11 Gas lift

Gas lift by injecting gas into the production pipe so that the fluid mixture will be lighter, is
a common lifting method for productive wells. Few moving parts are located in the well,
so the need for intervention becomes small. Gas lift handles water and sand better than
other methods.
Gas injection increase the total volume of the fluid mixture, thereby reducing the average
density. Reduced density helps to reduce the pressure gradient. However, larger volume
also implies greater flow speed that contributes to greater pressure gradient. As long as
the density reduction dominates, injection provides less pressure drop. ith constant
separator pressure, this provides higher production rate.
11.1 Surface facilities
Figure !!.! illustrates a gas lift system. "sually the gas must be compressed before
injection. #he gas is distributed between the wells. hen the field is producing, gas for
injection obtained from the processing plant. At startup, the gas must be obtained from
other sources.

Figure 11.1 Compression and injection for gas lift
11.2 The well
Figure !!.$ illustrates a well. #he gas is pumped into the annulus at the wellhead. %own
hole, the gas flows into the production pipe through a no&&le. #he pressure difference
across the no&&le may be about one atmosphere. 'alves further up the production pipe
is used at startup and will otherwise be closed.
.
Figure 11.2 Gas lifted well
(ressure in production tubing and annulus illustrated in figure !!.$ above are
calculated for well parameters listed in #able !!.!, with gas injection rate)
5 3
0.55 10 / Sm d and oil production rate)
3
430 / Sm d . *t is assumed that the annular
cross section is so large that we can ignore flow friction there.
Table 11.1: Well scenario
11.3 Safety
(roduction pipe) length and diameter L=1600 m, d=70 mm
Annulus ) length and diameter L=1600 m, d=150 mm
Reservoir) pressure, temperature, (*
100
R
p bar
, 60
R
t C , J=20 Sm3/d/bar
*njection valve opening, efficiency fac d
c
=12 mm, C=0.9
Fluid) gravities, viscosities
0.66
g

,
0.82
o
0.5
o
cP 0.0012
o
cP
Gas content in reservoir oil
10
t
R
+lip parameters
1.2
o
C
,
0.2 /
o
v m s
+eparator pressure
20
s
p bar
ith gas in the annulus, we lose the safety barrier that completion fluid otherwise
provides. ith lea,age through the inner casing, gas may then enter between the
casings. -uter casings are usually not designed resist high pressure, so that the gas
may finally reach the surface. #his is obviously very dangerous. *t is usually re.uired
pressure monitoring pressure between casings to detect such developments early.
*f the production pipe or valves lea,, the well fluid may flow in the annulus when the gas
shut off. #his need not be immediately dangerous, but ma,es the start up more difficult.
*f the inner casing lea,s, it can lead to uncontrolled blowout.
11.4 Oil rate with gas lift
e can calculate the bottom well pressure by integrating along the pipe
( )
2
0
1
,
2
L
m
g o s !P " !P m
p # # p g $ v d"
d

_
+ +

,

#he bottom well pressure may also be estimated from the reservoir to the well:
( )
1
o R o
p # p #
J

#he estimate above must necessarily be e.ual: ( ) ( ) ,
g o o
p # # p # , such that
2
0
1 1
2
L
m
R s !P " !P m o
p p g $ v d" #
d J

_
+

,

(11-1)
For a given gas injection rate, the resulting oil production must be such that /!!0$1 is
fulfilled.
11.4.1 Numerical solution
2.uation /!!0!1 can be solved by numerical integration of the pipe flow e.uation and
3R0iteration to determine the oil rate. Figure !!.4 illustrates the relationship) #o%#g& for
the scenario in #able !!.$. Red line indicates instability, according to criteria developed
in 5hapter !!.6
Figur 11.3 il production depending on gas rate, numeris! l"#sing
Figur !!.4 estimates max production) 250 Sm
3
/d reached at gas rate
5 3
0.6 10 / Sm d .
+omewhat larger gas rate will probably be necessary to achieve stability
11.$.2 %nal#tical approac&
ith the following simplifications can /!! to !1 be solved analytically
a1 'olume flow much gas than oil)
g o
' ' >>
, so that) ( )
2
/ / 4
m g
v ' d
b1 3o slip, so that)
g g o o
!P m
g o
' '
' '


+

+
c1 7ineari&ation of the integral in /!!0!1) ( ) ( )
0
L
$ p d" $ p L

'olume flow oil production pipe can then be expressed explicitly as in /!!0$1 below.
*nflow rate is here related to oil flow) /
o o o
# ' ( . 'olume factor, density and flows
related to average conditions along the production pipe
2
2 5
2
2 5
8
1
8
g
R s g " m g
o g
o
m g g o "
o
p p g L $ L'
d
' '
$ L' ' g L
d J(

+ +
(11-2)
Figure !!.8 illustrates /!!0$1 for the scenario in #able !!.$. hen density is estimated
neglecting slip, and slip still occurs, the interfacial friction should be included in the
friction factor, here assumed) fm 9 :.:8. /;y this, the friction factor .uantifies irreversible
energy transfer, while pressure and height terms .uantify reversible energy transfer.
#his is thermodynamically more consistent.1
Figure 11.$: il production for different gas rates, anal#tic appro'imation
At high gas rate is figure !!.8 in reasonably consistent with the numerical solution,
Figure !!.4. #his is reasonable since the analytical solution estimates velocity by gas
flow alone. For smaller rates, there is not true.
11.$.3 Analytical approximation for maximum oil production
<aximum oil production can be estimated by derivative) d=o > d=g 9 :. 2xplicit solution
is possible from /!!0$1, but will involve 8th0degree e.uation. +implification is obtained by
neglecting gas density, thus putting)
0
g

, in e.uation /!!0$1. #his provides optimum
gas flow)

2 5

8
"
g
m
g d
'
$

g
' inserted into /!!0$1,for)
0
g

provides maximum flow. At surface conditions

( )
( )
2 5

8
1 2
R s o
o
o o o m "
o g s
J p p '
#
( $ g
J R L
d

+ +
/!!041
#o arrive at /!!041, we have utili&ed the relationship)
o o
o o o g s
( R + . #his follows from
the blac, oil model.
-ptimal gas injection rate corresponds to optimal gas flow, corrected for injecting gas
dissolved in the oil phase. At surface conditions, optimum injection rate may then be
expressed as
( )
1

o
g g o s sr o
p !
# ' # R R
p ! )
+ /!!081
Rs here denotes gas solubility at pipe average conditions, similar to /!!041, while) Rsr,
denotes gas content in the oil reservoir
For the scenario above gave /!!041 and /!!081 estimates)
340
o
# and
5 3
0.82 10 /
g
# Sm d , somewhat higher than from the numerical solution, Figure !!.4.
11.( )tart up
11.(.1 Closed in well
hen the well is shut in, the bottom hole pressure e.ual to the reservoir pressure. #his
is greater than the bottom hole pressure during production. +tarting a closed in well thus
re.uired more injection pressure than when the well produces.
Figure !!.8 illustrates a closed well with atmospheric tubing head pressure and gas filled
annulus. #he oil level is at !,4:: m /*f a valve lea,s, the oil level will be e.ual in the
annulus and production pipe.1 For gas to flow through the injection valve, the pressure in
the bottom of the annulus is raised to !:: bar. #he purpose of the start0up valve is to be
able to start the well with the gas pressure only slightly above that re.uired for stable
production.
Figure 11.( *efore start up
11.(.2 +al,es
5ommonly, is commonly controlled by the annulus pressure and close when it falls
below a preset level. #his is simple and has the advantage that the wellhead pressure
indicates which valves are open.
Figure !!.6 illustrates a valve. #he pressure in the bellows) pd wor,s in the area) Ab and
provide closing force, possibly supplemented with coil springs. #he annular pressure) pg
wor,s on the same area and provides the force) Fo! 9 pgAb. #o ,eep the valve open this
must exceed the closing force. ;y setting the pressure in the bellows, the valve can be
set to close at the desired annulus pressure.
hen the valve is closed, pressure in the production pipe) pt acts on the area Ap. #he
area the annulus pressure acting on is correspondingly reduced. /Figure !!.6 illustrates
a valve seat without area, as approximation.1 hen the pressure in the production pipe
is smaller than in the annulus, larger gas pressure is re.uired to open the valve than to
,eep it open. #his prevents the valve from opening and closing, ?chattering?, by small
variations in annulus pressure.
Figure 11.- .rinsipal s!etc& for pressure operated gas lift ,al,e
5hec, valve prevents well fluid flows into the annulus when the well is shut off. #he
cho,e limits gas inflow through the valve.
Gas lift valves are often placed in wider sections of the production pipe /side poc,ets1,
Figure !!.@. #he valves can then be replaced by the string operations. 'alves may be
replaced with blind plugs.
Figure 11./ Gas lift ,al,e in side poc!et mandrel
11.3.3 )tart up
Figure !!.A illustrates a case where gas is injected through the middle valve, such the
gas content in the fluid column above increases and the well pressure decreases. hen
the bottom well pressure drops below the annular pressure, gas will begin to flow
through the lower injection valve and the start up completed.
Figure 11.0 )tart up, injection t&roug& t&e middle gas lift ,al,e
11.$.3 Gas lift design
ith gas lift design, we normally understand to decide the location, dimensions and
adjustment of the valves. #he continuous injection valve will always located as deep as
possible. 7ocation of valves along the pipe is based on worst possible start0up
conditions. #his is usually that both annulus and production pipeline is filled with
completion fluid.
hen gas pressure is increased, the li.uid level in the annular space is pressed down. *f
the fluid does not flow bac, to the reservoir, it will be forced up and out of the production
pipe. #his gives pressure in the production pipe) p /%1 9 pth B 7 gx % /?unloading
gradient? in the figure1. From Figure A.!! it follows that at the with gas pressure applied,
the fluid level in the annulus will sin, to $8:: foot. For the gas to flow into the production
pipe, the first start valve be here /or slightly shallower1.

Figure 11.0 1ocating start up ,al,es
Plassering av oppstartsventiler nedover !lger sa""e se#vens. $ette s#%lle si#re oppstart, og
avsetning or ventilprod%sentene
Gas injection through the first start up valve will reduce the pressure gradient up the
production pipe /e.uivalent) blue curve in Figure !!.@1. #his pressure gradient can
estimated for gas and li.uid rates considered. *n Figure !!.A this is called ?design
gradient.? From the figure it follows that with design gradient over valve and static li.uid
gradient below, the li.uid level in the annulus will drop to below 4C:: feet. 3ext boot
valve should then be placed there. hen the gas flows trough this valve, the pressure in
the annulus decreases so that the upper valve closes.
7ocation of startup valves down follows the same se.uence. #his would ensure startup,
and provide business for valves manufacturers
11. !ynamic response
11..1 O"ser#ed flow #ariation
Gas lift as described above assumes steady flow. Figure !!.A shows measured rates for
a well ;rage field. e see that the rates vary widely.
Figure 11.2 3stable well on *rage, 4alsmo 5 al 6277(6
Already Gilbert observed periodic flow variation in wells without production pac,er . He
interpreted the observations as that the gas would accumulate in the annulus when the
gas>li.uid level was above the inlet to the production pipe. #his would press the li.uid
level down, so that the level finally would reach the inlet and gas flow into the production
pipe. (roduction would then increase sharply, while the gas pressure in the annulus
dropped, till the li.uid level again would rise and bloc, the inlet. #he phenomenon is
related with gas lift and GilbertDs explanation consistent with current understanding
commonly occurring instability. #his involves as pressure changes in the annulus,
measured at the wellhead , and is therefore often called ? casing heading .?
11..2 $uantitati#e %odeling
Grupping et al . > !CA8 > considered flow variations as changes between pesudo steady
states and suggested measures based on this understanding. <ost recent wor,s are
based on the transient flow.
Fitremann a al.>!CAE > analy&ed two0phase transient flow in wells and developed a
numerical model to predict the dynamic response and stability . Asheim>!CAA > proposed
analytical criteria that may be considered simplification of Fitremanns model, or further
development of the Hjalmars>!C@4 > stability analyses of air lift pumps. #he criteria
identify parameters that affect stability and indicate how these can be changed for better
stability. Alhanati F al > !CC4 > , Fairu&ov F al > $::8 > , (oblano F al > $::E > has
developed modified criteria.
#he concept underlying the above mentioned wor,s is that flow will be subject to small
discrepancies. #he flow appears stationary with rate) =o , can realistically be expressed
as ) =o /t 1 9 =o B =o/t1 . *f the deviations are sufficiently small, the flow is considered
dynamically stable, with some noise. *n some cases, deviations may be self0reinforcing
/positive feedbac, 1, and over time build up large flow variations , for example, similar to
Figure !!.C above.
%ynamic response as outlined above can be simulated numerically. However, transient
two0phase simulation provides no direct relationship between cause and effect. #he
treatment below is analytical and based on lineari&ation.
11..3 !ynamic response of flow elements
Figure !!.!: illustrates a gas lifted well. e will assume stationary flow and investigate
how the system responds to small changes of pressures and rates
Figur 11.17 Gassl"ft, utgangspun!t for stabilitetsanal#se
*nflow of gas
*f the pressure drop across the no&&le is much smaller than the upstream pressure, the
gas volume will be approximately constant and the pressure drop can be expressed by)
2 2
0.5( )
g g c g
p p * ' . *ncluding variations, pressure and rate may be expressed as)
p%t&=p+p%t&, pg%t&=pg+pg%t& and: 'g='g+'g(t). *nserted in the pressure drop relation
provides)
g g
p p p p +

( )
2
2
0.5( )
g c g g
* ' ' + .;y subtracting the stationary
pressure drop relation and neglecting the second order term) =g
$
, /assuming) =g GG
=g1 we get

2
g
g g g
c
p p ' '
*

(11-2)
#he approach above is called perturbation. -perationally, it corresponds to
differentiation.
*nflow of oil
+tationary rate described by) ( )
o o R
' J( p p . (erturbation as above gives

o o
' J( p /!!041
Flow in the tubing
#o simplify, we neglect friction, consider averaged density and expresses the
relationship pressure relationship as)
t, m "
p p g L + . (erturbation gives
" m
p g L (11-4)
3eglecting slip, the density of the inflowing fluid mixture is expressed as)
( ) / ( )
m g g o o g o
' ' ' ' + +
. (erturbation gives
( )
( )
2
g
o
m o g o g
o
g o
'
'
' '
'
' '

_


+ ,
(11-5)
*f fluid with density) m &m flows into a pipe, while the fluid flowing out has the density
m, the will over time change the average density. <ass balance of the pipe gives
g o
m m
t
' '
t -

+

(11-6)
-t '*tL here denotes pipe volume.
;y differentiating /!!081 and using /!!061 and /!!0E1 we find

( )
g
" o
o g o g
t g o o
'
g L '
p ' '
t - ' ' '

_


+
,
(11-()
/!!0$1 and /!!041 entered into /!!0@1 .uantifies the dynamic response to pressure
changes
2
2
2
1
g g
" o c o
g
t g o g g c o
'
g L ' * J(
p p p
t - ' ' ' * '

_ _

+
, ,
(11-8)
#he stationary rates are considered constants. 2.uation /!!0A1 expresses thus the
dynamic response of the tubing by a differential e.uation in the format)
g
p a p b p

A nnulus
(ressure must follow the general e.uation of state) p' 9/ m><1&R#, so that the pressure
depends on the amount of gas in the annulus.
g
r
p
)R! m
t .- t

(11-))
Here we assume constant gas flow into the annulus. #he gas mass flow out, through the
injection valve, is expressed) g'g%t&= g'g+g'g , and gas density relates to pressure)
= pM / zRT. ith /!!0C1 this gives
g
g g
r
p
p '
t -

(11-10)
(ressure response, similar to /!!0!A1, we find by putting /!!041 into /!!0!:1
( )
2
g
c
g g
r g g
p
*
p p p
t - '

(11-11)
2.uation /!!0!!1 expresses thus the dynamic response of the annulus by a differential
e.uation in the format:
g g
p c p c p

11..4 System response
%ynamic response of the annulus is given by e.uation /!!0!!1, for the production of pipe
/!!0A1. #ogether, these provide the system response

g g
p p a b
p p c c t


1 1 1

1 1 1

] ] ]
(11-12)
#he coefficients) a,b,c follow from /!!0A1 og /!!0!!1 as shown above.

2
2
2
1
g g
" o c o
t g o g g c o
'
g L ' * J(
a
- ' ' ' * '

_


+
,
2
" o c
t g o g g
g L ' *
b
- ' ' '

+
2
g
c
r g g
p
*
c
- '

#he coefficients) a, b, c, contain constants and stationary variables and can thus be
found by solving for stationary flow, or measuring at stable gas lift.

Figure !!.!! illustrates the dynamic response calculated from /!!0!$1 for the well
scenario given in #able !!.!, gas injection @: ::: +m
4
> d. #his gave the oil rate) $86
+m
4
>d and response coefficients) a=0.00448, b=0.00629, c=0.00615. #he response is
estimated for an initial offset) p=1Pa, *+ multiply out /!!0!$1 for small time steps) t
Figure 11.11: 4ampened d#namic response 8stable9
Figure !!.!$ illustrates dynamic response when the gas rate is reduced to E: ::: +m
4
>
d. #his gave response coefficients) a=0.00971, b=0.0114, c=0.00839. *nitial offset as
above now resulted in fluctuations with period of) # 9$C minutes and amplitude that
increase with time

Figure 11.12: 3n:dampened d#namic response 8unstable9
3umerical experiments as above can of course also be made with transient,
numerical simulation models. #his re.uires more wor,, but provides greater
flexibility. Here, we go further and relate stability to well design and operation.
11./ 4#namic stabilit#
11./.1 ;at&ematical basis
%ynamic changes can be described by differential e.uations. A simple case is change
over time) / / / t beeing proportional to the current level
/ / (11-13)

For constant proportionality factor, , integration of /!!0!41 gives
( ) ( )
( )
o
t t
o
/ t / t 0

(11-14)
*f) G:, then) y /t1 obviously approaches &ero when the time increases. +uch a system is
obviously stable. e shall later show that under certain conditions, a gas lift system may
be described by /!!0!41.
#he response relationship derived above /!!0!$1 corresponds to /!!0!41, in vector form.
;y letting y denote the vector) Hy!, y$I and A a $x$ matrix li,e /!!0!41 we can write
/ */ (11-15)
A matrix) * with distinct eigenvalues)1, 2 can be diagonali&ed)
1
* 1 1

, where)
denotes the diagonal matrix with the eigenvalues of) * along its diagonal. And 1 is the
matrix with eigenvectors as columns. ith this we can express /!!0!E1 as) 1/ 1/ .
;y the coordinate transformation) ) =1/, /!!0!E1 is expressed as) ) ) , written out
as
,
1 1 1 1 1
,
2 2 2 2 2
0
0
) ) )
) ) )


1 1

1 1
] ]

#his provides two independent e.uations each of which can be solved accordingly /!!0
!41)


( ) ( )
( )
( ) ( )
( )
1
2
1 1
2 2
o
o
t t
o
t t
o
) t ) t 0
) t ) t 0

(11-16)
As for /!!0!41 the solutions of /!!0!61will be dynamically stable if the exponents are
negative. #he exponent
1
,
2
are the eigenvalues of the matrix) A. #hus, we can
diagonali&e the vector e.uation /!!0!E1, or /!!0!$1, to /!!0!61 and transform the solution
bac, to the original coordinates, by) /=1
21
).
For a gas lift system will eigenvalues often be complex. For a $x$ matrix can be complex
eigenvalues expressed as)
R
3 t . *nserted into /!!0!61 gives the solution)
( ) ( )
( ) ( )
( )
( )
( ) ( ) ( )
-os sin
R o o R o
t t 3 t t t t
o o o o
) t ) t 0 0 ) t 0 t t 3 t t



+ .
#his predicts dampened oscillations if the real component is negative) ./0, un0
dampened for positive real component. #he fre.uency is given to the imaginary
component. e see that the analysis predicts fluctuations as observed from the
numerical solution in Figure !!.!! and !!.!$.
;y transformations as outlined above, analytical solution is possible, Hirsch F +male >
!C@8 >. However, for our purposes it is usually fully satisfactorily to solve numerically. *n
many cases we will also not bother about fre.uencies and damping coefficients, but
want criteria that indicate stable or not.
11./.2 )tabilit# criteria
From the discussion above it follows that the dynamic response and stability depends on
the eigenvalues of response matrix. #hese can be found by solving the characteristic
e.uation) HA0I 9 :. For $x$ matrix of the response e.uation /!!0!$1 the eigenvalues
become
( ) ( )
( )
2
0.5 4 a c a c b a c t (11-1()
*n most cases the expression under the root0sign becomes negative, thus) complex
eigenvalues and periodic fluctuations. #he amplitudes will be dampened if the real part
of the eigenvalues are negative) R <0. From /!!0!@1 thus follows stability if) a-c <0. ;y
inserting the coefficients, a and c, we obtain a dimension0free stability criterion,
Asheim>!CAA >
( )
2
1
1
1
1
g g o
t
r o o g "
p ' '
-
4
- ' 4 g L
+
>

(11-18)
01ere:
2
1 2
g g
o
c o
'
J(
4
* '

(11-1))
11./.3 <ealities
#he criteria are based on substantial simplifications and should therefore be considered
indicative. 2xperience shows if the simple stability criteria /Asheim>!CAA > , Alhanati>!CC4
> and others1 predicts instability, the well will be unstable. ;ut wells predicted stable will
not always be so, Alhanati > $:!4 >. #he interrelations between design parameters,
operating conditions and stability that follows from the criteria appear to be reasonably
consistent with experience .
Figure !!.C above illustrated serious instability. %alsmo al. > $::E > state that the well
had high productivity, with long and complex completed interval. #he well was tried
started up against a cho,e, with critical flow of gas injection valve. #his should provide
constant gas inflow. Although the case differs is not consistent with the assumptions in
the stability analysis above, constant gas flow should contribute to stability. However, the
well turned out so unstable that normal startup was impossible. #he problem was
eventually solved by automatically controlling the throttle valve.
Figure !!.!4 shows a more typical case of unstable gas lift. <easured pressure varies
here between $A and 4$ bar, with a period of $4 minutes. #he fitted sine curve shows
good agreement with the measured pressure. -ther measured variables vary
accordingly, usually with a phase shift.
&igure 11.13 O"ser#ed' sta"ili(ed pressure fluctuations
-ur model is lineari&ed and distinguishes therefore only between stability and instability.
Gas lift as illustrated in Figure !!.!4 is therefore unstable according to our definition. *n
reality, $nd order terms will limit the amplitudes, so that a mildly unstable well may be
producible, albeit hardly optimal.
11.0 <eferences
1)(3 23al"ars, 4.
561e 7rigin o 8nsta*ilit+ in 9irlit P%"ps:
;. o 9pplied <e-1ani-s, ;%ne 1)(3, 3))
1)(4 2irs-1, <.=., 4"ale, 4.
$ierntial >?%ations, $+na"i-al 4+ste"s, and @inear 9lge*ra
9-ade"i- press, Aoston 1)(4
1)82 $e<oss, >.>, 6ie"ann, =.$.:
BCas @it 8n-reases 2ig1-Dol%"e Prod%-tion Ero" Fla+"ore EieldG
;. Pet. 6e-1., 9pril 1)82, 6)6
1)84-a Cr%pping, 9.=., @%-a, F.=.E., Der"e%len, E.$.
52eading a-tion anal+Hed or sta*iliHation:
7il and Cas ;., ;%l+ 23, 1)84, 4(
1)84-* Cr%pping, 9.=., @%-a, F.=.E., Der"e%len, E.$.
561ese "et1ods -an eli"inate or -ontrol ann%l%s 1eading:
7il and Cas ;., ;%l+ 30, 1)84, 186
1)85 Eitre"ann, ;.-<. I Dedrines P.
:Jon stead+ gas-li?%id lo0 and gas-lited 0ells:
2
nd
8nternational Foneren-e on <%ltip1ase Elo0, @ondon 1)-21 ;%ne 1)85
1)88 9s1ei", 2.: 5Friteria or Cas @it 4ta*ilit+:
;. Petr. 6e-1. , Jove"*er 1)88, 1452
1))3 9l1anati, E.;.4., 4-1"idt, K., $ot+, $..., @agerle, $.$.:
Fontin%o%s Cas-@it 8nsta*ilit+: $iagnosis, Friteria and 4ol%tions.
4P> 26554, Pro-. 9nn%al 6e-1. Foneren-e, 2o%ston, 6eLas, 3-6 7-t. 1))3, 401
1))5 de 7liveira, 9.
>sta*lidade 7pera-ional de Po-os -o" Cas-@it Fontin%o
$issertation, Mniversidade >stad%al de Fa"pinas, $e-.1))5
1))6 Carna%d, E., Fasagrande, <., Eo%illo%t, F., @e"eta+er, P.:
Je0 Eield <et1ods or a <aLi"%" @it Cas >i-ien-+ 61ro%g1 4ta*ilit+
4P> 35555, Pro-. >%ropean Prod%-tion 7perations Fon., 4tavanger, 1))6
1))) 9s1ei", 2.:NDerii-ation o 6ransient, <%lti-P1ase Elo0 4i"%lation or Cas @it
9ppli-ationsN, 4P> 5665), 9nn%al 6e-1ni-al Foneren-e, 2o%ston, 6eLas 3-6 7-t. 1))).
2002 $als"o, <., 2alvorsen, >., 4l%pp1a%g, 7.:
9-tive Eeed*a-# Fontrol o Mnsta*le =ells at t1e Arage Eield
4P> ((650, Pro-. 9nn%al 6e-1. Fon., 4an 9ntonio, 6eLas, 2) 4ept. -2 7-t., 2002
2003 2%, A., Colan, <.:
Cas-lit 8nsta*ilit+ .es%lted Prod%-tion @oss and 8ts .e"ed+ *+ Eeed*a-# Fontrol:
$+na"i- 4i"%lation .es%lts
4P> 84)1(, Pro-. 8nt. 8"proved 7il .e-over+ Foneren-e, O.@.,7-t. 2003
2004 C%errero-4ara*ia, 8., ., Eair%Hov, P.D.:
4ta*ilit+ 9nal+sis o Cas @it =ells Msed or $eep0ater 7il Prod%-tion
4P> 10403(, Pro-. Eirst 8nt. 7il Fon. , Fan-%n, <eLi-o, 2004
2004 Eair%Hov, P.D., C%errero-4ara*ia, 8., Falva-<orales, F., Far"ona-$iaH, ..,Fervantes-
AaHa, 6., <ig%el-2ernandeH, J., .o3as-Eig%eroa, 9.: 4ta*ilit+ <ap or Fontin%o%s
Cas @it =ells: 9 Je0 9pproa-1 to 4olving an 7ld Pro*le"
4P> )0644, Pro-. 9nn%al 6e-1. Foneren-e, 2o%ston, 6eLas, 26-2) 4ept. 2004
2005 2%, A. : F1ara-teriHation o gas-lit insta*ilities
$o#torav1andling J6JM, (-, 2005
2005 Po*lano, >., Fa"a-1o, .., Eair%Hov, P.D.:
4ta*ilit+ 9nal+sis o Fontino%s-Elo0 Cas @it =ells
4P> Prod%-tion and Ea-ilities, Ee*. 2005, (0