A Basis for Automated Control of Steam

Trap Subcool in SAGD

Dharmeshkumar R. Gotawala and Ian D. Gates, University of Calgary

Summary Fig. 2 displays interpreted seismic data of the heated zones at

Full steam conformance along the well pair of the steam-assisted the Surmont SAGD pilot (ConocoPhillips 2008) after 2 years of
gravity-drainage (SAGD) oil-sands-recovery process is essential operation. The images demonstrate excellent steam conformance
for high thermal efficiency. Conformance can be improved by around the top well pair, good steam conformance with a cold
controlling injection and production strategies to ensure that interval along the middle well pair, and poor steam conformance
steam is delivered to target regions in the reservoir. Smart wells along the bottom well pair. The interwell-pair spacing is on the
use interval-control valves (ICVs) that, conceptually, can be order of 100 m, which reveals that even over these short distances,
dynamically controlled to yield uniform steam injectivity along conformance can vary significantly during steam injection.
the well pair. Dynamic control algorithms, such as proportional- Although there are many reasons for heterogeneity of the
integral-derivative (PID) control and their associated controller steam chamber, there is a need for recovery processes that are
parameters, have not yet been developed for the SAGD processes adaptive; their evolution can help to promote maximum well-pair
that use ICVs. One control strategy would be to control the inter- use and steam-chamber uniformity along the entire length of the
well subcool temperature difference—that is, the difference well pair. The reasons for chamber heterogeneity include porosity,
between the steam-injection temperature and the produced-fluids permeability, fluid-saturation heterogeneity, oil-composition het-
temperature. If this temperature difference is small, then the liquid erogeneity, shale and mud layers, top and bottom thief zones,
pool above the production well is small and there is a likelihood wellbore flow and hydraulics, undulating trajectories of the injec-
of live steam production from the chamber. On the other hand, if tion and production wellbores, and insufficient steam rate injec-
the difference is large, the pool may rise above the injection well tion (i.e., loss of latent heat as steam flows down the pipe)
and gravity drainage is hindered because the chamber is largely (Gotawala 2011).
filled with liquid. Here, the focus is on developing a simple, ap- There are several means of promoting more-uniform steam
proximate theory for the behavior of the liquid pool at the base of injection into the reservoir:
the steam chamber to determine the ranges of values of control 1. One relatively simple method is enforcement of steam-trap
parameters to achieve a targeted interwell subcool temperature control between the injection and production wells (Edmunds
difference. 1998; Gates and Leskiw 2010). Steam-trap control maintains a
liquid pool above and surrounding the production well, which pre-
Introduction vents live steam production from the steam chamber. Otherwise,
There are two main requirements of a thermal bitumen-recovery injected steam will be directly produced and will not contact bitu-
process. First, the oil must be mobilized; this is often achieved by men, leading to poor thermal efficiency of the process. Also,
using high-pressure and high-temperature steam, which reduces direct steam production can prevent liquid production because the
the viscosity of the oil from its original value in the millions of liquid relative permeability drops at large gas saturations. In typi-
centipoises to less than 20 cp. Second, the oil must be moved to a cal practice, the temperature difference between the injected
production well—this means that a driving mechanism must be steam and produced fluids, known as the interwell subcool tem-
provided. In the SAGD process, displayed in cross section in Fig. 1, perature difference, is controlled to maintain condensed steam—
steam heats oil sands at the edge of the depletion chamber. The oil that is, liquid—at the production well (Edmunds 1998; Ito and
at the chamber edge becomes mobile and flows under the action of Suzuki 1999). The smaller the interwell subcool temperature dif-
gravity, the driving mechanism, to the production well at the base ference, the closer the produced liquids are to the steam tempera-
of the depletion chamber. ture, and thus the smaller the height of the liquid pool. The
In current practice, the length of the upper injection and lower greater the interwell subcool temperature difference, the cooler
production horizontal wells is between 500 and 1000 m and the the produced liquids, and thus the higher the liquid-pool height. If
vertical interwell spacing is between 5 and 10 m (Edmunds and the liquid-pool height is too small, live steam production occurs.
Gittins 1993; Singhal et al. 1998; Komery et al. 1999). If one of If the liquid height is too large, then liquid occupies a large frac-
these two requirements is not met, then the recovery process is tion of the steam chamber (Gates and Leskiw 2010). Given hori-
not technically feasible. These requirements are technical hurdles zontal-well drilling, horizontal wells can range both vertically and
that have to be met. In addition to these requirements are eco- horizontally by several m from the intended horizontal path; thus,
nomic ones: The steam, the largest operating expense in the pro- coupled with reservoir heterogeneity, steam conformance can
cess, must convey nearly all of its heat to productive oil sand. vary along the well pair because of inconsistent steam-trap control
This means that the amount of oil mobilized—the revenue stream along the well pair.
per unit steam injected—must be large. Heat transfer in the form 2. In current practice, the simplest steam-injection horizontal-
of steam from the wellbore to the chamber to the oil sand beyond well design consists of the injection point at the heel of the well;
the edge of the chamber is critical; the greater the efficiency of steam then flows down the well and passes through the slotted
heat transfer from steam to oil, the more productive and economic liner to the oil sand surrounding the well. This implies that,
the recovery process. This implies that a central objective of the depending on the quality profile of steam in the well, the pressure
process is control of energy delivery. However, this is impeded by will either drop from the heel of the well to the toe (because of
reservoir heterogeneity and other factors that cause nonuniform viscous effects if there is more-liquid-like flow) or the opposite
steam chambers, especially in the downwell direction along the will happen (because of inertial forces if there is more-gas-like
injector/producer well pair. flow). The interplay of the pressure profile, heat losses caused by
conduction and steam flow into the reservoir, and changes of the
elevation of the well (because real wells are not perfectly horizon-
Copyright V
C 2012 Society of Petroleum Engineers tal) mean that the steam quality and potential temperature profiles
Original SPE manuscript received for review 11 October 2011. Revised manuscript received
are not uniform along the injection well. This further implies that
for review 20 January 2012. Paper (SPE 159170) peer approved 24 January 2012. the energy delivery along the well to the reservoir is not uniform.

680 September 2012 SPE Journal

 SAGD Well Pairs
Steam
Chamber Gravi Drainage of
Gravity
150 m 150 m
at TC Mobilized Bitumen
Mobil
Steam Condensate
and S
Injection
j Well
Subcool
Liquid T Native
Pool Reservo
Reservoir
at TR
Production Well

Fig. 1—Cross-sectional view of SAGD steam chamber with the

steam-injection and fluid-production well pair. Typically, the
reservoir is between 15 and 45 m thick and the lateral width of
the steam chamber is set by the interwell-pair spacing, which is
typically between 80 and 120 m. This implies that the maximum
width of the chamber is between 80 and 120 m. 2005 2007

Fig. 2—Visualization of Surmont SAGD pilot steam-conform-

One simple method to make the flow more uniform is to place a ance zones interpreted from seismic processing, demonstrat-
second tubing string into the injection well—for example, having ing nonuniform conformance along the three SAGD well pairs
two steam-injection points, one at the heel and the other at the toe (ConocoPhillips 2008). Steam injection started in 1998.
(ConocoPhillips 2008). A dynamic means to control the steam
conformance might be accomplished by dynamically moving one
or all of the injection points along the well through time to “cold”
between the injection and production wells). Here, a simple heat-
spots along the well.
transfer and heat-loss model is developed to determine control pa-
3. The well configuration of the recovery processes can be
rameters to automate control of the SAGD interwell subcool tem-
designed to promote complete steam conformance along the well
perature difference.
as the process evolves as is done in the J-well assisted gravity-
drainage (JAGD) process (Gates et al. 2008). The JAGD well con-
figuration grows the depletion chamber from the toe to the heel, Heat Transfer in the SAGD Process
thus ensuring that there are no “cold” spots along the well as the In SAGD, steam flows from the injection well into the depletion
process evolves. However, reservoir heterogeneity can still con- chamber. Because the steam is at saturation conditions in the
trol the extent of chamber growth along the well pair as the pro- chamber, the temperature is largely constant (except for the small
cess evolves. Limited-entry perforations (LEPs) can help with effect of variable solution-gas content in the vapor phase, which
steam-injection uniformity (Bacon et al. 2000). These in-well leads to steam partial-pressure effects), and thus heat conduction
devices promote sonic flow across chokes, which implies that the within the chamber is small. The dominant heat-transfer mecha-
steam-flow rate depends only on the upstream pressure within the nism from the well to the outer edges of the depletion chamber is
well itself. This in turn means that heterogeneity of the pressure by convection. At the edges of the chamber, heat is conducted
drop across the perforations caused by reservoir heterogeneity is into the oil sands beyond it, leading to oil mobilization. At the top
not the key factor controlling the steam-flow rate into the reser- of the chamber, if it is exposed to the overburden, heat conduction
voir. By having a limited number of LEPs, the pressure within the in the overburden steals heat from the chamber.
well is more uniform than would be the case with a slotted liner.
The key problem with sonic-based LEPs arises when the flow
becomes subsonic; then, the flow rate into the reservoir depends Theory
on the downstream sandface pressure, and thus reservoir heteroge- For the analysis performed here, the physical domain (Fig. 3) con-
neity can still affect steam conformance along the well. sists of the pool at the base of the steam chamber. The production
4. Another, more novel method to promote uniform steam con- well is located at the base of the liquid pool. The steam chamber
formance is to use smart wells. Smart wells are wells with one or is at the saturation temperature TS, corresponding to the steam-
more flow-control devices that control flow along the length of injection pressure P. The original temperature of the oil-sands for-
the well. These methods have been tested in conventional oil mation is TR. The temperature of the liquid pool Tp is taken to be
fields (Brouwer and Jansen 2004; Saputelli et al. 2005; Brouwer constant throughout the volume of the pool, although its value
et al. 2004), but there are very few studies on how they should be changes through time. In the following analysis, to simplify the
used in thermal-recovery processes in bitumen reservoirs (Gota- calculations, the density and heat capacities of oil, water, and rock
wala and Gates 2009). For SAGD, a smart well pair would use have been taken to be constant (not dependent on temperature). In
ICVs to control injection or production of fluids at specified inter- the model, heat flows into and out of the liquid pool at the base of
vals of the well. The control valves permit targeted fluid injection the chamber as follows: First, there is heat transfer from the steam
to or production from specific regions of the reservoir. Given that chamber; second, steam condensate and bitumen flow into the
the steam chamber will rise, the ICVs can be controlled dynami- pool; and third, heat removal occurs with produced water and oil
cally to yield uniform steam distribution along the well pair, thus from the pool through the production well.
dynamically optimizing the productivity of the well pair. A material balance around the pool is given by
Dynamic control algorithms, such as PID control, for these wells
have not been explored before, and thus the parameters for these qo qo þ qw qw  qo qpo  qw qpw
control algorithms have not been established. d . . . . . . . . . . . . . . . . . . . ð1Þ
¼ ðqo Vo þ qw Vw Þ;
The method being focused on here is the use of smart-well- dt
control devices to enforce interwell subcool temperature differ-
ence. Here, the focus is on the dynamics of the temperature in the where qo and qw are the oil and steam-condensate volumetric flow
steam-trapped liquid pool at the base of a SAGD steam chamber. rates, respectively, entering the pool from the chamber; and qpo
Gates and Leskiw (2010) have shown that steam coning directly and qpw are the oil and water volumetric flow rates, respectively,
from the injector to the producer occurs if the interwell tempera- produced to surface. Vo and Vw are the volumes of oil and water
ture difference is below a critical value (their value was equal to in the pool, respectively, and qo and qw are the oil and water den-
approximately 21 C, although it depends on the permeability sities, respectively.

September 2012 SPE Journal 681

 where gS is the quality of the injected steam. After substituting
Steam Chamber
Eq. 6 into Eq. 5, the material balance becomes
" #  
Mobile Cpo ðTS  TR Þ qo
Bitumen and
Condensate
Injection Well qo q o 1 þ  qS  f qw
Flow in geff gS kS SOR . . . . . . . . ð7Þ
Interface Production Well
d
Liquid Pool
¼ ðqo Vo þ qw Vw Þ:
dt
As a first approximation, we assume that the ratio of the volumes
qw, qo of oil and water contained in the pool is proportional to the ratio
qw, qo
of the flow rates of oil and water that feed the pool; that is
TS
Q(t) Vw qw
 : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8Þ
Vo qo
p Heat Losses
Heat Losses T
TR Eq. 8 implies that the produced water largely consists of conden-
sate. In oil-sand reservoirs without topwater or bottomwater,
under typical steam/oil (3–6 m3/m3) and produced-water/injected-
p
qw , qo
p water (0.95 m3/m3) ratios, the produced formation water is equal
to approximately 3–6% of the total water (condensate and forma-
tion water) produced from the reservoir. Thus, the assumption we
Fig. 3—Exaggerated view of SAGD pool at base of SAGD steam
chamber. The steam chamber is at the saturation temperature
make here is reasonable, provided that there are no adjacent water
TS, corresponding to the injection pressure P. The original tem- zones. Together with Eq. 6, Eq. 8 implies
perature of the oil sands formation is TR. The temperature of
the liquid pool, Tp, is taken to be constant. Energy is injected to qw Vw Cpo ðTS  TR Þ
_ ðtÞ: qo and ¼ : . . . . . . . . . . . . . . . . . . . . . . . ð9Þ
the reservoir through the injection well at rate of Q qo V o geff gS kS
qw are the oil and steam condensate volumetric flow rates,
respectively. qop and qwp are the oil and water volumetric flow With Eq. 9, the accumulation term of material balance (Eq. 7) can
rates, respectively. be rewritten as
d
ðq Vo þ qw Vw Þ
Eq. 1 can be rearranged to give dt o   . . . . . . . . . . . . . . ð10Þ
d Cpo ðTS  TR Þ
qo ðqo  qpo Þ þ qw ðqw  qpw Þ ¼ qo Vo 1 þ ;
dt geff gS kS
d . . . . . . . . . . . . . . . . ð2Þ
¼ ðqo Vo þ qw Vw Þ:
dt which results in the following form of the material balance:
 
SAGD field data (ERCB website industrial annual reports) show 1 q
that the water-production rate from the pool qpw is related to the qS  wf
dVo SOR qo
steam-injection rate, expressed as cold-water equivalent, by ¼ qo  : . . . . . . . . . . . . . . . . . . ð11Þ
dt Cpo ðTS  TR Þ
1þ
qpw ¼ fqS ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð3Þ geff gS kS

where f tends to be between 0.9 and 1.1, depending on the nature An energy balance around the SAGD pool is given by
of the reservoir, and qs is the volumetric flow rate of steam qo qo Cpo ðTS  TR Þ
expressed as a cold-water-equivalent volume. If there is no water
þ qw qw Cpw ðTS  TR Þ þ QðtÞ_
zone in the region surrounding the well pair, f  0.95 [see, for
example, Saltuklaroglu (1999)], whereas if there is a water zone  qo qo Cpo ðTp  TR Þ  qw qpw Cpw ðTp  TR Þ
p
adjacent to the reservoir—for example, a top water zone—then f d  
is greater than unity because water is also produced from the ¼ qo Vo Cpo ðTp  TR Þ þ qw Vw Cpw ðTp  TR Þ ;
dt
water zone [see, for example, ConocoPhillips (2008)]. The oil-
production flow rate from the liquid pool qpo is given by the field                    ð12Þ
steam/oil ratio (SOR), as follows: _
where Cp* is the heat capacity of Phase * and QðtÞ is the heat
qS transferred from the steam chamber to the pool through the top of
qpo ¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ
SOR the pool. Heat losses from the pool to the region below the pool
are considered negligible compared with the heat transferred from
After substituting Eqs. 3 and 4 into Eq. 2, the result after rear- the steam chamber. After combining Eqs. 3, 4, and 7 with Eq. 12,
rangement is the energy-balance equation is given by
   
qS Cpo ðTS  TR Þ
qo qo  þ qw ðqw  fqS Þ qo qo Cpo ðTS  TR Þ 1 þ
SOR geff gS kS
 
d qo Cpo
¼ ðqo Vo þ qw Vw Þ:             ð5Þ  qS ðTp  TR Þ _
þ f qw Cpw þ QðtÞ . . . . . ð13Þ
dt SOR
d
The mass-flow rate of steam condensate produced can be esti- ¼ ½ðqo Vo Cpo þ qw Vw Cpw ÞðTp  TR Þ
mated by calculating the amount of steam required to mobilize dt
oil. The steam required can be computed from the energy required After making substitutions and simplifications similar to those
to heat the oil that drains into the pool, qo, from the reservoir tem- performed for the material balance, the energy-balance equation
perature to the steam temperature: becomes
qo qo Cpo ðTS  TR Þ dTp
m_ w ¼ qw qw ¼ ; . . . . . . . . . . . . . . . . ð6Þ þ AðTp  TR Þ ¼ B þ QðtÞ; . . . . . . . . . . . . . . . . ð14Þ
geff gS kS dt

682 September 2012 SPE Journal

 Subcool Measurement
Device –
ΔT s ε(t) y(t)
Controller Reservoir
+
ε(t) = s – y(t)
PC
Fig. 5—SAGD liquid-pool-feedback controller.
Process Control
Valve
ðt
deðtÞ
QðtÞ ¼ Qss  KP eðtÞ  KI eðtÞ dt  KD : . . . . . . . ð18Þ
dt
Fig. 4—SAGD liquid-pool-control diagram. The interwell sub- o
cool temperature difference is the difference between the
injected-steam and the produced-fluid temperatures. The pro- After inserting Eq. 18 into the SAGD pool-control equation, the
cess controller (PC) generates the ICV settings to control the result is
steam-injection pressure and rate into the reservoir.
ðt
deðtÞ A þ KP KI
þ eðtÞ þ eðtÞ dt ¼ 0: . . . . . . . . . ð19Þ
dt 1 þ KD 1 þ KD
o
where
2   0 13 For Eq. 19, the initial condition is given by
1 q
qS  wf
1 6 SOR qo 1 B qS C7 eð0Þ ¼ e0 ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð20Þ
A¼ 6 4qo  þ B
@
C7
A5
Vo C ðT
po S  T R Þ SOR C ðT
pw S  T R Þ
1þ 1þ where e0 is the value of the deviation at time t ¼ 0.
geff gS kS geff gS kS
0 1 Here, the ordinary-differential equation is solved by using the
Laplace-transform method. After applying the Laplace transform
qo f q Cpw B B 1 C
C to Eq. 19, the control equation of the deviation in the Laplace do-
B¼ ðTS  TR Þ  w
Vo Vo qo Cpo @ Cpw ðTS  TR ÞA main is given by
1þ
geff gS kS e0 s
eðsÞ ¼   ; . . . . . . . . . . . . . . ð21Þ
and A þ K P KI
s2 þ sþ
1 þ KD 1 þ KD
_
QðtÞ
QðtÞ ¼ : which can be transformed back to the time domain to give the
Cpw ðTS  TR Þ
1þ deviation e at any time t:
ggS kS  
eðtÞ r1 r2
At steady state, the energy balance simplifies to ¼ er1 t þ er2 t ; . . . . . . . . . . . . ð22Þ
e0 r1  r2 r2  r1
AðTp;ss  TR Þ ¼ B þ Qss : . . . . . . . . . . . . . . . . . . . . . ð15Þ where r1 and r2 are roots of the quadratic equation
 
After subtracting Eq. 15 from Eq. 14, a control equation of the A þ KP KI
SAGD pool results: s2 þ sþ ¼ 0: . . . . . . . . . . . . . . . . . ð23Þ
1 þ KD 1 þ KD
dðTp  Tp;ss Þ Eq. 22 provides an estimate of the normalized deviation of the
þ AðTp  Tp:ss Þ ¼ QðtÞ  Qss : . . . . . . . . .ð16Þ
dt produced-fluids temperature from the setpoint value. In terms of
If the steady-state value is thought of as a target value for the the interwell subcool temperature difference given by DTsc ¼ TS –
operation, a control equation for the deviation e(t) defined as the Tp, Eq. 22 can be rearranged to yield
difference between the pool temperature at time t and the target  
r1 r1 t r2 r2 t
value at steady state can be constructed from Eq. 16: DTSC ¼ 1  e  e ðTS  Tp;ss Þ;
r1  r2 r2  r1
deðtÞ                    ð24Þ
þ AeðtÞ ¼ QðtÞ  Qss : . . . . . . . . . . . . . . . . . . . ð17Þ
dt
where e0 ¼ TS – Tp,ss is the initial deviation (the case in which the
For automated control of the deviation, the control strategy should produced-fluids temperature is equal to the steam temperature,
attempt to force it to zero. Here, a PID controller is used. Fig. 4 implying no liquid-pool height). This initial condition represents
displays the application of feedback control to achieve a target the worst-case starting condition that could be encountered in a
value of the interwell subcool temperature difference. Tempera- SAGD operation. In the following, values of the PID control pa-
ture-measurement devices measure the steam-injection and pro- rameters have been explored to understand what ranges of values
duced-fluid temperatures. The temperature difference between the force the subcool temperature difference to the set point value.
steam and produced fluids can be measured at several locations
along the well pair or at the center of each control interval along
the well pair. These measurements are sent to the process control- Results and Discussion
ler and are based on control function of the system; the process Table 1 lists the values of operating conditions and material prop-
controller alters the controlled variables (for example, the valve erties, including thermal-expansion coefficients and thermal diffu-
settings at each ICV to change the steam-injection pressure at sivities of water, bitumen, and sand. Here, to demonstrate the
each ICV). If there are n ICVs along the steam-injection well, theory, the SAGD operating pressure is taken to be equal to 1.55
then n settings of the valves are altered to control steam conform- MPa, with corresponding saturation temperature equal to 200 C.
ance along the well pair. The initial reservoir temperature has been taken to be 4 C. The
Fig. 5 shows the feedback-control diagram. The deviation e(t) steam quality at the sandface of the SAGD injection well is equal
is the difference between the measured value y(t) obtained from to 0.9. The efficiency of the process (in terms of energy used in
the process and set point value s. The PID controller is defined as oil production) is 0.5; the remainder is lost to overburden (Gates
follows (Stephanopoulos 1984): 2011). The steam-flow rate (cold-water equivalent) is 200 m3/d.

September 2012 SPE Journal 683

 TABLE 1—PARAMETERS USED FOR SAGD LIQUID-POOL PID-CONTROLLER
GAINS ESTIMATION

Parameter Value

Reservoir temperature Tr ( C) 4
Steam temperature Ts ( C) 200
Pressure (kPa) 1553
Latent heat of vaporization of steam kS (J/kg) 1941(103)
Steam-to-oil ratio (steam expressed as cold water equivalent) (m3/m3) 3
Density of oil qo (kg/m3) 1005
Density of water qo (kg/m3) 980
Heat capacity of water cpw (kJ/kg C) 4.187
Heat capacity of oil cpo (kJ/kg C) 1.900
Oil flow rate qo (m3/day/m well length) 0.125
Steam flow rate qS (m3/day/m well length) 0.25
Steam quality (g) 0.9
Efficiency of the process (geff) 0.5
Set subcool temperature ( C) 20

25 25

20 20

15 15
KP = –5
ΔTsc(t), °C
ΔTsc(t), °C

10 10
KP = –2
5 All Values of KI
5 KP = –1

0
0

KP > 0
–5
–5 0 2 4 6 8 10 12 14
0 2 4 6 8 10 12 14
Time, day
Time, day

Fig. 7—The evolution of the subcool temperature difference

Fig. 6—The evolution of the subcool temperature difference
DTsc vs. time at different values of the integral controller gain KI
DTsc vs. time at different values of the proportional controller
(KP 5 KD 5 0).
gain KP (KI 5 KD 5 0).

The SOR has been taken to be equal to 3 m3/m3 for the analysis the set point of the interwell subcool temperature difference DTsc
performed here. The results do not change significantly at other is taken to be 20 C (Gates and Leskiw 2010). If the subcool drops
values of SOR. Given the steam-injection temperature, the set to zero, this means that the produced-fluids temperature is equal
point for the produced-fluids temperature is set to 180 C—that is, to the steam temperature, and live steam is being produced from
the chamber. This produced steam is not contacting native bitu-
men and is thus a heat loss.
25 Fig. 6 displays the results of the case in which proportional
control is used alone. The results reveal that, provided that the
20 proportional gain KP is lower than 1, the subcool temperature
difference evolves to its set-point value. Otherwise, the subcool
remains at the steam temperature and the controller is unable to
15
move the system from the state of having no liquid pool (that is,
ΔTsc(t), °C

KD = –2 live-steam production). The value of the proportional gain sets the

10 time scale of the evolution of the subcool to the set point. The
KD = –5 larger the absolute value of the gain, the more rapid is the evolu-
tion to the set point.
5 KD = –10 Fig. 7 presents the results for the case in which integral control
is solely used. The results reveal that under purely integral con-
0 trol, the algorithm is not capable of moving the operation, so that
the subcool temperature difference moves to its set point regard-
KD > 0 less of the value of the integral gain. These results suggest that the
–5
0 2 4 6 8 10 12 14 value of integral control—that is, the error obtained from the his-
Time, day tory of the deviation through time—is small. This is expected in
this case because no historical data are provided for the simple
Fig. 8—The evolution of the subcool temperature difference model used here.
DTsc vs. time at different values of the derivative controller gain Fig. 8 shows the results of the case where purely derivative
KD (KP 5 KI 5 0). control is used. The results demonstrate that derivative control is

684 September 2012 SPE Journal

 40 m_ w ¼ mass-flow rate of water, kg/s
35
P ¼ pressure, Pa
KP = 10, KI = –200 and KD = –20 Q ¼ flow rate per unit well length, m3/m/s
30 Q(t) ¼ heat rate per unit well length, kJ/m/s
KP = 5, KI = 0.01 and KD = –2
Q_ w ðtÞ ¼ heat input rate,  C day–1
25
r1 and r2 ¼ roots of control function quadratic equation
ΔTsc(t), °C

20 s ¼ set point and Laplace domain

t ¼ time
15 KP = 5, KI = 0.01 and KD = –20 T ¼ temperature,  C
10
V ¼ volume
y(t) ¼ process parameter,  C
5 e ¼ deviation,  C
geff ¼ efficiency of process, dimensionless
0
KP = –1, KI = 0.01 and KD = –20 gs ¼ steam quality, dimensionless
–5 kS ¼ latent heat of steam, kJ/kg
0 2 4 6 8 10 12 14 q ¼ density, kg/m3
Time, day

Fig. 9—The evolution of the subcool temperature difference Subscripts and Superscripts
DTsc vs. time at different values of the PID-controller gains KP, 0 ¼ initial condition
KI, and KD. c ¼ steam-chamber condition
o, w, s ¼ oil phase, water phase, and steam phase
P ¼ SAGD liquid pool
capable of evolving the subcool temperature difference to the set R ¼ reservoir
point for negative values of the gains. Otherwise, the subcool sc ¼ subcool
remains equal to zero, implying that the produced fluids are at the ss ¼ steady state
steam-injection temperature. The value of the derivative gain that S ¼ steam condition
yields the smallest time scale for the subcool to reach the set-point
value is immediately below 1. At derivative gains larger than
1, the subcool remains equal to zero. A more-negative deriva- Acknowledgments
tive gain below 1 yields a larger time scale of response of the The authors acknowledge financial support from Shell Canada
subcool temperature difference. and the National Science and Engineering Research Council of
Fig. 9 presents the results for several cases at various values of Canada.
the proportional, integral, and derivative gains. The cases show
that the control parameters can be tuned to obtain a desired time
scale for the subcool to evolve to the set point. With the combina-
tion of the controllers, the values of the gains are not constrained References
because one or more of the controllers may compensate for the Bacon, R.M., Scott, G.R., Youck, D.G. et al. 2000. Steam distribution and
others. Further research is needed to determine the optimum val- production of hydrocarbons in a horizontal well. US Patent No.
ues of the control parameters within the context of the time scales 6,158,510.
of the SAGD process. Brouwer, D.R. and Jansen, J.-D. 2004. Dynamic Optmization of Water
Flooding With Smart Wells Using Optimal Control Theory. SPE J. 9
Conclusions (4): 391–402. SPE-78278-PA. http://dx.doi.org/10.2118/78278-PA.
Brouwer, D.R., Nævdal, G., Jansen, J.-D. et al. 2004. Improved Reservoir
Conceptually, uniform steam conformance of the SAGD process
Management Through Optimal Control and Continuous Model Updat-
can be achieved with an application of smart-well technology, such
ing. Paper SPE 90149 presented at the SPE Annual Technical Confer-
as multiple ICVs along the well pair to obtain more-uniform steam
ence and Exhibition, Houston, 26–29 September. http://dx.doi.org/
conformance and fluid production. Here, a novel model of the
10.2118/90149-MS.
SAGD liquid pool has been developed by employing heat balance
ConocoPhillips. 2008. ERCB Annual Update: Surmont Project. In-Situ
and mass balance around the SAGD liquid pool that sits above the
Process Report, Alberta Energy Resources Conservation Board, Calgary,
production well. The SAGD liquid-pool model, coupled with a PID
Alberta (4 June 2008), http://www.ercb.ca/portal/server.pt/gateway/
controller, demonstrates that the subcool temperature difference
PTARGS_0_0_303_263_0_43/http%3B/ercbcontent/publishedcontent/
can be controlled to ensure that the liquid level of the pool prevents
publish/ercb_home/industry_zone/industry_activity_and_data/in_situ_
live-steam production from the steam chamber. Live-steam pro-
progress_reports/2008/ (accessed 27 February 2011).
duction can harm the thermal efficiency of the SAGD process.
Edmunds, N. and Gittins, S.D. 1993. Effective Application of Steam
Numerous cases have been evaluated to estimate bounds on the
Assisted Gravity Drainage of Bitumen to Long Horizontal Well Pairs.
values of controller parameters for ICVs along a SAGD well pair
J Can Pet Technol 32 (6): 49–55. PETSOC-93-06-05. http://
to ensure that the controller forces the interwell subcool tempera-
dx.doi.org/10.2118/93-06-05.
ture to a desired set-point value. The results show that the time
Edmunds, N.R. 1998. Investigation of SAGD Steam Trap Control in Two
scale of the evolution of the subcool temperature difference can be
and Three Dimensions. Paper SPE 50413 presented at the SPE Interna-
controlled by altering the proportional and derivative gains.
tional Conference on Horizontal Well Technology, Calgary, 1–4
November. http://dx.doi.org/10.2118/50413-MS.
Nomenclature Gates, I.D. 2011. Basic Reservoir Engineering. Kendall-Hunt, Inc. ISBN:
A ¼ constant, 1/s 978-0-7575-9062-7.
B ¼ constant, kJ/s Gates, I.D., Adams, J.J., and Larter, S.R. 2008. The Impact of Oil Viscosity
Cpo ¼ specific-heat capacity of oil, kJ/kg C Heterogeneity on the Production Characteristics of Tar Sand and Heavy
Cpw ¼ specific-heat capacity of water, kJ/kg C Oil Reservoirs. Part II: Intelligent, Geotailored Recovery Processes in
f ¼ fraction of steam condensate into water, Compositionally Graded Reservoirs. J Can Pet Technol 47 (9): 40–49.
dimensionless JCPT Paper No. 08-09-40. http://dx.doi.org/10.2118/08-09-40.
KD ¼ derivative-control gain, day  C–1 Gates, I.D. and Leskiw, C. 2010. Impact of steam trap control on perform-
KI ¼ integral-control gain,  C–1 day–1 ance of steam-assisted gravity drainage. J. Pet. Sci. Eng. 75 (1–2):
KP ¼ proportional-control gain,  C–1 215–222. http://dx.doi.org/10.1016/j.petrol.2010.11.014.

September 2012 SPE Journal 685

Gotawala, D.R. 2011. Non-Uniform SAGD Steam Chambers: Evolution, Dharmeshkumar R. Gotawala is a Research and Technology
Control, and Optimization. PhD thesis, University of Calgary, Calgary, Advisor in the Scientific Research and Experimental Develop-
Alberta. ment (SR&ED) directorate at the Prairie Region Division of Can-
Gotawala, D.R. and Gates, I.D. 2009. SAGD Subcool Control with Smart ada Revenue Agency (CRA). Before joining CRA, he worked
Injection Wells. Paper SPE 122014 presented at the EUROPEC/EAGE as a post-doctoral research fellow in the Department of
Chemical and Petroleum Engineering at the University of Cal-
Conference and Exhibition, Amsterdam, 8–11 June. http://dx.doi.org/ gary. He also spent 2 years as a lecturer in the Department of
10.2118/122014-MS. Chemical Engineering at the S.V. National Institute of Technol-
Ito, Y. and Suzuki, S. 1999. Numerical Simulation of the SAGD Process ogy, India. His primary areas of SR&ED technical review are oil
In the Hangingstone Oil Sands Reservoir. J Can Pet Technol 38 (9): and gas and chemical, biochemical, environment, and man-
27–35. PETSOC-99-09-02. http://dx.doi.org/10.2118/99-09-02. ufacturing industry sectors. Gotawala holds a BE degree from
Komery, D.P., Luhning, R.W., and O’Rourke, J.G. 1999. Towards Com- Gujarat University, India, in chemical engineering, and an
mercialization of the UTF Project Using Surface Drilled Horizontal MTech degree from the Indian Institute of Technology-Roor-
SAGD Wells. J Can Pet Technol 38 (9): 36–43. PETSOC-99-09-03. kee, India, in industrial pollution abatement. He also holds an
MSc degree from the University of Manchester, UK, in environ-
http://dx.doi.org/10.2118/99-09-03.
mental process design, and a PhD degree in petroleum engi-
Saltuklaroglu, M., Wright, G.N., Conrad, P.R., McIntyre, J.R., and Man-
neering from the University of Calgary.
chester, G.J. 1999. Mobil’s SAGD Experience at Celtic, Saskatchewan.
Paper 99-25 presented at the Petroleum Society of Canada Annual Ian D. Gates is an associate professor in the Department of
Technical Meeting, Calgary, Alberta, 14–18 June. Chemical and Petroleum Engineering at the University of Cal-
gary. He spent 7 years in the industry before joining the univer-
Saputelli, L., Nikolaou, M., and Economides, M.J. 2005. Self-Learning
sity. His primary research interests are in thermal and thermal/
Reservoir Management. SPE Res Eval & Eng 8 (6): 534–547. SPE
solvent methods and optimization of these technologies,
84064-PA. http://dx.doi.org/10.2118/84064-PA. application of smart-well technologies for adaptive produc-
Singhal, A.K., Ito, Y., and Kasraie, M. 1998. Screening and Design Crite- tion of heavy-oil fields, support of vector machine learning for
ria for Steam Assisted Gravity Drainage (SAGD) Projects. Paper SPE reservoir characterization, in-situ gasification (in-reservoir reac-
50410 presented at the SPE International Conference on Horizontal tion engineering), biofilm development in porous media, and
Well Technology, Calgary, 1–4 November. http://dx.doi.org/10.2118/ coal gasification. Gates holds a BSc degree from the University
50410-MS. of Calgary, an MASc degree from the University of British Co-
Stephanopoulos, G. 1984. Chemical Process Control: An Introduction to lumbia, and a PhD degree from the University of Minnesota,
Theory and Practice. Englewood Cliffs, New Jersey: International Se- all in chemical engineering. He is a registered professional
engineer in Alberta. For more information, see http://schulich.
ries on the Physical and Chemical Engineering Sciences, PTR Prentice ucalgary.ca/chemical/IanGates.
Hall.

686 September 2012 SPE Journal