Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
) in series, using a
gas control valve for controlling GLCC liquid level and liquid control valve for
controlling LLCC underflow watercut, was studied experimentally and theoretically to
investigate its performance as a three-phase oil-water-gas separator.
Experimental data acquired for the GLCC
/LLCC
/LLCC
/LLCC
) 2
Figure 1.2 Schematic of the Liquid-Liquid Cylindrical Cyclone (LLCC
) 4
Figure 1.3 Schematic of Two-Stage GLCC
and LLCC
)
1
, as
shown in Figure 1.1. The GLCC is a simple, compact, low weight and low-cost
separator. It is a vertical pipe section, mounted with a downward inclined, tangential inlet
located approximately at the middle. Neither moving parts nor internal devices are used,
reducing the need for maintenance. Separation in this equipment is achieved by
1
GLCC
)
Mechanistic models for design and performance prediction of GLCC have already
been developed (Gomez, 1998, Gomez, L.E., 2001) and are in use by the industry. In
these models, the oil-water mixture is treated as a single liquid-phase flow. Also,
strategies for the GLCC liquid level and pressure control have been developed (Wang,
3
2000, Wang et al., 2000). Following the theoretical development of the GLCC control
strategies, field implementations have demonstrated the concept validity. The GLCC has
recently gained popularity in the industry, with more than 500 units installed in the field
around the world.
Liquid-Liquid Separation: The Liquid-Liquid Hydrocyclone (LLHC) is utilized
by the industry to clean produced oily water for disposal, reducing oil concentrations to
levels below 40 ppm. This equipment is suitable for cleaning water with low oil content.
Attempts have been made in the past to utilize cylindrical hydrocyclones for oil-water
separation. The use of cylindrical hydrocyclones for oil-water separation has been
hindered due to the fact that at high velocities they perform as mixers rather than
separators. However, by operating at moderate velocities, the cylindrical hydrocyclone
can be used to perform at least partial oil-water separation (free-water knockout).
Recently, studies have been conducted (Mathiravedu, 2001) on the performance
of Liquid-Liquid Cylindrical Cyclone (LLCC
)
2
as a free-water knockout. The LLCC
has a similar configuration as that of the GLCC, namely, a vertical pipe section, but with
a horizontal tangential inlet, as shown in Figure 1.2. The horizontal inlet promotes oil-
water stratification. The liquid phase mixture enters the vertical section through a
reducing area nozzle, increasing its velocity. The swirling motion in the LLCC produces
a centrifugal separation, whereby the oil phase moves to the center, and an oil-rich stream
exits through the top (overflow). The water moves to the pipe wall, flows downward and
exits through the bottom (underflow). Mathiravedu (2001) has also demonstrated the use
of a unique quality control strategy in order to ensure the condition of maximum clean
2
LLCC
)
Gas-Liquid-Liquid Separation: The GLCC
and LLCC
and LLCC
separators.
In this configuration, the three-phase gas-oil-water mixture enters the GLCC, the first stage
5
device, through the inclined tangential inlet. The gas flows to the top of the GLCC and
exits out of the system. The liquid, an oil-water mixture, flows through the GLCC liquid
leg into the second stage LLCC, where the oil-water separation occurs.
In the proposed combined GLCC
/LLCC
and LLCC
/LLCC
/LLCC
).
Wang et al. (2000) developed a unique optimal control strategy, capable of
optimizing the GLCC operating pressure. Detailed simulations and experimental
investigations were conducted to evaluate the performance of the proposed optimal
control system. The significant advantages of this strategy (Wang, 2000) are: The system
can be operated at optimum separator back pressure; the system can adapt to the changes
of liquid and gas flow conditions; and finally, the strategy can be easily implemented
using simple PID controllers available in the market.
LLCC control dynamics have been recently studied both theoretically and
experimentally by Mathiravedu (2001). A unique control strategy has been developed and
12
implemented, capable of obtaining clear water in the underflow line and maintaining
maximum underflow (optimal split ratio). A linear model has been developed, for the
first time, for LLCC separators equipped with underflow watercut control, which enables
simulation of the system dynamic behavior. Comparison of simulation and experimental
results showed that the control system simulator is capable of representing the real
physical system and can be used to verify the controller design and dynamic behavior.
2.4 Watercut Measurement
The measurement of water content in crude oil is an important and widely
encountered practice in the petroleum industry. The watercut measurement is utilized in
multiphase flow meters. Also, monitoring watercut at various points throughout a
processing facility may optimize the separation efficiency in production operations. A
review of available watercut meters is reported below.
Agar (1988) developed a watercut meter using microwave method, by measuring
the energy absorption properties of an oil-water mixture. Lew (1988) developed a method
and a device for determining the concentration of each phase in an oil-water mixture
utilizing Nuclear Magnetic Resonance (NMR) analysis. In this method, a direct and
accurate measure of the desired component, oil for example, can be achieved on a real-
time basis in the field, without the need to interrupt operations.
Durrett et al. (1989) developed a watercut monitor that uses microwave principle
to measure the watercut in a multiphase flow stream. Measurement accuracy was
maintained despite changes in temperature, salinity, crude properties and the presence of
13
gas. Gaisford et al. (1992) used Radio Frequency (RF) bridge technique to determine the
composition of oil and water in an oil-water mixture. The device is operated by using the
metal pipe of the process stream as an electromagnetic waveguide.
Cobb (1995) developed a method and an apparatus to monitor the composition of
a fluid mixture traveling through a conduit, using ultrasonic propagation. These
measurements were used to determine fluid mixture composition based on a relationship
derived from measurement of samples of the fluid mixture. Al-Mubarak (1997) proposed
a new method for calculating watercut using a Coriolis device such as Micromotion
mass flow meter. The method provided reliable well testing results that are comparable
with a three-phase conventional test facility, when there is no gas present in the
multiphase mixture. Recently, Lievois (2000) developed a narrow band infrared watercut
meter that can detect a full watercut range of a flow stream.
As shown above, there is no previous work regarding the integration process of
separators. However, the work of Contreras (2002) is very helpful since it considers one
of the most important factors in the cascaded configuration, which is the effect of the gas
carry-under coming from the GLCC into the LLCC, on the LLCC performance. The
scope and contribution of the present study is to assemble previous models and
experimental studies performed for each separator separately, and construct a coupled
model for the GLCC/LLCC system.
14
3. MATHEMATICAL MODEL
The mathematical model for the GLCC that was developed by Wang (2000)
consists of two parts, namely GLCC model and control system model. The GLCC model
is developed based on gas and liquid mass balance equations, flow behavior of the
respective phases in the GLCC, and pressure drops across the gas and liquid legs. The
control system model facilitates the design of the controllers required to optimize the
separator.
The LLCC model, developed by Mathiravedu (2001), is based on the water
concentration in the underflow as the measuring parameter. The methodology for LLCC
control system is established as a design tool, and simulators are developed using
Matlab/Simulink
=
2
(3.1)
where
L
C is the overall flow coefficient of the liquid leg that is given by,
18
(
(
+ =
= =
n
i
m
j
Lj
j L
Li
i L i L
L
d d
L f
C
1 1
2 4 2 5
8
8
(3.2)
LCV
P is the pressure drop across the liquid control valve, which can be solved from the
liquid control valve flow rate equation (Fisher, 1998), as follows
( )
L
LCV
v Lout
P
C Q
= 002228 . 0 (3.3)
Solving
LCV
P from equation (3.3) gives
( )
2 2
2
002228 . 0
v
L Lout
LCV
C
Q
P
= (3.4)
Substituting equation (3.4) into equation (3.1) yields an expression for the total pressure
drop across the liquid leg of the GLCC, namely,
+
=
c
L Lout L L
Lout
g
gH Q C
P P
2
( )
2 2
2
002228 . 0
v
L Lout
C
Q
(3.5)
Taking the derivative of equation (3.5) with respect to time, and assuming the
liquid discharge pressure,
Lout
P , to be constant, gives an expression for the rate of change
of the GLCC pressure caused by the change of the liquid control valve position, namely,
19
( )
( ) ( )
( )
2
2 2
2 2 2 2
002228 . 0
002228 . 0 2 002228 . 0 2
2
v
v
L Lout v v
Lout
L Lout
c
L
Lout
Lout L L
C
dt
dC
Q C C
dt
dQ
Q
g
dt
dH
g
dt
dQ
Q C
dt
dP
|
|
.
|
\
|
+
=
(3.6)
Gas Leg Pressure Drop. The pressure drop across the gas leg is given by
GCV
c
G Gout G G
Gout
P
g
gH Q C
P P +
=
2
(3.7)
where
G
C is the overall flow coefficient of the gas leg that is given by,
(
(
+ =
= =
n
i
m
j
Gj
j G
Gi
i G i G
G
d d
L f
C
1 1
2 4 2 5
8
8
(3.8)
GCV
P is the pressure across the gas leg of the GLCC, which can be solved from the gas
control valve flow rate equation (Fisher, 1998),
( )
Deg
GCV
g
G
Gout
P
P
C
C P
T
Q
|
|
.
|
\
|
(
=
1
3417
sin
520 3600
7 . 14
(3.9)
Solving
GCV
P from equation (3.9) gives,
20
( )
( )
( )
2
2
1
520
7 . 14
3600
sin
3417
(
(
|
|
.
|
\
|
|
.
|
\
|
=
T P C
Q
arc P
C
P
G
g
Gout
GCV
(3.10)
Substituting equation (3.10) in equation (3.7) gives the total pressure drop across the gas
leg, namely,
+
=
c
G Gout G G
Gout
g
gH Q C
P P
2
( )
( )
( )
2
2
1
520
7 . 14
3600
sin
3417
(
(
|
|
.
|
\
|
|
.
|
\
|
T P C
Q
arc P
C
G
g
Gout
(3.11)
Taking the derivative of equation (3.11) with respect to time, assuming constant
liquid discharge pressure
Lout
P and operating temperature T , gives an expression for the
rate of change of the GLCC pressure caused by the change of gas control valve position,
namely,
( )
( )
( )
( )
( )
( )
( )
( )
( )
|
|
|
|
.
|
\
|
+
|
|
.
|
\
|
|
|
|
|
|
|
|
.
|
\
|
(
(
|
.
|
\
|
+
(
(
|
|
.
|
\
|
|
.
|
\
|
+
|
|
.
|
\
|
+
|
|
.
|
\
|
=
2 2
2
2
1
2
2
1
2
520
7 . 14
3600
520
7 . 14
3600
1
1
3417
520
7 . 14
3600
sin
3417
2
g
g
g
Gout g
Gout
G
g
G
g
Gout
G
g
Gout
Gout
Gout G
c
G G
c
Gout G
C P
dt
dP
C
dt
dC
P Q P C
dt
dQ
T C
T P C
Q
P
C
dt
dP
T P C
Q
arc
C
dt
dH
g
dt
dQ
Q C
g
dt
d
g
gH Q C
dt
dP
(3.12)
21
The rate of change of gas density can be found from the equation of state, given by,
ZRT
PM
G
G
= (3.13)
Taking the derivative of equation (3.13) with respect to time gives the rate of change
of gas density, namely,
( ) dt
dP
ZRT
M
dt
d
G G
=
(3.14)
Liquid Mass Balance. Taking the liquid-phase mass balance in the GLCC gives the rate
of change of liquid level, namely,
dt
dV
d dt
dH
L
2
4
= (3.15)
where the rate of change of liquid volume in the GLCC is given by,
Lout Lin
L
Q Q
dt
dV
= (3.16)
Gas Mass Balance. The gas-phase mass balance in the GLCC gives an expression for the
rate of change of gas mole number, namely,
22
( )
G
G
Gout Gin
G
M
Q Q
dt
dn
= (3.17)
Differentiating the equation of state ( ZRT n PV
G G
= ) with respect to time gives the
relationship between the rate of change of GLCC pressure and the rate of change of gas
mole number and the rate of change of gas volume, namely,
dt
dV
P
dt
dn
ZRT
dt
dP
V
G G
G
= (3.18)
As the volume of the GLCC is constant, the rate of change of gas volume and liquid
volume are related as,
dt
dV
dt
dV
L G
= (3.19)
Substituting equations (3.16), (3.17) and (3.19) in equation (3.18) yields a relationship
between the rate of change of GLCC pressure and the rate of change of gas and liquid
volumes, namely,
( ) ( )
Lout Lin Gout Gin
G
G
G
Q Q P Q Q
M
ZRT
dt
dP
V + =
(3.20)
where the gas volume is defined as, ( )
2
4
d H H V
GLCC G
= .
23
Equation (3.6), (3.12), (3.15) and (3.20) form the GLCC model. The unknowns
are GLCC pressure P , liquid level H , liquid outflow rate
Lout
Q , gas outflow rate
Gout
Q ,
and liquid control valve and gas control valve flow coefficients, namely,
v
C and
g
C ,
respectively. Thus, there are four equations and six unknowns. We need two more
equations for solving the LCV and GCV flow coefficients. These equations were derived
by Wang (2000).
For the metering loop configuration, the GLCC can be operated without control
systems for a limited range of inlet flow variations. Instead of using control valves,
manual choke valves with constant flow coefficients can be used to balance the pressure
drops across the liquid leg and gas leg. For this case, the GLCC model is a static model
and can be solved for equilibrium liquid level and pressure at any liquid and gas inflow
conditions, without any additional equations for flow coefficients.
System Specifications. The system specifications, including all the parameters for the
system components, are given below.
GLCC body: diameter
GLCC
d , total height
GLCC
H ;
Gas leg: diameter
Gi
d , length
Gi
L , friction factor
Gi
f and fittings
Gi
;
Liquid leg: diameter
Li
d , length
Li
L , friction factor
Li
f and fittings
Li
;
Gas control valve: flow characteristics
g
C and response time
oG
C (experimentally
determined);
Liquid control valve: flow characteristics
v
C and response time
oL
C (experimentally determined);
24
Pneumatic line time constants
oL
and
oG
for the liquid control loop and gas
control loop, respectively.
Initial Conditions. It is assumed that initially the system operates at steady-state
conditions. The liquid level and pressure are at the set point liquid level and set point
pressure. The flow conditions are the designed inlet liquid and gas flow rates. The liquid
and gas control valve positions are designed to be 50% open. The pneumatic pressure
signal corresponding to 50% control valve opening is the set point pneumatic pressure (9
psig for this case). For any liquid and/or gas flow rate disturbances from their steady-state
values, the system equations can be solved for the dynamic liquid level and GLCC
pressure.
For liquid level control by LCV and pressure control by GCV, the initial conditions
are given as follows: ( ) ( ) ( ) ( )
Go Gout Lo Lout set set
Q Q Q Q P P H H = = = = 0 ; 0 ; 0 ; 0 ;
( ) ( )
gset g vset v
C C C C = = 0 ; 0 .
For liquid level control by both LCV and GCV, the initial conditions are given as
follows: ( ) ( ) ( ) ( )
Go Gout Lo Lout o set
Q Q Q Q P P H H = = = = 0 ; 0 ; 0 ; 0 ; ( ) ; 0
vset v
C C =
( )
gset g
C C = 0 .
For primary liquid level control by LCV and secondary liquid control valve position
control by GCV, the initial conditions are also given by: ( ) ; 0
set
H H =
( ) ( ) ( )
Go Gout Lo Lout o
Q Q Q Q P P = = = 0 ; 0 ; 0 ; ( ) ( )
gset g vset v
C C C C = = 0 ; 0 ;
set
x x = ) 0 ( .
25
Controller Settings. PID controller is assumed for the control loop. Actually, the kind of
controller to be used for a given system is unknown until the system is analyzed and the
controller is designed.
As expected, the system equations cannot be solved without the controller
specifications. The nonlinear model is difficult to solve for control system design
purposes.
3.1.2 Liquid Level Control by Gas Control valve
Linear Model. The block diagram of the liquid level control loop using GCV is shown in
Figure 3.3. The corresponding linear difference equation model is shown in Figure 3.4.
The linear model is derived based on the assumption that the gas inflow rate remains
constant.
Block 1. This is a pure integrator relating the liquid volumetric flow rate to the liquid
volume. In the form of deviation variable, the liquid inflow and outflow rates can be
expressed as,
Ls Lin Lin
Q Q Q = and
Ls Lout Lout
Q Q Q = . The rate of change of liquid
volume in the GLCC is given by,
Lout Lin Lout Lin
L
L
Q Q Q Q
dt
V d
V = =
=
.
(3.21)
where
L
V is the net liquid volume change in the GLCC.
Taking the Laplace transformation of equation (3.21) gives,
( )
( )
s
s V
s V
L
L
1
.
=
(3.22)
Figure 3.3 Block Diagram of Liquid Level Control System by GCV
Figure 3.4 Linear Model of Liquid Level Control by GCV
CONTROLLER
GAS RATE IN
TRANSMITTER/SENSOR
RELATION 4
+ GAS
CONTROL
VALVE &
ACTUATOR
GAS
RATE
OUT
LIQUID
RATE IN
+
LIQUID
RATE
OUT
RELATION 2
PGLCC
RELATION 1
RELATION 3
+
LIQUID
LEVEL
Q
Lin
+
L
V
&
1
s
L
Q
1
D
H
e
v
p
1
100
'
lim
+
s C
p
o
v
x
set
x x
g
x
C
=
4
D
g
C
Q
Lout
1
1
'
+ s
o
c
E
min max
16
H H
Gout
Q
) (s G
cG
3
D
P
5
D
Q
Lout
16
lim v
p
c
p
s
1
G
V
1 2 3 4 5 6 7 8 9
10 11 12
1
Liquid
Level
Setpoint
2
6
27
Block 2. This block presents the linear relationship between liquid level change and
the change of liquid volume in the GLCC. Using deviation variables in equation (3.15),
the liquid level can be expressed as,
set
H H H = . Substituting the deviation variables
and taking Laplace transform of equation (3.15) gives,
( )
( )
2
1
4
d s V
s H
D
L
=
= (3.23)
Block 3. This is the liquid level transmitter gain. As developed by Wang (2000),
based on the error signal and using deviation variables gives, ) (
16
min max
H
H H
e
= .
Taking the Laplace transform yields,
( )
( )
min max
16
H H s H
s e
(3.24)
Block 4. This is the unknown controller block, which needs to be specified in the
controller design. From Wang (2000), the general form of the mathematical description
for a PID controller is given by,
( )
|
|
.
|
\
|
+ + = s t
s t
K s PID
d
i
c
1
1 (3.25)
or
( ) s k
s
k
k s PID
d
i
p
+ + = (3.26)
28
Block 5. This is the gain, which converts the controller output current signal (4-20
mA) to pneumatic pressure signal (typically 3-15 psi) to actuate the control valve.
( )
( ) ( ) 16 4 20
lim min max v v v
c
c
p p p
s E
s p
=
(3.27)
Block 6. This is the transfer function for the pneumatic line delay.
1
1
'
+
s p
p
o c
v
(3.28)
Block 7. This is the transfer function for the control valve. As shown by Wang
(2000), taking derivative of the pneumatic control valve equation with respect to time
using deviation variables and taking the Laplace transform gives,
( )
( ) 1
100
'
lim
+
s C
p
s p
s x
o
v
v
(3.29)
where
min max lim v v v
p p p = . The negative sign comes from the reverse action of the
control valve.
29
Block 8. This block is the transfer function of the relationship between the control
valve flow characteristic and control valve position. In this study, a linear flow
characteristic around the set point for the control valves is assumed.
( )
( )
set
x x
g g
x
C
s x
s C
=
|
|
.
|
\
|
(3.30)
Block 9. This is the transfer function for the gas flow rate calculation related to the
control valve. In equation (3.9), assuming the pressure is constant and
set
P P = ,
substituting the deviation variables and taking the Laplace transform gives the linear
relationship of the change of gas outflow rate and the change of the gas control valve
position or flow coefficient, namely,
( )
( )
Deg
set
Gout set
set
G g
Gout
P
P P
C
P
T
s C
s Q
D
(
(
=
1
4
3417
sin
520 3600
7 . 14
(3.31)
Block 10. This is a pure integrator relating the gas volumetric flow rate to the gas
volume. In the form of deviation variable, the gas inflow and outflow rates can be
expressed as,
Gs Gin Gin
Q Q Q = and
Gs Gout Gout
Q Q Q = . The rate of change of gas
volume in the GLCC is given by,
Gout Gin Gout Gin
G
G
Q Q Q Q
dt
V d
V = =
=
.
(3.32)
where
G
V is the net liquid volume change in the GLCC.
30
Taking the Laplace transformation of equation (3.32) gives,
( )
( )
s
s V
s V
G
G
1
.
=
(3.33)
Block 11. This is the transfer function for the relationship between the GLCC
pressure and the net gas volume change in the GLCC. In equation (3.20), it is assumed
that the gas column volume
G
V is constant. This assumption is valid provided the liquid
level is controlled around the set point, which implies that the liquid outflow matches the
inflow. Also, it is assumed that the gas density to be constant (when the GLCC pressure
doesnt change much). Using deviation variables and taking the Laplace transform gives,
( )
s
D
V
s P
G
1
3
.
(3.34)
where,
( )
Gset
set
G
V
P
V
s P
D
=
3
.
Block 12. This is another transfer function for the LCV liquid outflow rate
calculation. In this case, the control valve position or flow coefficient is assumed to be
constant at the initial position corresponding to the set point flow conditions. The liquid
outflow is assumed to be driven by the GLCC pressure alone. Using deviation variables
and taking the Laplace transform of equation (3.3) gives,
( )
( )
Lout set L
vset Lout
P P
C
s P
s Q
D
=
1
2
1 002228 . 0
5
(3.35)
31
Controller Design: The open loop transfer function can be obtained from the linear model
of liquid level control by GCV that is given by,
( ) ( ) ( )
) 1 )( 1 (
' ' 2
+ +
=
s s C s
K
s G s G s H
o o
s
cG
(3.36)
where:
) (s H - feed back path transfer function
) (s G - feed forward path transfer function
) (s G
cG
- controller transfer function, need to be determined from the design.
s
K - the system gain, which is given by,
set
x x
g
v
v
s
x
C
p
p
H H
D D D D K
=
|
|
.
|
\
|
|
|
.
|
\
|
|
.
|
\
|
|
|
.
|
\
|
=
lim
lim
min max
5 4 3 1
100
16
16
(3.37)
Further details on the controller design and additional calculations can be found in the
work of Wang (2000).
3.1.3 GLCC Gas Carry-Under
Due to the complex nature of the GLCC dynamic system, a simple method is
required to study the gas carry-under in the liquid leg. Marrelli, et al. (2000) developed a
correlation to predict the gas void fraction in the GLCC underflow based on the in-situ
32
gas volume fraction at the GLCC inlet (GVF
i
), Reynolds Number in the liquid leg (Re
l
)
and a dimensionless equilibrium liquid level (Led). This equation is presented below:
51 . 3
095 . 0
307 . 0
1000
Re
1 . 46
|
.
|
\
|
= Led GVF GVF
l
i
(3.38)
3.1.4 GLCC Liquid Carry-Over
Liquid carry-over in the GLCC is a very complex phenomenon under transient
conditions. However, for limited gas flow rates, simple correlations can be applied to
predict this effect. Ishii and Mishima (1989) introduced a correlation to predict the
entrainment fraction under the quasi-equilibrium condition. Consequently, it should be
applied in regions distant from the inlet. Thus, in terms of an entrainment Weber number
and liquid Reynolds number, the equilibrium entrainment correlation becomes:
( )
25 . 0 25 . 1 7
Re 10 25 . 7 tanh
f
We E =
(3.39)
where,
3
1
2
|
|
.
|
\
|
=
g
g g
D j
We
f
f f
f
D j
= Re
j
g
: volumetric flux of gas-phase (superficial velocity)
j
f
: volumetric flux of liquid-phase
f
: liquid viscosity
: density difference
33
3.2 LLCC System
A schematic of LLCC equipped with control system is shown in Figure 3.5. The
LLCC geometrical parameters and dimensions specified are derived based on specific
design criteria corresponding to the operating conditions. The LLCC separator has a two-
phase flow inlet and single-phase water and oil rich outlets. Watercut meters, such as
Micromotion Mass Flow Meter and/or Starcut Watercut Meter, are used to determine the
water concentration in the underflow.
Starcut
Micromotion
wcC
wcCV
Oil
rich
Oil / Water
Inlet Mixture
Water
leg
Figure 3.5 Schematic of LLCC Control System
Free
Water
34
The actuating signal from the watercut meter is sent to the controller to control the
position of the control valve, which is mounted in the LLCC underflow line. The
operation of LLCC strongly depends upon the split of inlet flow rates. Split Ratio, S.R, is
an important parameter used in this study to quantify the performance of LLCC. It is the
ratio of the underflow rate to the total inlet flow rate, as given below:
S.R =
Qin
Qunderflow
100 % (3.40)
The main objective of using a LLCC compact separator is to provide an effective
alternative for oil-water separation in the form of a free-water knockout device. Hence,
there exists an optimal split ratio that depends upon the LLCC inlet flow conditions.
Optimal Split Ratio is defined as that particular split in which maximum underflow in
LLCC is obtained and at the same time maintaining clear water in the underflow.
However, the inlet water and oil flow rate fluctuations will cause the watercut in the
underflow to fluctuate during operation. These dynamics affect the performance of the
LLCC since watercut at the inlet is an indirect parameter of the optimal split ratio.
Therefore, the objective of the control system is to maintain the optimal split ratio for
different inlet oil and water flow rates.
A schematic of the control strategy developed in this study is shown in Figure 3.6.
The sensor/transmitter (Micromotion Mass Flow Meter and/or Starcut Watercut Meter)
measure the control parameter, in this case the underflow watercut, directly and converts
the watercut signal into current signal in the range of 4-20 mA. This signal is compared
to the watercut set point and the error signal is sent to the controller. The controller
35
output is sent to the control valve in the form of pneumatic actuating pressure signal to
control the valve position accordingly.
Liquid
Underflow
Flowrate
Watercut
Setpoint
Downstream
Controller
Pneumatic
Line
Water
Control Valve
Relation
1 2 3,4,5,6 7,8
Watercut
Sensor/Transmitter
Actual
Watercut
Inlet Oil
Flowrate
Inlet Water
Flowrate
+
-
+
-
Figure 3.6 Schematic of LLCC Control Loop
Design Elements of the Control System
The sequence of procedure followed for LLCC control system design and analysis
is discussed below in detail.
a) Control Objectives:
The central element in any control configuration is the process that needs to be
controlled. Thus, the control objectives for LLCC control are two-fold:
i) obtain clear water in the underflow and
ii) maximize the amount of water that can be separated.
b) Selection of Measurements:
Some means to monitoring the performance of a process is needed in order to
achieve the control objectives. This is done, by measuring the values of certain process
variables that represent the control objectives. In this case, the variable that is used to
monitor the performance of LLCC is watercut in the underflow.
36
c) Selection of Manipulated Variables:
Manipulated variables are those that can be used to control a process. In this case,
position of control valve is the manipulated variable.
d) Selection of Control Configuration:
A control configuration is the information structure that is used to connect the
available measurements to the available manipulated variables. For the LLCC watercut
control, the feedback control configuration is applied.
Feedback Control Configuration: This configuration uses direct measurement of the
controlled variables to adjust the value of the manipulated variables, as shown in Figure
3.7. The objective is to keep the controlled variables at desired levels (set points).
.
.
. .
Measured
outputs
Unmeasured Outputs
Manipulated
Variables
Disturbances
.
.
Process
. .
Controller
Set points
.
.
.
.
. .
Measured
outputs
Unmeasured Outputs
Manipulated
Variables
Disturbances
.
.
Process
. .
Process
. .
Controller
Set points
.
.
Figure 3.7 Structure of Feedback Control Configuration
Non-measured Outputs
37
3.2.1 Linear Model for LLCC Control
The main reason for developing a linear model is to approximate the dynamic
behavior of a nonlinear system in the neighborhood of specified operating conditions.
This approach, in principle, is always feasible and is widely used in the study of process
dynamics and design of control systems. The main advantages are as follows:
Analytical solutions can be obtained for linear systems. It helps to obtain a
complete and general picture of systems behavior.
Most significant developments towards the design of effective control
systems are available for linear systems.
The block diagram of the watercut control loop using a control valve in the water
leg is shown earlier in Figure 3.6. The corresponding linear model is shown in Figure 3.8.
The transfer functions given in the blocks are stepwise mathematical descriptions of the
physical subsystems in Laplace domain. The transfer functions are related to deviation
variables instead of actual variables. A deviation variable is the deviation of a variable
from its steady-state or set point value, denoted by a preceding . The reason for using
deviation variables is that the ratio of the output to the input of a block can be expressed
linearly. This is done assuming that the real controlled variable is at the set point value
from steady state conditions.
Figure 3.8 Linear Model of LLCC Control Loop
1. Feedback Controller 5. C.V Characteristics
2. Pneumatic Line Gain 6. Underflow Calculation
3. Actuator Delay 7,8. Watercut Calculation
4. C.V. Response Time 9. Transmitter Gain
1
D
set
x x
x
v
C
=
1
1
+ s
o
) s (
1 c
G
1 2 3 4 5 6
v
p
c p u Q
c E
x
v
C
6
in
Q
min max . .
16
C W C W
Qo
in
Q
1
C
7 8
R S. 1
C W.
+
1
p / 100 vlim
+
s C
o
e
16
lim v
p
3
8
39
Block 1:
This is the unknown controller block that needs to be determined from the
controller design. As reported by Mathiravedu (2001), the transfer function for a PI
controller is given by,
( )
|
|
.
|
\
|
+ =
s t
1
1 K s PI
i
c
(3.41)
or
( )
s
k
k s PI
i
p
+ = (3.42)
Block 2:
This is the gain, which converts the controller output current signal (4-20 mA) to
pneumatic pressure signal (typically 3-15 psig) to actuate the control valve.
) (
) (
s E
s p
c
c
( ) 4 20
min max
v v
p p
=
16
lim v
p
(3.43)
Block 3:
This is the transfer function of pneumatic line delay.
1
1
) (
) (
+
s s p
s p
o c
v
(3.44)
Block 4:
This is the transfer function for the control valve. Mathiravedu (2001) verified
that taking derivative of the pneumatic control valve equation with respect to time using
deviation variables and taking the Laplace transform gives,
( )
( ) 1 s C
p
100
s p
s x
o
v
v
+
=
lim
(3.45)
40
where,
min max lim v v v
p p p = (12 psig in this study)
Block 5:
As presented before, this block denotes the transfer function of the relationship
between the control valve flow characteristics and control valve position. In this study, a
linear flow characteristic around the set point is assumed.
( )
( )
set
x x
v v
x
C
s x
s C
=
|
.
|
\
|
(3.46)
Block 6:
This is the transfer function of the liquid flow rate calculation for the control
valve. Assuming that the pressure drop across the control valve is constant, the liquid
flow rate is only a function of flow coefficient. Taking derivative of equation (3.3), using
deviation variables and taking the Laplace transform gives,
( )
( )
L
LCV
1
P
0.002228
s
v
C
s
u
Q
D = = (3.47)
Block 7:
This is the transfer function that converts the overflow rate to the split ratio.
Recall that the split ratio is defined as the ratio of the underflow rate to the total inlet flow
rate, namely,
S.R. =
in
u
Q
Q
and dS.R. =
in
u
dQ
dQ
Using deviation variables,
41
( )
( )
( ) s Q
s Q
s R S
in
u
= .
1- ( ) = s R S. 1-
( )
( ) s Q
s Q
in
u
1- ( ) = s R S.
( ) ( )
( ) s Q
s Q s Q
in
u in
1- ( )
( )
( ) s Q
s Q
s R S
in
o
= .
( )
( ) ( ) s Q
1
s Q
s R S 1
in o
=
.
(3.48)
Block 8:
This is the transfer function that converts the split ratio to the underflow watercut.
The conversion value is obtained from the graph plotted using the experimental data.
Using deviation variables, this transfer function is given by:
Slope (C) =
( )
( )
=
s R S 1 d
s C dW
) . (
. .
( )
( ) s R S 1
s C W
.
. .
(3.49)
Block 9:
This is the watercut sensor/transmitter gain. As confirmed by Mathiravedu (2001),
using deviation variables in the controller error signal equation and taking the Laplace
transform gives,
42
( )
( )
min max
. . . .
16
. C W C W s C W
s e
(3.50)
Controller Design
The system open loop transfer function for watercut control using a control valve
in the underflow can be easily derived from the linear model, and is given by,
( ) ( ) ( )
( )( ) 1 1 + +
=
s s C
K
s G s G s H
o o
s
cl
(3.51)
where,
H(s) - feed back path transfer function
G(s) - feed forward path transfer function
G
cl
(s) - controller transfer function (needs to be determined)
o
C - Control Valve Delay (2 sec)
o
- Time Constant of the Actuator (0.2 sec)
K
s
System gain, given by
|
|
.
|
\
|
|
.
|
\
|
|
|
.
|
\
|
|
.
|
\
|
|
|
.
|
\
|
=
= in x x
v
v
v
1
s
Q
C
x
C
p
100
16
p
C W C W
16
D K
set
lim
lim
min max
. .
.
Further information about the controller design and other details in the
development of the LLCC control strategy can be found in the work of Mathiravedu
(2001) and Mathiravedu et al. (2002).
3.2.2 LLCC Model With Gas
In order to be able to predict the hydrodynamic flow behavior in the LLCC
43
operating with small amount of gas, it is important to understand the associated physical
phenomenon. Little amount of gas into the system does not affect the inlet flow patterns
and behavior of the flow in the nozzle. However, when the fluids reach the LLCC body,
and as a result of the vortex forces produced by the swirling phenomenon, gas is attached
to the oil and the water phases. The gas gets attached mainly to the oil, reducing its
density. This causes lower drag between the oil droplets and the continuous water-phase,
causing an improvement in the separation efficiency. The water density is also affected,
because some part of the gas is attached to the water phase.
At the LLCC inlet, the gas phase splits, whereby part of the gas flows upwards
into the upper LLCC part and the other part flows downwards into the lower LLCC part.
Using experimental data, Contreras (2002) developed correlations for the gas void
fraction in the oil phase (
G(o)
) and in the water phase (
G(w)
) in the underflow of the
LLCC, as follows:
Fgl GVF Fgl
Vsl Vsg
Vsg
o G
=
+
=
) (
(3.52)
) 1 ( ) 1 (
) (
Fgl GVF Fgl
Vsl Vsg
Vsg
w G
=
+
= (3.53)
Next the densities of the oil and water phases (with the attached gas) are,
respectively:
) ( ) (
mod
) 1 (
o G g o G o o
+ = (3.54)
) ( ) (
mod
) 1 (
w G g w G w w
+ = (3.55)
The GVF is the inlet gas void fraction and Fgl is a factor determined
44
experimentally, which depends on the inlet watercut, as shown in Table 3.1.
This model is able to describe the phenomenon when the LLCC works with small
amount of gas, up to the maximum efficiency point of the LLCC. Beyond this point, a
behavior reversal occurs and the LLCC efficiency decreases. The LLCC model with gas
does not describe this reversal process. However, the description of this process is not
critical, because the LLCC under these conditions is not efficient.
Table 3.1 Values of Empirical Factor, Fgl
Inlet Watercut
%
Fgl
60-75 0.58
75-85 0.54
>85 0.53
GVF Effect on LLCC Performance:
The effect of the gas void fraction on the LLCC performance was studied by
Contreras (2002), reporting results like those shown in Figure 3.9. This demonstrates that
for a fixed watercut introduced to the LLCC, the presence of gas improves the separation
efficiency of the equipment until reaching a maximum point where the effect of
additional gas drops the efficiency until the stage where the separator performance
deteriorates drastically.
Based on the experimental data a correlation is proposed in order to capture this
phenomenon. This correlation establishes the relationship between the underflow
watercut and the split ratio for different GVF values.
( ) % exp % SR B A WC = (3.56)
where A & B are functions of the GVF%, given by
45
A = -1.739E-01*GVF%
2
+ 1.521E+00*GVF% + 1.481E+02
B = 2.161E-06*GVF%
3
- 4.585E-05*GVF%
2
+ 2.779E-04*GVF% - 8.403E-03
Vsl=0.8 ft/sec Inlet Wcut=60%
50
55
60
65
70
75
80
85
90
95
100
0 20 40 60 80 100
SR(%)
W
c
u
t
(
%
)
No Gas 1.10%
3.00% 5.00%
8.00% 12.00%
15.00%
Figure 3.9 GVF Effect on the LLCC Split Ratio for 60% Inlet Watercut
It is important to notice that this correlation requires further development since
effects such as LLCC inlet watecut and LLCC geometry are not considered. Figure 3.10
presents the prediction of the developed correlation, which follows the physical
phenomenon.
Using the proposed correlation, Block 8 of the LLCC control loop can be replaced
by one transfer function, which depends not only on the split ratio but also on the gas
void fraction.
( ) B SR B A
SR
WC
=
% exp
%
%
(3.57)
46
SR% in LLCC
correlation prediction
0
10
20
30
40
50
60
70
0 5 10 15 20
GVF%
S
R
%
SR%
Figure 3.10 Prediction of LLCC Split Ratio as a Function of the GVF
for 60% Inlet Watercut
The values obtained using equation 3.56 are between 0.4 and 1.0. Although this
correlation was developed for LLCC inlet watercut of 60%, the correction for the
watercut to split ratio gain for other inlet watercuts should not make a significant
difference in the analysis of the system.
GVF Effect on LLCC Control System Stability:
As a consequence of the modification in the LLCC control loop due to the new
split ratio to watercut transfer function (equation 3.56), the root-locus stability map
should be re-evaluated in order to make sure that the control strategy developed is still
stable in the presence of gas. For split ratios from 30 to 100% and for GVF from 0 to
17% the value of this transfer function (gain) oscillates between 0.9672 and 0.4154. The
47
result of the root locus for this range is shown in Figure 3.11 where the arrows indicate
the trajectory covered by the different values.
Figure 3.11 Root Locus Plot for LLCC with Gas
It is clear that for the range studied, the system remains stable in the presence of
gas and under the 40% overshoot region.
3.3 GLCC / LLCC System
Founded on the developments through the years of the GLCC and LLCC
technologies, the next step is to combine both separators as a three-phase separation
system.
48
3.3.1 GLCC / LLCC Separation System
As a first approach, it was proposed to connect the GLCC and the LLCC with the
control strategies proposed by Wang (2000) and Mathiravedu (2001), as shown in Figure
3.12.
Starcut
DP
G
a
s
/
O
i
l
/
W
a
t
e
r
G
a
s
/
O
i
l
/
W
a
t
e
r
Gas Gas
Free Water Free Water
Oil Rich Oil Rich
Starcut
DP
Starcut
DP DP
G
a
s
/
O
i
l
/
W
a
t
e
r
G
a
s
/
O
i
l
/
W
a
t
e
r
Gas Gas
Free Water Free Water
Oil Rich Oil Rich
Figure 3.12 Initial GLCC / LLCC System Approach
However, after some experiments with gas and an oil-water mixture through the
GLCC, it was found that the fluid shearing in the liquid leg of the GLCC compromised
the performance of the LLCC located downstream. The main source of this shearing is
49
the valve located between the GLCC and the LLCC. Therefore, it was proposed to
control the GLCC only by means of the gas control valve, as shown in Figure 3.13. By
doing so, not only was the effect of the shearing due to the liquid control valve
eliminated, but also that the LLCC could take advantage of the pre-coalescing effect
created by the GLCC in the oil-water mixture.
Starcut
DP
G
a
s
/
O
i
l
/
W
a
t
e
r
G
a
s
/
O
i
l
/
W
a
t
e
r
Gas Gas
Free Water Free Water
Oil Rich Oil Rich
Starcut
DP
Starcut
DP DP
G
a
s
/
O
i
l
/
W
a
t
e
r
G
a
s
/
O
i
l
/
W
a
t
e
r
Gas Gas
Free Water Free Water
Oil Rich Oil Rich
Figure 3.13 Current GLCC / LLCC System Configuration
50
3.3.2 Droplet Size Behavior Through Control Valves
As explained before, an accessory such as a valve could produce a dramatic
change in the two-phase mixture flowing from the GLCC to the LLCC, even to the point
of creating an emulsion that is impossible to separate. Thus, previous research done in the
area of emulsion formation in valves and orifice plates is applied to this particular case.
Janssen (2001) studied this problem based on the work from Hinze (1955) deriving an
expression for the maximum droplet diameter:
5
2
5
3
5
3
max
|
|
.
|
\
|
=
c
crit
We d (3.58)
where We
crit
is the Weber number of the maximum stable droplet:
|
|
.
|
\
|
+
=
2
3
24
2
c
d
crit
We
(3.59)
The average amount of energy that is being dissipated per time and mass unit can be
estimated by:
dis
o perm
L
U P
(3.60)
where,
U
o
: average fluid velocity through the restriction
P
perm
: permanent pressure drop
L
dis
: length of the dissipation zone
Using this model and equation 3.3, which describes the flow through a control
valve, one can analyze how the closing-opening action of a liquid control valve will
affect the oil droplets dispersed in water.
51
For a perturbation of Q
gas
= -0.5 ft
3
/s and Q
liquid
= 0.06 ft
3
/s, using the
differential model implemented by Wang (2000) with the purpose of controlling the
liquid level in the GLCC, the results of the valve action on the droplet size distribution
can be observed in Figures 3.14 and 3.15. For the given perturbation, LCV position and
pressure drop are plotted as a function of time in Figure 3.14 and the droplet size is
plotted as a function of time in Figure 3.15.
LCV position and Pressure Drop
through LCV
0.0
10.0
20.0
30.0
40.0
50.0
0.0 5.0 10.0 15.0 20.0 25.0 30.0
t (s)
L
C
V
p
o
s
i
t
i
o
n
(
%
)
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
0.55
D
P
(
p
s
i
d
)
LCV
DP
Figure 3.14 LCV Position and Pressure Drop in Control Valve
While the control valve is closing, the shearing applied to the oil-water mixture
flowing through the valve is increased. As a consequence, the droplet size of the oil
dispersed in water is reduced up to 60% of its original size, as shown in Figure 3.15.
Hence, it is desirable to eliminate the usage of LCV between the GLCC and the LLCC.
)
52
Droplet Size after a Control Valve
300
350
400
450
500
550
600
650
0.0 5.0 10.0 15.0 20.0 25.0 30.0
t (s)
D
r
o
p
l
e
t
S
i
z
e
(
m
i
c
r
o
n
s
)
dpmax
Figure 3.15 Maximum Droplet Size Downstream of a Control Valve
3.3.3 Pressure Losses Between GLCC and LLCC
Calculation of the pressure losses of a two-phase mixture can be a complex task,
depending on the model used. Considering the pressure losses between the GLCC and the
LLCC in the dynamic models, a simplification has been made. A non-slip homogeneous
flow of two incompressible liquids (oil and water) is assumed. This led to simplifications
to use average properties, such as density and viscosity, for the mixture. Therefore, based
on the continuity and momentum equations the pressure losses can be computed as
shown below:
( ) ( ) 0 =
m m m
V A
x
A
t
(3.61)
obtaining, 0 =
x
V
m
53
x
P
A A g S
x
V
A
t
V
A
m m
m
m
m
m
sin
2
finally obtaining,
A
A g S
t
V
A
x
P
m m
m
m
|
.
|
\
|
+ +
sin
(3.62)
where,
A = area of the pipe
S = perimeter of the pipe
V
m
= mixture velocity
= angle of inclination of the pipe
m
= mixture density
m
= mixture viscosity
m
= sheer stress of the mixture against the pipe wall
3.3.4 GLCC / LLCC Control System
Integration of the system based on the models for the GLCC and LLCC, requires
a more realistic approach, in which a common vector is defined in order to transmit all
the information of each system from one separator to the other. This vector would share
the same format for every pipeline containing information, such as flow rates, pressure
and continuous-phase of oil and water mixture. A diagram of this vector is shown in
Figure 3.16. The Emulsion variable has a value of 1 for a water-continuous system and
0 for an oil-continuous system.
54
Qoil
Qwater
Qgas
Qsolids
Emulsion
Pressure
Qoil
Qwater
Qgas
Qsolids
Emulsion
Pressure
Figure 3.16 Common Vector for Each Pipeline
Simulator Development
Since the method of describing the system as real as possible is desired, the
approach of expressing the results of the simulations in terms of absolute values is easier
to understand and more likely to capture the process performance for situations such as
different set points, start-up operations and other transient circumstances. The control
models developed previously by Wang (2000), and Mathiravedu (2001) were expressed
in terms of deviation variables, which require converting the results from the differential
models into absolute values. The absolute (realistic) input vector is also, required to be
expressed in appropriate terms that the differential models would recognize.
The simulator was built using Matlab/Simulink
coriolis mass flow meters. The water and oil flow rates are controlled
using Fisher control valves mounted on the water and oil lines, respectively. Both the
water and oil pipelines have check valves mounted on the lines downstream of the control
valves to avoid back flow. The flow rates and densities of both water and oil are
measured using the Micromotion
application program. It integrates measurements, data acquisition, and interactive data
processing and analysis for the feedback control, and data and results display. It provides
accurate and interactive control and display of measured and analyzed variables. The
control of all functions and data acquisition settings is conveniently provided through the
virtual instrument's "front-panel" interface.
The LabVIEW
)
b) Microwave Watercut Meter (Starcut).
4.2.1 Coriolis Mass Flow Meter (Micromotion
)
A Coriolis device such as Micromotion
was
successfully tested for different scenarios, including moving setpoints and changes in
input flowrates. The main contribution of this simulator is not only to simulate the
integrated control system for the GLCC and LLCC working simultaneously, but also to
incorporate other phenomena occurring in the system. Models such as gas carry-under,
liquid carry-over, pressure drops, dead time, gas effect on LLCC peformance and others
are included to provide a more realistic performance prediction of the real system.
However, the integrated control system is still based on the linearizations made by Wang
(2000) and Mathiravedu (2001), extending these differential models for an absolute value
model. The next stage in this process is to compare the advantages that a truly absolute
model, based on the original differential equations for the system, would provide instead
of using a differential model. Nevertheless, it has been proven that the proposed approach
is capable of providing more insight, as the simulator is sensitive to the setpoints, and
also since simulations for start-up operations and two-stage LLCC are now available.
128
5.2 Experimental Program Discussion
Before testing the GLCC/LLCC system dynamics in the experimental program, it
was revealed how important it was to validate the proper performance of watercut meters
in the presence of gas. This was demonstrated to be critical in order to improve the
quality of the system investigation. As shown in Chapter 4, it is important to consider the
presence of gas, since this factor improved the readings obtained by Micromotion, while
comparing the results with a watercut meter, such as the Starcut, which is not sensitive to
gas. The inclusion of an uncertainty analysis in this comparison is significant in order to
validate the experimental investigation. In addition, it was essential to validate the values
obtained for the other control variable, the GLCC liquid level. As shown in Chapter 4, the
method used to convert the GLCC differential pressure to liquid height is crucial while
comparing it with the actual level.
5.2.1 Uncertainty Analysis for Watercut Meters
The uncertainty analysis for the watercut meters was based on the data collected
by Contreras (2002), who compared the watercut measurements from Micromotion with
the values obtained using Starcut, while simultaneously taking representative samples of
the mixture to verify the real watercut. The discrepancies between the actual watercuts
and the values reported by the watercut meters were assumed to be equivalent to the
systematic errors given by each instrument. This criterion was established in order to take
into account not only the uncertainty claimed by the manufacturer, but also the real
uncertainty of the instrument installed in place.
129
The uncertainty model used was the simplified U
95
, as described by Dieck (1997),
namely,
( )
2
1
2
,
2
95 ,
95 , 95
(
(
+
|
.
|
\
|
=
R X
R
S
t
B
t U
(5.1)
where: B
R
: normally distributed systematic uncertainty at 95% confidence
R X
S
,
: standard deviation of one result. It is the random uncertainty.
t
,95
: students t
95
for a given number of degrees of freedom
An example of the results obtained using the uncertainty model for the watercut
meters is shown in Table 5.1. For an input mixture of 95% watercut, the U
95
for
Micromotion is 2.978 wc% and the U
95
for Starcut is 1.663 wc%.
Table 5.1 Uncertainty Analysis for Watercut Meters
Average WC% t
95
B
R
S
X,R
U
95
Rel. Uncert. (%)
Micromotion 95.80 2 2.91 0.5238 2.978 3.11
Starcut 94.91 2.571 0.69 0.5945 1.663 1.75
The validation data for the watercut meters, including the error given by the
uncertainty analysis procedure, are plotted in Figure 5.1. The error bars in this figure are
within the 5% discrepancy range, mainly for watercuts higher than 65%. A closer look
on the validation data, including the error bars, is seen in Figure 5.2, where the range
from 90 to 100% watercut is plotted. The values obtained with Micromotion differ from
the values obtained with Starcut in the 1% range. However, the uncertainty bars are
larger than this range.
130
Micromotion vs. StarCut
0% gas
50
60
70
80
90
100
50 60 70 80 90 100
wc StarCut (%)
w
c
M
i
c
r
o
m
o
t
i
o
n
(
%
)
+ 5%
- 5%
Figure 5.1 Starcut Watercut Meter Validation using Single-Phase Measurements with
Micromotion (0% Gas, 50%-100% Watercut Range)
Micromotion vs. StarCut
0% gas
90
92
94
96
98
100
90 92 94 96 98 100
wc StarCut (%)
w
c
M
i
c
r
o
m
o
t
i
o
n
(
%
)
+ 1%
- 1%
Figure 5.2 Starcut Watercut Meter Validation using Single-Phase Measurements with
Micromotion (0% Gas, 90%-100% Watercut Range)
131
Including the uncertainty bars in the comparison for the watercut meters
performance in the presence of gas between Micromotion and Starcut, it is important to
notice that the offset in the watercut given by Micromotion is larger than the uncertainty
of the instrument, as shown in Figure 5.3. This is relevant in order to prove the significant
distortion in the readings obtained by Micromotion in the presence of gas.
Micromotion vs. StarCut
86
88
90
92
94
96
98
100
86 88 90 92 94 96 98 100
wc StarCut (%)
w
c
M
i
c
r
o
m
o
t
i
o
n
(
%
)
+ 5%
- 5%
Figure 5.3 Watercut Measurement Performance Comparison in the Presence of Gas
On the other hand, when the comparison between the Micromotion reading
compensated for gas and the values obtained with Starcut, including the uncertainty bars,
as seen on Figure 5.4, are plotted, not only a significant improvement is observed, since
the data is between the 2% discrepancy range, but also the uncertainty bars are
overlapping the 2% discrepancy range.
132
Micromotion +GVF correction vs. StarCut
86
88
90
92
94
96
98
100
86 88 90 92 94 96 98 100
wc StarCut (%)
w
c
M
i
c
r
o
m
o
t
i
o
n
c
o
r
r
e
c
t
e
d
(
%
) + 2%
- 2%
Figure 5.4 Compensated Watercut Measurement Performance in Presence of Gas
5.2.2 Uncertainty Analysis at the Inversion Point
The inversion point graph for the oil-water mixture studied, including uncertainty
bars for the watercut determination using Starcut, can be observed in Figure 5.5. In this
figure, the uncertainty bars around the inversion point overlap. This can lead to several
conclusions, including:
a.) It is not clear if the inversion point actually presents a hysterisis effect.
b.) The hysterisis band could be actually larger than that shown.
133
Continuous Phase based on StarCut
0
0.5
1
44 45 46 47 48 49 50
wc (%)
C
o
n
t
i
n
u
o
u
s
P
h
a
s
e
Water Continuous
Oil Continuous
Figure 5.5 Inversion Point for Oil-Water Mixture based on Starcut
5.2.3 Uncertainty Analysis for GLCC Liquid Level Determination
Additionally, the uncertainty model described by Dieck (1997) has been applied
to the GLCC liquid level determination. As shown in Table 5.2, the systematic error
changes, based on which method is used to convert the measured differential pressure
into equivalent liquid height. The criterion to determine the systematic error (B
R
) is
similar to that applied for the watercut measurement uncertainty. As explained in Chapter
4, there is a difference between the value obtained for the liquid height assuming water
density or mixture density, and the actual value measured in place. It is assumed that this
difference includes not only the uncertainty of the instrument, but also the additional
uncertainties related to its installation in place. Subsequently, even the results for the
uncertainty in the GLCC liquid level show a difference in the systematic errors,
depending on the method used to compute the equivalent liquid height, the values for U
95
134
due to the standard deviation contribution, result in similar values for the ultimate U
95
(2.948 inches while assuming pure water and 2.829 inches using a mixture density).
Table 5.2 Uncertainty Analysis for GLCC Liquid Level Determination
Actual
level (in)
t
95
B
R
S
X,R
U
95
Rel. Uncert.
(%)
in H
2
O 37.6 2 2.10 1.0352 2.948 7.84
mix. Density 35.4 2 0.07 1.4138 2.829 7.99
GLCC level
Changes in setpoint (35.5 in to 45 in)
0.0
10.0
20.0
30.0
40.0
50.0
60.0
0.0 50.0 100.0 150.0 200.0 250.0 300.0
t (s)
L
i
q
.
L
e
v
e
l
(
i
n
)
in H2O
mix. Density
OS=5.1%
OS=6.8%
Figure 5.6 GLCC Liquid Level Comparison With and Without
Mixture Density Correction
The differences in the GLCC liquid level when assuming pure water or when
using a mixture density criterion can be observed in Figure 5.6. The respective
uncertainty bars are displayed as well. The U
95
values obtained for both criterions are
similar to each other and at the same time, these values are equivalent, as compared to
their respective overshoots.
135
5.2.4 Transient Data Discussion
The measurements based on the GLCC liquid level and the LLCC underflow
watercut can be analyzed after validating these two most important variables for the
GLCC/LLCC system, considering that both of them are the variables to be controlled in
the investigation.
The experiments performed using the GLCC/LLCC system include the response
of the GLCC liquid level control and LLCC watercut control system for different
conditions, as described in Chapter 4. The values reported for the GLCC liquid level in
each run present larger changes than the values found for the uncertainty of this
measurement for most of the experiments. This means that the results obtained for the
GLCC liquid level are conclusive, even if the offset between the real liquid level and the
value measured is not considered. This offset is neglected because all the experiments
were performed using high watercut mixtures, and as discussed previously, for these
cases the offset is not as large as it would be for a low watercut mixture.
On the other hand, the values acquired for the LLCC underflow watercut for some
particular cases such as changes in the watercut setpoint from 97% to 98% (Section
4.4.2), the uncertainty of the measurement is actually larger than the change imposed on
the setpoint of the equipment. Based on this, it can be implied that for such small changes
in the watercut, the action of the control system is affected by the uncertainty of the
instruments. Thus, in order to fine-tune the operation of the separator, improvements in
the watercut metering accuracy would be required.
136
6. CONCLUSIONS AND RECOMMENDATIONS
6.1. Conclusions
The study of the GLCC/LLCC system is not a trivial problem. Even though
separate investigations of each of the existing parts of the system have already been
executed, once integrated in a system, the complexity of the real interactions between the
separators and their control systems might lead to operational problems, such as
instabilities and dramatic reduction in separation efficiency. On the other hand, a proper
management of the resources of the GLCC/LLCC system can also lead to improvement
of the separation efficiency of each stage separately, achieving enhancements in the
performance of the system, not possible before combining them.
Based on the detailed theoretical and experimental investigations of this study, the
following conclusions can be drawn:
1. The complexity of the GLCC
/LLCC
/LLCC
/LLCC
operation.
Given that the operation of the GLCC liquid level control produces substantial
changes in the amount of gas carry-under, and as the LLCC efficiency is affected by
the gas fraction in the mixture, any change in the GLCC liquid level may improve or
deteriorate the LLCC operation. If the GLCC liquid level setpoint is lowered below
the inlet level, the LLCC split ratio increases as the gas carry-under contributes to
improve the liquid-liquid separation efficiency. However, there is a limit (7%)
beyond which the amount of gas carry-under is so large that the effect is reversed and
the LLCC performance is then compromised.
6.2. Recommendations
The following recommendations are made for further studies to expand the scope
of this investigation:
1. Improvements in the control models can be performed by solving fundamental
differential equations, which are non-linear. As explained before, the previous
investigations accomplished for the control systems in the GLCC and the LLCC are
based on linearizations of the basic equations for each system. A more detailed
evaluation of the dynamics involved is necessary so as to consider factors such as
initial conditions and the dampening that the system could provide for different
perturbations under particular conditions.
2. GLCC
underflow
watercut control is crucial, in particular if the inlet liquid mixture is from the GLCC
liquid leg. As it was demonstrated, the performance of the LLCC can be improved by
proper selection of the watercut meter.
140
NOMENCLATURE
Symbols
A
= Constant for linear control valve characteristics
C
= Overall flow coefficient
1
C
= Constant for gas control valve:
v
g
C
C
g
C
= Gas control valve flow coefficient
o
C = Control valve response time, t, seconds
v
C
= Liquid control valve flow coefficient
d = GLCC diameter, L, ft
1
D = Constant for GLCC geometry
2
D = Constant for liquid flow rate calculation
3
D = Constant for gas mass balance
4
D = Constant for gas flow rate calculation
5
D = Constant for effect of pressure on liquid flow rate
6
D = Constant for effect of liquid level on pressure
7
D = Constant for effect of pressure on gas flow rate
e = Error signal
E = Controller output
f = Friction factor
Fgl = empirical factor
141
g = Acceleration due to gravity
GVF = Gas Void Fraction (%)
c
g = Factor of proportionality defined by Newton's second law
( ) s G = Feed forward loop transfer function
H = Liquid level, L, ft
( ) s H = Feed back loop transfer function
j = Volumetric Flux
k = Controller gain
K = Gain
L = Length of pipe segments
M = Molecular weight
n = Mole number
p = Pneumatic pressure, m/Lt
2
, psi
P = GLCC pressure, m/Lt
2
, psi
( ) s PD = PD compensator transfer function
( ) s PI = PI compensator transfer function
( ) s PID = PID compensator transfer function
Q = Volumetric flow rate, L
3
/t, /s ft
3
R = Universal gas constant, 10.7317 (lbf/in.
2
)-ft
3
/lbmol-R
s = Laplace variable
S.R = Split Ratio (%)
142
t = Time, t, seconds
T = Temperature, T,
o
R
u = Controller output
U = Controlled variable
V = Volume, L
3
, ft
3
sg
V = Superficial gas velocity, L/t, ft/s
sl
V = Superficial liquid velocity, L/t, ft/s
V
&
= Net volume rate, L
3
/t, ft
3
/s
We = Weber Number
W.C. = Water Cut (%)
x = Control valve position, %
z = Compensator zero
Z = Compressibility factor
Greek Letters
= gas void fraction
= Specific gravity
= Fitting flow coefficient
= Density, m/L
3
, lbm/ft
3
o
= Time constant, t, seconds
= Incremental deviation
= 3.141592
143
= Viscosity
Superscripts
'
= Denotes parameters in the gas control loop
mod
= Modified variable
Subscripts
c = Controller
Deg = Degree
d = Derivative
G = Gas
GCV = Gas control valve
GLCC = Gas-liquid cylindrical cyclone
i = Integral
in = Into GLCC
L = Liquid
LCV = Liquid control valve
lim = Limit
LLCC = Liquid-liquid cylindrical cyclone
max = Maximum
min = Minimum
m = Number of fittings
144
n = Number of pipe segments
o = oil phase
out = Out of GLCC
p = Proportional
s = System
sl = Superficial-liquid
sg = Superficial-gas
s , set = Set point
test = Test
T = Transmitter
v = Valve
w = water phase
145
REFERENCES
1. Afanador E.: Oil-Water Separation in Liquid-Liquid Cylindrical Cyclone
Separators, M.S. Thesis, The University of Tulsa, 1999.
2. Agar, J.: Method and apparatus for net oil measurement, U.S Patent 4774680,
September 27, 1988.
3. Al-Mubarak, A.M.: A new method in calculating Watercut and Oil and Water
volumes using Coriolis Meter, SPE 38786, presented at SPE Annual Technical
Conference, San Antonio, October 5-8, 1997.
4. Arpandi I.A., Joshi A.R., Shoham, O., Shirazi, S. and Kouba, G.E.: Hydrodynamics
of Two-Phase Flow in Gas-Liquid Cylindrical Cyclone Separators, SPE 30683
presented at SPE 70
th
Annual Meeting, Dallas, October 22-26, 1995, SPE Journal,
December 1996, pp. 427-436.
5. Bandyopadhyay, P.R., Pacifico, G.C. and Gad-el-Hak, M.: Gas Core Configurations
in a Cyclone-Type Gas-Liquid Separator, Naval Undersea Warfare Center Division,
Division Newport, Rhode Island, Technical report No. 10308, January 1994.
6. Bednarski, S. and Listewnik, J.: Hydrocyclones for Simultaneous Removal of Oil
and Solid Particles from Ships' Oily Waters, Filtration and Separation, March/April
1988, pp. 92-97.
7. Cobb: Ultrasonic Fluid Composition Meter, U.S. Patent 5473934, December 12,
1995.
146
8. Contreras, W.: The effect of inlet gas void fraction on the performance of liquid-
liquid cylindrical cyclone (LLCC