback
from downhole (PWD) tools. The goal of the outer control loop is to maintain the desired BHP and is the goal of the
complete system.
Putting all these pieces together, we can now explain how the entire process conceptually functions (Fig. 11). The entire
process starts at various data acquisition sensors, which transmit their signals to a dedicated data acquisition system (DAS),
whose primary function is to condition, convert, and communicate this information to an appropriate decision system. The
decision system comprises specialized interfaces and models, which are used to evaluate current conditions, perform
calculations, and accept user input. Resultant computations and information then flow to a set point module, which
establishes a preliminary WHP set point. This set point is then routed to a validation module, which confirms and finalizes
set point selection. This final value is transmitted back via the data acquisition system to a control module, which converts
this value into a controllable parameter and, using an appropriate feedback algorithm, physically actuates the choke to hold
the desired pressure. Once this loop is accomplished, the process starts all over again with the sensors detecting the effect of
the controllers change and feeding the system to generate another set point, which is then used to control the chokes. The
entire process is analogous to a wave, with information flowing in and results and commands flowing out. Each wave is
dependent on the results of the previous and improves system accuracy and control. Therefore, this process needs to be
repeated on a continual basis.


Fig. 11Conceptual MPD Automated Workflow.
8 IADC/SPE 151416
How Is It Done? (Control Implementation)
The next question that arises is how is this automation achieved? There are a variety of architectures and routes used to
provide this functionality, but irrespective of implementation there are several components common to all approaches. These
o two broad categories:
(i.e. equipment)
2. Software

ach in more detail. The Hardware category encompasses all the physical components
Control Device (RCD)
Choke

the annulus by restricting flow of the fluid through the choke. This is achieved using


Rig Pump Diverter (RPD)

ereof, collectively form the physical foundation for all automated MPD
perations. Lets briefly examine each in turn.
), which is a
re-requisite for all automation. The way in which a RCD functions and operates can be clearly seen in Fig. 12.

Fig. 12 Rotating Control Device (RCD).
core components can be divided int
1. Hardware
Only with a close coupling of hardware modules with associated supervisory software subsystems, can a successful
system be achieved. So lets examine e
required for automation and includes:
Rotating
Sensors
Controller
The above components are the minimum requirements for automation and assume the availability of a pressure source
(such as rig pumps) to achieve control. The choke(s) cannot be used to control backpressure applied to the annulus (for
control of BHP) unless fluid is flowing through the choke(s). Most often, this lack of circulation occurs whenever a
connection is made in the drill string. Conventionally, this would be handled by regulating the BHP solely by adjusting the
density of the drilling fluid. However, this is not sufficient in a MPD operation, so a technique is required where the flow of
fluid through the choke is maintained, even though the fluid does not circulate through the drill string and annulus. Thus, in
this case, pressure can still be applied to
the following secondary components:
Back Pressure Pump (BPP)
These components, or a combination th
o

Rotating Control Device. Or RCD for short, though usually not directly involved in the automation, is essential in forming a
closed-loop system (by dynamically sealing off the annulus and diverting all return flow to a designated choke
p


























IADC/SPE 151416 9
Sensors. Though many sensors can assist and augment automation, one sensor set is essential to all automation. The
upstream choke sensor, also commonly referred to as the wellhead pressure (WHP) sensor is the fundamental variable upon
which all critical automation relies. The WHP reading is particularly significant in the pressure control loop. To protect this
pressure control against failure, redundant readings are usually recommended, which are implemented as multiple sensors
installed at close proximity. The chosen controller system will then have internal logic to choose the most relevant/valid
alue, which, in turn, is used in all subsequent calculations. A sample dual WHP sensor setup on a MPD autochoke system
an be seen in the following Figure.

d duri ed to):
(downstream injection pumps)
njection pumps)
)
oke skid)
oke skid)
D) am BPP Coriolis meter (BPP skid)
pstream BPP Coriolis meter (BPP skid)
) CD (MWD tool)
d) HT (MWD tool)
)
hoke A position (Choke skid)
It controls automated components, such as solenoids, to drive
hokes and valves using a chosen control algorithm (e.g. PID). In addition, it acts as a conduit or translator to transfer (or
communicate) information between equipment and software.


v
c
















Fig. 13Pair of WHP sensor installed upstream of the choke.

Other common sensor readings that may be acquire ng MPD operations include (but not limit

Injection flow rate Hydraulic pressure sensor (Choke skid)
Stroke counter (i Rig air pressure Sensor (Choke skid)
SPP (stand pipe Fluid temperature upstream choke (Choke skid)
Hook load (rig) Flow rate downstream choke Coriolis meter (Choke skid)
Bit depth (rig) Fluid density downstream choke Coriolis meter (Ch
RPM (surface sensor on power swivel) Fluid temperature downstream choke Coriolis meter (Ch
WHP (RCD) Flow rate upstream BPP Coriolis meter (BPP skid)
Hydraulic oil diff pressure to WHP (RC Fluid density upstre
WHT (RCD) Fluid temperature u
U/S choke pressure 1 (Choke skid) BHP (MWD tool)
U/S choke pressure 2 (Choke skid E
U/S choke pressure 3 (Choke ski B
D/S choke pressure (Choke skid
C
Choke B position (Choke skid)

Controller. A generic controller system is illustrated in Fig. 14 below. The controller system is composed of a number of
functional blocks, which are usually implemented in a dedicated computer known as a PLC. A PLC is an industrial computer
used for automation of many industrial processes. Unlike general purpose computers, the PLC is designed for complex
inputs and output arrangements, immunity to electrical noise, extended temperature ranges and resistance to vibration and
impact. Programs to control machine operation are typically stored in non-volatile memory, such as a Flash memory card. A
PLC is an example of a real-time system since output results must be produced in response to input conditions within a
bounded time; otherwise unintended operation will result (Wikipedia 2009). The PLCs primary functions are to monitor,
control, and communicate. It monitors data from sensors (such as on-skid pressures, temperatures, and flow rates) relaying
information pertaining to the drilling and automation process.
c
10 IADC/SPE 151416


Fig. 14Generic Controller System (using the PID control algorithm).

Many different architectures can be used to implement a PLC arrangement, but to achieve efficient control, a master-slave
device philosophy is usually adopted. In this setup, the main choke PLC is set as the Master PLC. This PLC not only
monitors sensor information, such as on-skid pressures, temperatures, and flow rates, but also regulates a provided pressure
set point (i.e., WHP directly, BHP indirectly) by controlling and driving the choke position using a PID control algorithm.
The master PLC also coordinates other hardware subsystems (such as the continuous choke flow controller) to provide an
integrated, automated control system. It is important to note that the master PLC is the only gateway into the software process
and shields all hardware subsystems from this abstract domain. It also independently drives the entire automation process,
thereby providing complete redundancy in the case of software failure. Internally, the PLC can be functionally divided into a
number of modules or blocks:

Set point validation block: To reduce the chance of a big step change in pressure set-point change, the set point
coming to the pressure control loop is restricted to a maximum-allowed change per second. This means that if there is
a big step change in pressure set point, the set point given to the pressure control loop will be a ramp. This regulates
control and provides for smoother and more accurate control.
WHP validation block: WHP is also validated concurrently to ensure control integrity. If any problems are found
with the value, the redundant reading is used. If there is a problem with the redundant reading, the system demotes
from automatic control to user-led control.
PID block: The desired pressure is compared with the measured pressure, and the deviation, or error "e", is used in
the PID algorithm to calculate the response of the hydraulic control valve (which ultimately controls the actual choke
actuation).
Boost block: The Boost Block is there to avoid a dead band that will otherwise be present in the control.
Pulse Gen block: The Pulse Gen Block is needed if a solenoid-type of hydraulic control valve is used. The purpose
of the Pulse Gen Block is to translate the output from the PID block into pulses that are proportional to the output.
This means that a 50% positive output from the PID block gives half the speed in the opening direction, 100%
negative output gives full speed in the closing direction, and so on. If a proportional valve is used in the system, the
signal from the Boost block will be going directly to an Analog output (Ao) card in the PLC.

A sample controller implementation is presented in Fig. 15. The two metal boxes at the forefront of the figure enclose an
advanced PLC and barrier system (housing components illustrated in Fig. 12 above), coupled to an on skid hydraulic control
panel.



IADC/SPE 151416 11

















Fig. 15PLC and HCP arrangement.

Choke. When the industry thinks of an automated managed pressure drilling system, the thought is typically of software
components. Little consideration is placed on the mechanical configuration. Because the automated choke is the pressure
regulator for the closed-looped drilling system, the choke trim selection will determine how precisely the chokes regulate the
bottomhole pressure. The precision of the chokes is dependent on the operating environment, and from the pre-well
engineering, the choke is fitted with the proper trim size for the job. Sizing the trims or flow orifice is important in that each
trim size has a controllable area that is a function of the coefficient of flow (Cv) profile of the trim size, which is a derivative
of the chokes geometry. The following equation defines Cv as:

Flow Rotc (q) |Spcciic 0ro:ity (sg) Prcssurc rop (Jp) ]
1 2
(2)

With the particular geometry of gate seat chokes, there are regions where small movements in position while closing will
have profound pressure changes, and as the choke continues to move toward closing, the flow will be restricted to the point
of no flow, even with the choke partially open. Conversely, as the choke moves toward full-open, there is a region of the
movement that has very little effect on the pressure. For these two reasons, the trims are selected for the operational flow
rates, fluid density, operational pressures, and where these parameters fall within the Cv profile. Fig. 16 shows a graphical
representation of the Cv profile and operational window for a 1--in. trim with fresh water.



Fig. 16Cv Profile and Operational Window for a 1--in. Trim with Fresh Water.

Again, to enhance the automation of the system, the flow rates and pressures will be derived from the pre-well
engineering hydraulics modeling. This modeling will not only define the circulating pressure for a project, but must also take
into consideration the connection pumps off and surge/swab pressures for trim size selection. When tripping or making
connections, flow across the choke must be maintained to keep the chokes within the controllable area of the trims. Typically,
12 IADC/SPE 151416
continuous choke flow is provided with a dedicated offline pump or with a rig pump diverter system. The only requirement
for these devices is that they deliver a pressure at a flow rate that is within the controllable area of the Cv profile.



























Fig. 17(a)Top left: Autochoke, (b)Top right: Opening choke, (c) Bottom left: Choke body, (d) Bottom right: Choke internals.

Back Pressure Pump. As mentioned earlier, the choke cannot be used to control backpressure applied to the annulus for
control of the BHP, unless fluid with a direct connection to the annulus is flowing through the choke. In conventional
overbalanced drilling operations, a lack of circulation can occur whenever a connection is made in the drill string (e.g., to add
another length of drill pipe to the drill string as the wellbore is drilled deeper). This loss of fluid (and subsequently control)
has traditionally been handled by MPD systems by using a dedicated pump, known as the backpressure pump (BPP), which
in place of the rig pumps, can be used to supply a flow of fluid to the return line upstream of the choke manifold by pumping
fluid into the annulus when needed (such as, when connections are being made in the drill string). The BPP has several
distinct advantages in that it provides a redundant source of flow (& pressure) which enables automatic control to continue
even when injection operations and rig pumps are stopped. But the BPP also has a number of serious disadvantages.
Primarily, it introduces additional complexity of control into the system (especially during the transitional phases of starting
and stopping the pumps), which if not managed correctly can lead to unacceptable BHP fluctuations.


















Fig. 18(a)Back pressure pump (BPP), (b) Sample connection showing transition from rigs pumps to BPP.

IADC/SPE 151416 13
Rig Pump Diverter (RPD). To counter some of the inherent disadvantages of the BPP and provide more accurate granular
control during periods of no injection, a technology was devised (RPD) whereby fluid could be diverted from the standpipe
manifold to the return line when needed, without the need of a separate pump. Restriction by the choke of such fluid flow
from the rig pump will thereby cause pressure to be applied to the annulus and enable continuous automatic control. The
RPD is a dedicated skid, which using a combination of coordinated valves and chokes diverts flow from the drill string to the
annulus (upstream of the choke) in a completely automated manner. The RPD has several distinct advantages over traditional
BPP approaches, by removing the synchronization requirement of managing two separate pumping sources.

















Fig. 19 Rig Pump Diverter.

Equipment in isolation cannot provide the required functionality, level of control, and accuracy that is required to drill
tight margined MPD wells. Rather, the equipment must be unified under a supervisory framework which coordinates,
communicates, and control operational activities. This framework is made up of software applications and dedicated
hardware appliances. Having detailed the hardware above, lets examine the software components of the system, which can,
itself, be divided into the following logical segments:
Data Source and Control
Data Acquisition
Data Storage and Distribution
Data Computation and Visualization























Fig. 20Conceptual Software Overview.

14 IADC/SPE 151416
The Data Source and Control segment comprises the components, which acquire data (such as sensors), serve data to
other processes, receive feedback, and perform actual control actions. The entire process is conceived at the sensors, where
data is acquired and transferred as electrical signals to the PLC. The PLC transforms this electrical information to digital
data. This data then needs a mechanism by which it can be transferred in real time to other processes. This is achieved by
using a Communication protocol.

Communication Protocol. As shown above, the equipment collects and collates a lot of physical information. Furthermore,
this information does not only originate from within the MPD system, but comes from a variety of sources, including rig
system, mud loggers, directional drillers, etc. All this information plays an important role in accurate control. However, to
effectively access this information in an efficient and timely manner, a suitable communications protocol needs to be chosen.
The requirements for this protocol include:
The ability to handle large amounts of data in a near real-time domain
To provide seamless interaction between systems (whether they be hardware or software)
Feature the ability for full bi-directional communication
Ensure security and stability

There are a number of protocols, both standard and proprietary, that fit this build, but one in particular has seemed to gain
wide acceptance, with most providers implementing it within their systemsthat is OPC. OPC is a standardized protocol
(language), which defines a set of rules that govern the interaction (transfer of information) between systems. Apart from
meeting all the requirements above, another big advantage of OPC is that it is an open standard. Because of this fact, many
vendors implement OPC communication interfaces, allowing systems to communicate with minimal configuration.
In a MPD automated system, OPC is usually used in communicating information between the PLC (or data logger) to the
data acquisition systems. It can also effectively be used for communication with the decision management system, human
machine interface (HMI), and hydraulic model.
This protocol discussion so far has only focused on inter-system communication; how about communication between
external systems? OPC can definitely fulfill these needs as well. But, as OPC is still relatively new to drilling systems, not all
contractors have systems that support it. Therefore, it is found that for external communication, companies usually fallback to
established/legacy protocolsthe most popular of which is WITS. So in summary WITS, is generally used to transfer
physical/sensor information in a uni-directional manner between systems. Whereas, OPC is used to transfer physical/sensor
information, control and command information within a system.
The Data Acquisition segment processes, computes, and analyses the acquired data. This segment usually provides the
only entry point (communication conduit) to control equipment (i.e. master PLC) and is where the majority of system-level
conversation takes place. To enable easy user interaction and supervision a mechanism is required for accepting and
processing user instructions and commands. This requirement is fulfilled by a human machine interface .

HMI. Is implemented to provide a convenient centralized control panel (or dashboard) to the entire system. This has to be
done in a manner that is not only effective/efficient, but also simple, intuitive, and user friendly. The HMI not only provides
both a display of acquired and calculated information, but also provides interactive control elements in one unified location.
This is vital, as it allows operators to control all aspects of the operation, from selecting computational algorithms (that
perform advanced hydraulic computation), to controlling valves and chokes, to setting levels of automation. From here, the
data flows to the Data Storage and Distribution segment.

IADC/SPE 151416 15

Fig. 21HMI.

The Data Storage and Distribution segment is where data is marshaled, consolidated, and organized into specialized
structured databases. To provide data redundancy and computational efficiency, a primary-secondary architecture may be
implemented. Data is received into a primary database, which processes all input parameters and is then replicated to a
secondary database, which is used to provide a redundant archive as well as a base for further computationally-intensive
processes. Once in the database, the data may be manipulated and distributed via real-time applications to a variety of
destinations ranging from explosion-proof monitors around the rigsite to real-time operations centres (RTO). From this
segment, data flows to the Data Computation and Visualization segment.
The Data Computation and Visualization is where acquired data is fed into customized hydraulic models to determine the
surface pressure set-point that is required to provide a pre-defined bottomhole pressure.

Hydraulics Model. The next vital piece of the automation puzzle is the hydraulics model. The hydraulics model forms the
heart of the decision system and is not only the major enabler in automation, but is also the major limiting factor in the level
of automation that can be achieved.
A hydraulic model is a mathematical model that tries to emulate and encapsulate physical phenomena in a well. Once set
up, it calculates all the components in the BHP equation (in addition to a whole host of physical parameters), thereby
enabling computation of a suitable WHP set point (P
s
) to attain a desired BHP. The hydraulics model is the missing piece,
which allows us to link the directly controlled surface pressure parameter (P
s
) to our indirectly-controlled downhole pressure
parameter. If this link is accurate, then our control will be accurate, but if it is not, it will have a major impact on the accuracy
of the system.
A model is inherently dependent on the quality of information fed into it. So what other factors need to be considered
when selecting a model for MPD? The first and foremost is complexity. Usually, the more complex a model, the more
physical phenomena it will have taken into account so usually the better the results it will calculate. But with added
complexity comes longer computational times. This is usually not an issue when doing pre-engineering, where some models
make multiple iterations to come up with answer sets/iterations that can take significant time to complete (from tens of
seconds to minutes). But, such a delay is a huge issue for a time-bounded control system, which requires a valid set point in a
discrete near real time manner. Therefore, a balance has to be made between model complexity and computation. As part of
this, the vast majority of real-time MPD models are single phase. But, as one can see, this also limits when the model can be
used. If a second phase is introduced into the annulus, the model immediately loses its validity, thereby nullifying automation
and requiring user intervention.
Another point to consider is that model selection is the debate between static and transient models. The vast majority of
MPD models are static in nature, but approximate transient behavior by being run continuously.


16 IADC/SPE 151416
Other points to consider include:
Data input requirements (Fig. 22)
Validation
Fault tolerance
Calibration requirements
Communication interfaces

In the end, it is important to note that a model is simply a model and may not reflect actuality (owing to a vast set of
assumptions underlying their foundations). This is where having systems use actual acquired/measured data, such as PWD, to
calibrate/correct in real-time is essential. Furthermore, models are usually event (or context) unaware. By that, we mean that
models do not know if they are taking a kick, experiencing a loss, etcas a result, they cannot be relied on (in isolation) to
drive full automation.























Fig. 22Sample data requirements for a Hydraulic model.

Once determined, the set point is then transferred back through the system to the control valves, i.e., travelling through
the Data Storage and Distribution segment to the Data Acquisition segment (where it is validated and transformed if
required) to the Data Source and Control segment (via OPC) and is transformed by the PLC into a control action by
translating the pressure set point to a choke position, which provides the set-point pressure (and ultimately the desired
bottomhole pressure).
Having established how a set point is used by the system to control bottomhole pressure, the next major question that
arises is how is the set point actually determined? This begins early in the projects lifecycle in the pre-engineering (front end
engineering) phase. Preliminary models are run to establish an operational window, bounded on one side by pore/collapse
pressure, and on the other by fracture pressure. But for real-time MPD control, such a static envelope is only of very limited
value owing to the dynamic nature of MPD operations. To provide accurate and efficient control, a static model cannot be
used in isolation, neither can isolated sensor information, such as PWD. Rather, a specialized real-time hydraulics model is
used, which continually evaluates acquired data and dynamic conditions and computes new set points at predetermined times.
But, as with all models, the real-time MPD control hydraulics model has some inherent deficiencies resulting from
assumptions that the models are based on, including pumps operating at an ideal efficiency (e.g., 100%), accuracy of acquired
data (density, rheology), ideal borehole conditions (e.g., no collapse), and a theoretical fluid system. The real-time models
also may not take into consideration important physical principles, such as transport efficiency. To try and counteract such
weaknesses, a number of approaches have been used to add additional mechanistic and empirical computations to cover all
practical scenarios. In theory, this works well, but adding further complexity to a real-time hydraulics model can have serious
performance implications, which can result in the model being unable to generate viable results in a bounded time essential
for a control application. To alleviate this serious condition, the hydraulics model results may be corrected with specialized
algorithms using acquired data (e.g., PWD). This method inherently models natural occurrences, without the need to have
discrete mathematical models for all possible physical phenomena. These corrected results are then forwarded to the PLC (as
above) for control.

IADC/SPE 151416 17
Level of Automation
This is one of the most confusing aspects of MPD automated systems today. Contrary to urban (marketing) legend, there
are no truly fully automated MPD systems on the market today. This is not only a result of limitation in decision systems
(hydraulic model, event detection, etc), but also a result of the unpredictable nature of drilling operations, where not all
scenarios may be correctly identified/handled and modeled. Another limiting factor is that true full automation also cannot be
realized until rig systems are fully integrated and controlled by the MPD automation frameworkwhich today is usually not
the case.
Having now covered the automated work flow, let us examine the level of automation and the factors that influence it.
This can best be seen in the figure below which lists inputs to the systems. As shown, there is a mix of automatic processes
and manual workflows. From this alone, one can see that it is difficult for the entire process to be fully and truly automated.
Further complications arise with a variety of well events. Normal well phases, like drilling and connection, are easily handled
by the automated systems, with transitions also being handled automatically. But, well control situations, like kick and loss
events, are more of a challenge. As the systems are highly focused on controlling WHP, if these events are not recognized
and analyzed early, systems can, in fact, react incorrectly, making situations worse. Hydraulic models play a large part in
determining an accurate set point, but hydraulic models are based on certain technical assumptions and have inherent
limitations. They, in isolation, are usually incapable of identifying and rectifying all well conditions. Rather, as we head
down the path of full control, an autonomous real-time event detection and action system is vital. Such a system would work
in conjunction with the hydraulic model to analyze well events and, once identified, take appropriate actions based on
operational guidelines. Fig. 23 shows the mix of manual and automatic workflows.



Fig. 23Automatic and Manual Processes.














18 IADC/SPE 151416
Case Study (Example)
Let us now look at a concrete example of an actual MPD automated system. The system setup can be seen in the following
figures.

























Fig. 24Simplified Process Flow Diagram (PFD).



Fig. 25Software Overview / Data Flow Diagram.

The setup consists of a RCD to seal the annulus with returns diverted to an autochoke system. Upstream of the autochoke,
the discharge line of the BPP is piped in to provide flow and pressure continuation during connections. Following the
autochoke, the fluid stream is channeled through a Coriolis flow meter which not only measures mass flow, but gauges fluid
density, which in turn is used to calculate a flow rate. Following the flow meter all fluid is sent to fluid processing and
IADC/SPE 151416 19
handling facilities where it is then either recirculated to the rig pumps for injection down the drill string, or routed via the
BPP upstream of the choke.
Information and data is acquired and flows through the system as described earlier, with specific implementation details
provided in Fig. 25. A notable difference to the generic setup presented earlier is that the system is not limited to a single
hydraulics model, but employs a plug-in type architecture, thereby enabling the hydraulic model to be chosen based on
project requirements (e.g., single phase, multi phase).
So, what were the results for such a setup? Many scenarios and tests were run, with selected results presented below.
Overall, the system does an excellent job of maintaining BHP in a variety of scenarios from changing rig flow rates, to
connections, to more challenging surge and swab. The most remarkable feat is that pressure control for this project is usually
within 35 psi, which would be virtually impossible in a manual setup.

Test A. The ability to keep BHP at desired pressure during stable circulation rate (Manual Choke Control Mode).



















Fig. 26Test A results.

The test showed that the system can be easily controlled Manually in the field, and BHP can be kept well inside 35 psi
for steady state conditions. The test also showed that the hydraulic model used gave good offline model results to be used
with Manual Control Mode.

Test B. The ability to maintain desired BHP during stable circulation rates, in Choke Pressure Control Mode. Verify the PID
tuning parameters through step changes of desired pressure at various injected rig pump flow rates.


















Fig. 27Test B results.

The test showed that the system controlled pressures within 35 psi both for Choke pressure (ChkP) and BHP while in
Surface Control Mode in steady state conditions, and that during 70 bar step changes the ChkP remained within 35 psi. The
20 IADC/SPE 151416
test confirmed that the PID settings used gave accurate pressure control with only moderate overshoot and a reasonably fast
response.

Test C. The ability to keep a desired BHP during stable rig pump circulation rates, in BHP Control Mode.


















Fig. 28Test C results.

The test showed that the system maintains BHP within the 35 psi when controlled in BHP Control Mode using the real
time hydraulics model as input, for steady state conditions.

Test D. The ability to keep desired BHP during variation of rig pump circulation rates, in BHP Control Mode.


















Fig. 29Test D results.

The test confirmed that changes in flow rate through the chokes present a challenge to the system. The test where the flow
rate was changed from 264 gpm (1000 lpm) to 528 gpm (2000 lpm) in 60 s fell slightly outside 35 psi in BHP control. The
max deviation for ChkP was 46 psi and for BHP was 52 psi. The test where the flow rate was ramped in 120 s was inside 35
psi both for ChkP and BHP. It should be expected that the limit for the system to remain inside the criteria would be a ramp
time of between 60 to 120 s. The time for ramping up and down the rig pumps is a critical factor for the Autochoke system to
keep BHP inside a tight pressure window.





IADC/SPE 151416 21
Test E. The ability to keep desired BHP during variation of Back Pressure Pump injection rate, while also circulating with
the Rig Pumps in BHP Control Mode.



















Fig. 30Test E results.

The tests with 120 and 60 s ramping time up and down of BPP were performed within the BHP criteria of 35 psi. During
the test with 30 s ramping time of BPP ChkP fluctuated up to 45 psi and BHP up to 52 psi. To enable improved pressure
control during rapid BPP flow rate changes alternative Autochoke control algorithms may provide better results.

Test F. The ability to keep desired BHP during a scheduled connection, in fully automated BHP Control Mode.



















Fig. 31Test F results.

In the initial phase of the connection where the rig pumps are ramping down the ChkP and BHP were both 6 psi outside
the criteria of 35 psi. In the rig pumps off phase both ChkP and BHP were inside the criteria. When the rig pumps come on
and ramp up after the connection, ChkP spikes up to 22 psi and BHP spikes up to 32 psi outside the criteria were noted. This
is mainly due to manually controlled diesel frac pumps being used as rig pumps. These were very hard to control smoothly.
After the initial spikes in the pressure the control was inside the criteria of 35 psi. Almost any pressures falling outside the
test criteria on this test can be attributed directly to the inherantly poor control of the rig (frac) pumps.






22 IADC/SPE 151416
Test G. The ability to keep desired BHP during a connection where rig pumps are controlled manually and Back Pressure
Pump ramping is manually initiated, while Autochoke is in BHP Control Mode.



















Fig. 32Test G results.

During this test the ChkP was 16 psi and the BHP 19 psi outside the criteria of 35 psi. This was caused by the rig pumps
falling directly from 53 gpm (200 lpm) to zero when the pumps were coming to the end of the ramping down phase. When
the rig pump was started it was impossible to get a smooth start to the ramping phase, due to poor control of the pump, and
this led to an initial pressure spike during start up. ChkP was up to 36 psi and BHP 49 psi outside the criteria. To enable
improved pressure control alternative Autochoke control algorithms may provide better results.

Test H. The ability to keep desired BHP during a connection where Chokes, Backpressure Pump and Rig Pump are
controlled manually.



















Fig. 33Test H results.

The test result both from the ramping down of rig pumps and ramping up of rig pumps shows that the operator kept ChkP
and BHP inside the criteria of 73 psi. The result also shows that the hydraulic model works very well as an offline model
and the pressure versus flow rate calculations that were made before start of test were correct.







IADC/SPE 151416 23
Test I. The ability to keep desired BHP during controlled changes in drill string rotation, in BHP Control Mode.


















Fig. 34Test I results.

The system was inside the criteria of 35 psi throughout the test.

Test J. The ability to keep desired BHP by compensating for swab and surge effects during movement of drill string, while
circulating using Rig Pump, in BHP Control Mode.



















Fig. 35Test J results.

All the tests were inside the criteria of 35 psi of ChkP. Also for BHP the test result was inside the criteria for all the tests
where the hydraulic model worked correct. During the test the model ended up in a non-physical state where the bit depth
was depeer than the well depth. This led the model to calculate incorrect predicted pressures in parts of this test. To prevent
faulty data to influence the model it is suggested to implement a mechanism that outputs an alarm/warning in the event of
situations such as bit depth value being larger than well depth value.









24 IADC/SPE 151416
Test K. The ability to keep desired BHP by compensating for swab and surge effects during movement of the drill string,
while circulating across the well head using the Back Pressure Pump, in BHP Control Mode.



















Fig. 36Test K results.

The test of 39 ft/min was inside the criteria of 35 psi for ChkP. Data shows that also BHP is inside 35 psi. Due to the
test setup the hydraulics model ended up in a non-physical state where the bit depth was deeper than the well depth. This led
the model to calculate incorrect predicted pressures in parts of this test. As a lesson learned to prevent faulty data to influence
the model a mechanism that outputs an alarm/warning in the event of situations such as bit depth value being larger than well
depth value.

Test L. The ability to keep desired BHP during displacement of multiple weight fluids to simulate a Balanced Mud Cap
spotting operation, in BHP Control Mode.



















Fig. 37Test L results.

The test had to be repeated due to incorrect mud weights being used in the first attempt. In the second (and also first) test
ChkP was inside the criteria of 35 psi, but the BHP was 20 psi outside the criteria. Before the second test the mud weight
was measured using a mud balance; however it can be questioned as to how accurate the mud weight was for the mud pill. 55
psi corresponds to 0.04 sg and an error of this size can be plausible for the given conditions during the test. During the test
one of the rig pumps failed several times. This led to pressure fluctuations, however since this was not part of the test these
pressure fluctuations are not included when the result from this test is evaluated.



IADC/SPE 151416 25
Test M. The ability to keep desired BHP during unplanned loss of power to rig pumps with BPP ready and available, in BHP
Control Mode.



















Fig. 38Test M results.

The tests show that the impact of suddenly losing the rig pumps is very difficult for the system to handle, and that starting
the BPP to compensate for the lost circulating will involve a BHP drop for up to 20 s before the pressure is back inside
criteria of 73 psi. The ChkP dropped 267 psi and BHP dropped 277 psi when the pumps were stopped while running at 528
gpm (2000 lpm). The pressure dropped up to 254 psi for BHP and 241 psi for ChkP when the pumps were stopped when
running 396 gpm (1500 lpm). The pressure dropped up to 122 psi for BHP and 115 psi for ChkP when the pumps were
stopped while running at 264 gpm (1000 lpm). Suggestions to improve the system is proposed, however it should be
discussed if a higher deviation can be allowed or if emergency trapping of pressure should be used even if the BPP is
available in case of emergency stop of the rig pumps.

Test N. The ability to trap pressure during unplanned loss of power to rig pumps when Backpressure Pump is not available,
in BHP Control Mode.



















Fig. 39Test N results.

The test of emergency trapping of pressure during an event of sudden rig pump failure passed both the test with initial rig
pump flow rate of 528 gpm (2000 lpm) and 396 gpm (1500 lpm). In both tests the trapped pressure exceeded the criteria of
minimum trapping 50 % of initial pressure. This only tested the emergency pressure trapping function. However the
triggering of this function will need to originate from the Rig system and will require further work to be clearly defined.


26 IADC/SPE 151416
Test O. The ability to trap pressure during unplanned loss of power to rig pumps when Choke is controlled manually and
Backpressure Pump is not available.



















Fig. 40Test O results.

The manual control from choke panel in the field was performed by a trained operator and the results were inside the
criteria of trapping a minimum of 50 % of initial pressure. This was the case for the test with initial flow rate of 528 gpm
(2000 lpm) and 396 gpm (1500 lpm). The testing showed that some training is needed especially for when a scenario of a
sudden stop of Rig pumps.

Conclusion
Automated MPD systems have truly heralded in an era on accurate pressure control enabling drilling of wells with tight
drilling envelopes. Systems have evolved from crude, clunky experiments to professional, sophisticated setups that can
handle a wide variety of conditions. But, it is still important to note that with all the advancements of today, there is still need
for more. Todays systems provide supervised automation, where certain phases can be handled automatically and certain
phases require operator intervention. The further advancement of hydraulics model, event detection and, action systems is
driving us towards a cruise control/start and forget paradigm.

References
McMillan, S., An, W.H., Zhang, Y.M., Xu, Z.X., Lovorn, R., Ospina, J., and Saeed, S. 2010. Overcoming Drilling Challenges in Northwest
China Using an Innovative MPD Technique. Paper SPE 130317 presented at the SPE/IADC Managed Pressure Drilling and
Underbalanced Operations Conference and Exhibition held in Kuala Lumpur, Malaysia, 2425 February.
Wikipedia contributors, "Programmable logic controller," Wikipedia, The Free Encyclopedia,
http://en.wikipedia.org/w/index.php?title=Programmable_logic_controller&oldid=316946109 (September 29, 2009).
Rachel Johnson, Halliburton Energy Services; Julio Montilva, Shell Exploration & Production Company; Mohamed Sati, Jeff Grable and
Saad Saeed, Halliburton Energy Services; and Richard Billa, William Derise, Shell Exploration & Production Company. Field
Demonstration of a New Method for Making Drillpipe Connections During Managed Pressure Drilling Operations. Paper 147278
presented at the SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 30 October2 November 2011.

Vous aimerez peut-être aussi