Vous êtes sur la page 1sur 24

Energy Conversion and Management 47 (2006) 16211643

www.elsevier.com/locate/enconman

Numerical modeling of ow processes inside geothermal wells:


An approach for predicting production characteristics
with uncertainties
O. Garca-Valladares a, P. Sanchez-Upton

b,c

, E. Santoyo

a,*

Centro de Investigacion en Energa (UNAM), Privada Xochicalco s/n, Temixco, Mor. 62580, Mexico
Posgrado en Ingeniera (Energa), UNAM, Privada Xochicalco s/n, Temixco, Mor. 62580, Mexico
Gerencia de Proyectos Geotermoelectricos (CFE), Alejandro Volta 655, Apdo. Postal 7-31, Morelia, Mich., 58290, Mexico
b

Received 27 March 2004; received in revised form 10 February 2005; accepted 13 August 2005
Available online 22 September 2005

Abstract
One dimensional steady and transient numerical modeling for describing the heat and uid dynamic transport inside
geothermal wells has been conducted. The mass, momentum and energy governing equations were solved using a segregated numerical scheme. Discretized governing equations for the uid ow were coupled and solved with a fully implicit
step by step method. The mathematical formulation used suitable empirical correlations for estimating the convective heat
transfer coecients as well as the shear stress and the void fraction parameters. Heat conduction across the wellbore materials was solved by an implicit central dierence numerical scheme using the tri-diagonal matrix algorithm (TDMA). The
ow characteristics of producer geothermal wells (pressure, temperature, enthalpy, heat uxes, etc.) at each depth node
were computed. Analytical data reported in the literature were used to validate the numerical capability of the wellbore
simulator developed for this study (GEOWELLS). This simulator, together with another computer code (ORKISZEWSKI), was applied for modeling the heat and uid ow processes inside some wells drilled in Mexican geothermal elds.
The simulated pressure and temperature proles were statistically compared against stable measured eld data (through
the computation of the residual sum of squares and Chi-square). A good agreement between the simulated and measured
proles of pressure and temperature was consistently obtained, having the best matching results for the GEOWELLS predictions. An analysis of the sensitivity and uncertainty was nally conducted to estimate the condence to be accorded the
simulation results predicted by GEOWELLS. Matching the sensitivity to variations in some input parameters (e.g., pressure, temperature, enthalpy and void fraction) was examined. The void fraction was identied as one of the most important
parameters that aect the GEOWELLS simulations for matching measured eld data correctly.
 2005 Elsevier Ltd. All rights reserved.
Keywords: Sensitivity and uncertainty analysis; Heat losses; Geothermal wellbore; Two phase ow; Error propagation

Corresponding author. Tel.: +52 (55) 56229774; fax: +52 (55) 56229791.
E-mail address: esg@cie.unam.mx (E. Santoyo).

0196-8904/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2005.08.006

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1622

Nomenclature
A
Cp
CV
D
e
f
fG
g
G
h
m
_
m
n
Nu
p
P
Pr
q_
Re
t
T
v
xg
Xtt
z

cross section area [m2]


specic heat at constant pressure [J kg1 K1]
control volume
diameter [m]
specic energy (h + m2/2 + gz sin h) [J kg1]
friction factor: s = U(f/4)(G2/2q)
Gnielinski friction factor
acceleration due to gravity [m s2]
mass velocity [kg m2 s1]
enthalpy [J kg1]
mass [kg]
mass ow rate [kg s1]
control volume number
Nusselt number: Nu = aD/k
pressure [Pa]
perimeter [m]
Prandtl number: Pr = lcp/q
heat ux per unit area [W m2]
Reynolds number: Re = qvD/l
time [s]
temperature [K]
velocity [m s1]
mass fraction (steam quality)
 0:5  0:1 

q
1xg 0:9
ll
turbulentturbulent Lockhart Martinelli parameter X tt qg
lg
xg
l
axial coordinate

Greek letters
a
heat transfer coecient: a q_ wall =T wall  T f [W m2 K1]
Dr
radial discretization step [m]
Dt
temporal discretization step [s]
Dz
axial discretization step [m]
e
roughness [m]
eg
void fraction
u
dependent variables (mass ow rate, pressure and enthalpy)
U
two phase frictional multiplier
c
surface tension [N m1]
k
thermal conductivity [W m1 K1]
l
dynamic viscosity [Pa s1]
h
angle [rad]
q
density [kg m3]
s
shear stress [Pa]
wshah Shah correction factor
Subscripts
f
uid
g
gas or steam

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

l
meas
r
sim
tp
w, e, n,
z

1623

liquid
measured
radial direction
simulated
two phase
s west, east, north, south (neighbor)
axial direction

Superscripts
0
previous instant


 ui ui1 =2

arithmetical average over a CV: u


 1  R zDz
u dz

integral average over a CV: u Dz
z

1. Introduction
The evaluation of the production performance of geothermal wells is a fundamental engineering task required for exploitation and optimization of their resources in power generation [13]. The optimum design
of the wellbore geometry, the identication of permeable zones inside the wellbore, the design of surface equipment and the wellbore deliverability studies are commonly evaluated from wellbore production data [4]. The
production characteristics are usually inferred from stable ow measurements carried out in situ. Studies of
the uid ow and heat transfer processes inside wells play an important role for a better understanding of
the wellbore production [5,6]. Numerical modeling has been widely recognized as an eective and cheaper tool
used to study the heat and uid ow processes inside geothermal wells [57]. This tool has usually been designed for replacing experiments in geothermal wells, in which the experiments could be dicult, costly or
unfeasible to perform.
A wellbore physical model is usually considered as a vertical circular pipe with liquid ow at the
deeper zone, which ashes in the upper zone due to the reduction of the static pressure and temperature
[8,9]. A two phase zone (liquid and steam) is formed above the ashing point with an increase of
mixture velocity and steam quality as the two phase ow gradually rises to the wellhead. The analysis of
two phase ow inside geothermal wells requires the correct calculation of pressure, temperature, enthalpy,
heat uxes and velocity proles along the well depth, as well as the knowledge of the uid thermophysical properties. These parameters are feasible to be computed by solution of the mass, momentum
and energy governing equations using numerical schemes that represent the complexity and multidimensionality of two phase ow [9]. Steady state or transient one dimensional ow models have been
suggested as suitable tools for analyzing the wellbore ow [1,9]. Although numerous simulators for
modeling the uid and heat ow in geothermal wells have been already developed [1,8,10,11]; their
prediction capability for matching measured eld data has been only partially successful. In some
matching analyses (related to pressure or temperature proles), the x-axis scale of the output variable (e.g., pressure or temperature) is manipulated for trying to demonstrate a suitable matching
between the simulated and measured data (especially when a by eye graphical method is being used)
[7].
Inaccurate predictions of geothermal wellbore simulators are generally associated with: (i) measurement
errors of temperature and pressure logs, enthalpy and mass ow rate [7]; (ii) the use of physical models that
do not describe accurately the uid ow and heat exchange inside a well [12]; (iii) the use of non-representative empirical correlations for computing two phase ow parameters, such as friction factors or void fractions [13] and (iv) the use of thermophysical properties of water that do not represent the actual
thermodynamic behavior of geothermal uids [14,15]. Moreover, when modeling two phase ows inside
geothermal wells, major prediction diculties also arise due to input data uncertainties [16]. Statistical

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1624

approaches, therefore, should be focused on evaluating the propagation of these uncertainties in numerical
simulations [17]. To our best knowledge, statistical studies for determining the eects of the input data
uncertainties on the simulations have been rarely reported in the geothermal literature. In other areas, these
studies are recognized as indispensable tasks for a better evaluation of the predictive capability of simulators
[18,19].
The goal of this work was to conduct a numerical modeling to study the uid and heat ow inside geothermal wells by means of the following tasks: (1) to develop a new wellbore simulator (called GEOWELLS)
for describing the physical processes that occur inside a well; (2) to validate the numerical capability of
GEOWELLS using analytical data reported in the literature; (3) to apply GEOWELLS for the prediction
of temperature and pressure proles in some producer wells of Mexican geothermal elds; (4) to compare statistically all the results predicted by GEOWELLS against both eld measurements and simulation results predicted by an existing wellbore simulator ORKISZEWSKI [20]; and (5) to evaluate the eect of input data
uncertainties on the GEOWELLS simulations.

2. Physical model of the GEOthermal WELLbore Simulator (GEOWELLS)


2.1. Features and assumptions
The developed simulator has the following numerical capability: (a) to analyze liquid, two phase and superheated steam ows; (b) to perform wellbore simulations from the bottom hole to the wellhead section using an
adaptable depth mesh (upto 0.1 m) without stability problems. A reverse situation could be easily modeled
using an iterative method with suitable boundary conditions; (c) to specify multiple changes of the wellbore
geometry (casing, diameters and roughness), including deviated wells; (d) to quantify transient heat losses between the well and the surrounding formation and (e) to obtain simulated data (pressure, temperature, enthalpy, steam quality, mass ow rate, ow regime, as well as friction, acceleration and gravitation pressure
gradients) at each wellbore depth node.
The following general assumptions are also considered for development of the wellbore simulator:
(i)
(ii)
(iii)
(iv)
(v)

One dimensional and transient ow is dened (i.e., p(z, t), h(z, t), T(z, t), etc.),
Two phase ow is analyzed by a separated model,
The thermodynamic behavior of geothermal uids is approximated using water properties,
The heat exchange by radiation and the axial heat conduction inside the uid ow are neglected and
The heat conduction equation through the wellbore materials (pipe, cement and rock formation) is
solved using a three-dimensional transient axial-symmetric temperature distribution.

2.2. General equations


The mathematical formulation used for describing the uid ow inside a control volume (CV) of a geothermal well is based on the scheme shown in Fig. 1. Semi-integrated governing equations for analyzing the uid
ow over a nite CV are given by
Continuity:
_ i1
m
i


om

0
ot

Momentum:
i1
m_ g vg i

i1
m_ l vl i

 ~
om_
i1
Dz
pi A  ~sP Dz  mg sin h
ot

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1625

Fig. 1. Schematic diagram of a geothermal wellbore showing the characteristic control volumes (CVs). For a uid CV, inlet and outlet
sections are represented as i and i + 1. P is the central node, E, W, N and S represents the neighbors, e, w, n and s are the faces
of a solid CV.

Energy:
~ l ii1 mg eg  el i1
me
~eg  ~el
i
 
om
~_ Dz
~el  el
qP
ot

 


 
 
omg
o~eg
o~el
o~p
mg
ml
 ADz
ot
ot
ot
ot
3

~_ ~e; ~p; q;
~_ ~s for a given CV, u
~ represents the average integral volume of any variable ue.g.; m;
 el is the
where u
average value estimated between the inlet and outlet of a CV. The subscripts and superscripts of the brackets
i1
represent the dierential value between the outlet0and inlet sections, i.e., ui ui1  ui . The transient terms
uu
ou
of these equations are approximated by ot Dt , where the superscript 0 denotes the value of a given variable at the earlier time step.
Eqs. (1)(3) are solved subject to the inlet boundary conditions given by the mass ow rate, the owing
pressure and either the temperature or the uid enthalpy, depending on the thermodynamic state (single or
two phase) that exists at the bottomhole conditions (Table 1). For non-adiabatic cases, the static formation
temperature distribution is also required as an initial boundary condition [10].
The uid ow regions inside a wellbore are dened with the following thermodynamic criteria:
Liquid region: h(p) < hl(p); xg = 0; eg = 0,

1626

Wellbore
diameter,
ID (m)

Mass ow
rate (kg s1)

Bottom-hole
pressure (bar)

Bottom-hole
temperature
(C)

Lip
pressureb
(bar)

Mass velocityc
[ton m2 s1]

Fluid
enthalpyd
(kJ kg1)

0
0
5.1

0.220
0.157
0.157

28.0

67.8

274.2

959
1250

0
0

0.220
0.157

26.7

47.2

2.68 0.03
2.68 0.13

1.38 0.01
1.38 0.07

1620 20
1620 102

850
1198

0
15

0.220
0.220

7.8

34.3

0.63 0.01
0.63 0.03

0.205 0.002
0.205 0.010

2592 33
2592 163

Geothermal
well

Depth
(m)

H-1

814
918
1200

Az-18
H-37

Wellbore
inclinationa
(degrees)

Wellbore inclination with respect to the vertical along the hole length; a roughness height of 45.52 106 m was considered for the three cases.
Minimum (upper-position) and maximum (lower-position) errors for the lip pressure measurement (1% and 5%, respectively).
c
Minimum (upper-position) and maximum (lower-position) errors for the mass velocity measurement (1% and 5%, respectively). Fluid enthalpy was computed by the RussellJames
equation (Eq. (12)).
d
Minimum (upper-position) and maximum (lower-position) errors of the calculated uid enthalpy.
a

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

Table 1
Input data used for analyzing the ow characteristics of the three production wells: H-1, Az-18 and H-37

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1627

Two phase region: hl(p) 6 h(p) 6 hg(p); 0 < xg < 1; 0 < eg < 1,
Steam region: h(p) > hg(p); xg = 1; eg = 1,
where hl(p) and hg(p) represent the liquid and steam saturation enthalpy for a given pressure p,
respectively.
For analyzing the heat conduction across the wellbore materials (pipe, cement or rock formation), a characteristic solid CV is additionally dened (Fig. 1). Heat conduction through these materials is simulated by
integrating the energy equation over a CV
!
o~
h
~
~
~
~
q_ w P w  q_ e P e Dz q_ s  q_ n A m
4
ot
_ are evaluated depending on the neighboring CVs. If the neighbor CV is a solid matewhere the heat uxes (~
q)
rial (e.g., rock formation, wellbore materials), the heat ux is computed by the equation: ~q_ koT =oz).
Otherwise, if the neighbor CV is the geothermal uid, the heat ux is estimated using the equation:
~
q_ aT wall  T f .
2.3. Empirical correlations and closure equations
Various empirical correlations are used for estimating the uid and heat ow parameters. Wall shear stress
is estimated through the friction factor (f). Two phase ow structure is calculated using the void fraction
parameter (eg). The heat transfer between the uid and the pipe wall temperature is computed through the
convective heat transfer coecient (CHTC:a).
2.3.1. Single phase ow
The CHTC is calculated by the dimensionless Nusselt number (Nu) and the Gnielinski equation for turbulent ow regimes [21]
fG =8Re  1000Pr
p
Nu
for Re > 2300
5
1 12:7 fG =8Pr2=3  1
2

where fG 1:82 logRe  1:64 .


The accuracy of the Gnielinski equation (Eq. (5)) is within 6% for Re ranging from 2000 to 106, and for Pr
from 0.6 to 2000 [21]. According to the denition of the Nu number (Nu aD=k) and considering the variable
uncertainties (sNu, sk, sD), the CHTC error (sa) can be computed by the following error propagation equation:
sa

a2

sNu
Nu

sk

k2

sD

D2

Since the uncertainties associated with the calculation of the water thermal conductivity and the wellbore
diameter measurement are very small, the quadratic terms of the Nu (Eq. (6)) would represent the main error
source to estimate the CHTC total propagated error. An approximated relationship 1:1 between Nu and
CHTC errors was found (i.e., 6%).
The friction factor (f) is computed with the empirical correlations cited by Sanchez-Upton [10].
" 
#1=12
12
8
1
f 2

7
1:5
Re
A B
where
"
A 2:457 ln

!#16

7=Re0:9 0:27e=D


and

37530
Re

16

The accuracy of this equation is within 3% when typical uncertainties of their variables are transported
[i.e., Re (20%, mainly due to the mass ow rate errors), e (5%) and D (1%)].

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1628

2.3.2. Two phase ow


(a) Condensation: If the wall temperature (Twall) is lower than the uid temperature (Tf), the CHTC is calculated by the Dobson and Chato correlations [22], which were proposed for annular ow:


2:22
0:8 0:4
Nu 0:023Re Pr
1 0:889
8
X tt
Dobson and Chato [22] reported that the accuracy of their correlation is within 5.3% for Re ranging from
1125 to 105 and Pr between 1 and 10.
(b) Evaporation: If the wall temperature (Twall) is higher than the uid temperature (Tf), the CHTC is computed by the Shah equation [23], which is given by
Nu 0:023Re0:8 Pr0:4 wshah

where the correction factor wshah is calculated depending on the heat ow pattern. The accuracy reported for
this equation is within 14% [23].
The void fraction and the friction factors, both for condensation and evaporation, are calculated as follows:
The void fraction (eg) is computed by a semi-empirical equation proposed by Rouhani and Axelsson [24] for a
drift ux model that considers the eects of the mass velocity (G), surface tension (c) and buoyancy (g), which
is given by
!(
! 
)
"
#
0:25 1
1:181  xg gcql  qg 
xg
xg
1  xg
eg
10
1 0:121  xg 

qg
qg
ql
Gq0:5
l
This equation generally provides good agreement with the water properties (including heavy waters) for owing pressures ranging from 19 bar to 140 bar. It has been reported that the void fraction estimation can be
obtained accurately to within 5% [24].
The friction factor (f) for two phase ow is calculated with the single phase ow correlation (Eq. (7)) and
multiplied by a correction factor (U), according to Beattie [25].
"
U 1 xg

ql
1
qg

!#0:8 (

"

3:5lg 2ll
1 xg
1
lg ll qg

#)0:2
11

2.3.3. Fluid enthalpy


The geothermal uid enthalpy is estimated by the well known RussellJames method [26], which is given by
the following equation:
h

917p0:871
G0:907

12

where h, p and G are the uid enthalpy (kJ kg1), the lip pressure (bar) and the mass velocity (ton m2 s1),
respectively. The enthalpy error (sh) can be accurately calculated by means of an error propagation equation
that transports all the individual variable uncertainties (sp, sG):
sh

h 2

0:759sp
p2

0:823sG
G2

13

According to this equation, it is expected that the uid enthalpy estimation can be obtained accurately to within 6.2%.
2.3.4. Thermophysical properties of uid and wellbore materials
The water thermophysical properties are computed using the commercial software NIST/ASME Steam
v.2.2 [27]. For the wellbore solid materials (cement, steel pipe and rock), these properties are obtained from
dierent databases reported in the literature [28].

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1629

2.4. Computational procedure


2.4.1. Wellbore model discretization
The spatial discretization of a geothermal wellbore is schematically shown in Fig. 2. The discretization
nodes for the uid ow process are dened at the inlet and outlet sections of the CVs. These nodes contain
nz+1 control volumes of length Dz (represented as shadow blocks in Fig. 2). For the wellbore solid materials,
the discretization nodes are considered at the solid CV centers with nz nr control volumes of length Dz and
width Dr (Fig. 2).
2.4.2. Discretization of the uid ow equations
For each CV of uid, a set of algebraic equations is obtained by discretization of the continuity, momentum
and energy governing equations, Eqs. (1)(3):

Fig. 2. Distribution of nodes along the geothermal wellbore model.

1630

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

For calculating the mass ow rate at the CV outlet, the discretized continuity equation is given by
"
#
0tp
ADz
qtp  q
m_ i1 m_ i 
Dt

14

where the two phase density (qtp) is dened by


qtp qg eg ql 1  eg

15

and the steam and liquid velocities by


!


_ g
_  xg
mx
m1
vg
; vl
ql 1  eg A
qg eg A
For computing the pressure at the CV outlet, the discretized momentum equation is given by
!
!)
  (
_  m
_ 0
_ 2 P
_ g vg 1  xg vl i1
Dz
mx
m
f m
i

pi1 pi 
 U

qtp Ag sin h
A
Dz
Dt
8
qtp A2

16

17

The rst three terms in the curly braces of this equation represent the pressure loss due to friction, gravitational and acceleration forces, and the last one corresponds to the transient pressure loss.
For calculating the enthalpy at the CV outlet, the energy Eq. (3) is combined with the continuity Eq. (1) for
obtaining the following equation:
"
#
2q_ wall  c1 m_ i1 c2 m_ i c3 ADz=Dt
hi1
18
m_ i1 m_ i 
q0tp ADz=Dt
where the variables c1, c2 and c3 are computed as follows:
2

c1 xg vg 1  xg vl i1 g sin hDz  hi


2

c2 xg vg 1  xg vl i  g sin hDz hi


2
0
0tp hi  2
v2i  q
0v0i
c3 2
pi  
p0i  q
hi  q
Eqs. (14), (17) and (18), together with the water thermophysical properties, are used to analyze the transient
two phase ow. Specic cases regarding steady state ow and/or single phase ow (liquid or steam) can be
analyzed by a simplied version of this model formulation (i.e., by neglecting the time terms and/or by considering the steam or liquid quality in the single phase ow).
2.4.3. Solution algorithm: Fluid ow
The discretized Eqs. (14), (17) and (18) are coupled in the ow direction (z) using a fully implicit step by step
_ in a given CV are iteratively calculated using their
method (Fig. 2). The output variables (e.g., p; T ; h; m)
known values at the CV inlet and taking into account the pipe wall temperature distribution. These output
calculations constitute the input data for the next CV. This iterative procedure is continued up the wellbore
until the wellhead is reached where the boundary conditions established for Eqs. (1)(3) are considered. The
ow transition zones are identied using the thermodynamic criteria already dened in Section 2.2. The accurate position of these zones is determined by dividing the corresponding CV into two new CVs. The length of
the rst generated CV is computed using the momentum equation and the thermodynamic criteria, while the
second one is calculated by a dierence in lengths between the initial CV (which was divided) and the rst generated CV. This ow transition criterion enables the consumed CPU time to be optimized [29].
2.4.4. Solution algorithm: Heat conduction through the wellbore materials
The following equation is obtained for each solid node of the mesh (Fig. 2)
aP T wall;i;j aN T wall;i1;j aS T wall;i1;j aE T wall;i;j1 aW T wall;i;j1 d P T 0wall;i;j

19

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1631

where the coecients are given by


aN

kn A
;
Dz

aS

ks A
;
Dz

aW

kw P w Dz
;
Dr

aE

ke P e Dz
;
Dr

dP

ADz
qcp ;
Dt

aP aN aS aE aW d P

These coecients are valid for 2 6 i 6 nz  1 and 2 6 j 6 nr  1. For other nodes, the following considerations
are applied:
When j = 1, forced convection is assumed for the node West face. The North and South faces are represented by the pipe thermal conductivity, while for the East face, either the cement or formation thermal
conductivity is employed. All these properties are evaluated at the mean temperature of the neighboring
CVs.
When j = nr, the static formation temperature and the heat conduction through the formation at a distance
of Dr/2 are considered.
For i = 1 and i = nz, the axial heat conduction or the temperature boundary conditions are used.

2.4.5. Global solution algorithm


The coupled solution of the uid ow inside a geothermal wellbore and the heat conduction across the wellbore materials is reached using a global numerical algorithm, which is solved in a segregated manner. The steady state solution is solved using a pseudo-transient calculation with a guessed initial condition (e.g., uid ow
and uniform temperature conditions in the whole domain). For the transient solutions, the initial conditions
must be specied (i.e., a complete distribution of the variables in the whole domain).
3. Methodology
The capability of GEOWELLS to model the ow characteristics in producer geothermal wells is demonstrated. The wellbore simulator is rst validated using a well known heat transfer problem reported in the literature. To verify that GEOWELLS can be reliably employed for modeling two phase ow inside geothermal
wells, numerical analyses of three producer wells drilled in Mexican geothermal elds are used. The simulated
results provided by GEOWELLS are statistically compared with both actual downhole stable measurements
logged in each well and with those simulation results predicted by an existing wellbore simulator
(ORKISZEWSKI) [20].
Matching analyses between the simulated and measured data are evaluated not only by eye graphical
method but also through a statistical analysis based on the residual sum of squares (RSS) and the non-parametric method of Chi-square (vi). Finally, sensitivity and uncertainty analyses for evaluating the eect of some
input data variations on the numerical simulations was performed.
3.1. Validation of GEOWELLS
The simulator is validated against the analytical solution of a well known heat transfer problem proposed
by Carslaw and Jaeger [30]. A steady state analysis based on the uid ow inside a hollow cylinder (0.2205 m
I.D. and 2.0 m O.D.) of constant thermal conductivity (1.686 W m1 K1) is employed. For this analysis, a
uid with a constant convective heat transfer coecient of 200 W m2 K1is owing at a constant inner temperature of 250 C, whereas the outer cylinder face is maintained at a constant temperature of 300 C.
3.2. Potential applications of GEOWELLS
The capability of GEOWELLS for predicting downhole pressure and temperature proles in producer wells
using a suitable number of CVs to provide grid independent solutions is also demonstrated. Such numerical
simulations enable the total pressure drop inside a geothermal wellbore, the wellhead enthalpy, the ashing
point, the steam quality, and steam and liquid velocities to be computed.

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1632

3.2.1. Wellbore data sources


Three data sets based on eld measurements logged during production tests of geothermal producer wells
of the Los Azufres, Mich. (Az-18) and Los Humeros, Pue. (H-1 and H-37) are used (Table 1). These wells are
currently under exploitation for the generation of electricity in Mexico [31]. The geothermal wells have varied
geometries (vertical and inclined cased sections) and a wide range of discharge rates and ow enthalpies (i.e.,
low, moderate and high enthalpy wells: H-1, Az-18 and H-37, respectively).
3.2.2. Matching analysis: RSS and Chi-square (vi)
The matching analysis between the simulated and measured data is evaluated through the residual sum of
squares (RSS) and non-parametric statistical analysis based on the absolute value of dierences between expected (measured) and observed (simulated) frequencies (i.e., sum of residuals: Chi-square, vi) [32]. The RSS is
computed as follows:
RSS

n
X
usim  umeas 2

20

i1

where subscripts sim and meas represent the simulated and measured values, either for pressure or temperature.
The sum of residuals (vi) is calculated as
vi

n
X
usim  umeas 2
umeas
i1

21

3.2.3. Sensitivity and uncertainty analysis


Most of the computer simulations that analyze the uid ow inside geothermal wells are usually performed
without a rigorous scrutiny of the input data (including the measurement errors) [7], which has invariably revealed discrepancies between simulated and measured eld data. To minimize this weakness, sensitivity and
uncertainty analyses were conducted for all the GEOWELLS simulations to investigate how variations in input data (due to measurement errors) aect the results.
The mathematical model of GEOWELLS uses input variables (e.g., p, T, h) and ow correlations (e.g., such
as void fraction, friction factor, etc.) whose values are not strictly known with precision and accuracy, since
they actually vary within an experimental error interval [16]. Furthermore, the numerical methods used by
GEOWELLS to solve the equations also introduce errors themselves. The eect of such errors and/or input
data uncertainties is quantied to assess the simulators range of validity.
4. Results and discussion
4.1. Numerical validation of GEOWELLS
For solving the validation problem mentioned in Section 3.1, GEOWELLS was allowed to run under pseudo-transient conditions to reach steady state before the simulated temperature data were compared with the
analytical solution. An excellent agreement between the simulated and analytical proles was found (Fig. 3).
Errors less than 1.4% were consistently obtained for two CVs. However, when the number of CVs was increased to 10, the prediction errors were less than 0.22%.
4.2. Application cases
4.2.1. Los Humeros geothermal well (H-1)
According to the deeper inow conditions (Table 1), the uid enters as a liquid with a subcooling degree of
9.5 C. These production characteristics were considered as input data for performing the numerical analysis
with GEOWELLS (using 1200 CVs; Dz = 1 m) and ORKISZEWSKI. In both cases, the simulated pressure
and temperature proles were successfully computed. All these simulated results together with the eld measurements are summarized in Tables 2 and 3.

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1633

300

Temperature [oC]

290

280

270

260
Numerical solution
Analytical solution

250

0.1

0.2

0.3

0.4 0.5 0.6 0.7


Radial position [m]

0.8

0.9

Fig. 3. Numerical validation results obtained by GEOWELLS (using 10 grid nodes) for a heat transfer problem described in Section 3.1.

The total pressure drop calculated by GEOWELLS showed that 97.3% was mainly attributable to the gravitational pressure gradient, 2.5% to the friction forces and the remaining 0.2% was due to the acceleration effects. The pressure proles predicted by GEOWELLS and ORKISZEWSKI show good agreement with the
measured data from 1200 m to 800 m (Table 2, H-1). However, at shallower wellbore depths (between
20 m and 700 m), some small dierences (probably due to an overestimation eect of the gravitational forces)
were found, especially for those simulated results predicted by ORKISZEWSKI. These observations were conrmed by a matching analysis between the simulated results (predicted by GEOWELLS and ORKISZEWSKI) and measured data. The analysis was conducted through the statistical evaluation of RSS and vi
(Table 2). RSS values upto 26.4 and 105.7 were computed for the GEOWELLS and ORKISZEWSKI simulations, respectively; whereas the calculated vi values varied from 0 to 1.3 and from 0 and 5.2, in that
order. Lower values of the sum of vi residuals were also found for GEOWELLS (4.8) in comparison with
those corresponding to ORKISZEWSKI (19.8). From these calculations, it is statistically evident that the
best agreement results were provided by GEOWELLS.
According to the temperature proles predicted by GEOWELLS and ORKISZEWSKI, the ashing point
for this well was located at ca. 1078 m. From this depth, a two phase mixture ows upto the wellhead. The
simulated temperature proles by both simulators showed good matching with the measured data (from
the 1200 m to 800 m), with some exceptions at shallower depths (between 20 m and 700 m) where the ORKISZEWSKI results show some signicant dierences (Table 3, H-1). RSS values upto 21.9 and 203.9 were
calculated for the GEOWELLS and ORKISZEWSKI simulations, respectively; whereas the vi values varied
from 0 to 0.098 and from 0 and 0.913, in that order. The sum of vi residuals also shows lower values for
GEOWELLS (0.4) in comparison with those related to ORKISZEWSKI (3.1). The best matching results
were also provided by GEOWELLS simulations.
4.2.2. Los Azufres geothermal well (Az-18)
In relation to the inow conditions of this well (Table 1), the uid enters as wet steam with a quality of 0.29.
These ow characteristics were considered as input data for executing the GEOWELLS (using 1250 CVs and a
similar mesh size Dz = 1 m) and ORKISZEWSKI simulations. In both cases, the pressure and temperature
proles were successfully obtained. The results of these simulations, together with the measured data, are also
summarized in Tables 2 and 3, respectively (Az-18). The total pressure drop calculated by GEOWELLS
showed that 85.1% was attributable to the gravitational pressure term, 14.7% attributable to friction forces
and 0.2% was due to the acceleration term.
The pressure proles predicted by GEOWELLS show an excellent agreement with the measured data for
the whole wellbore depth, whereas the ORKISZEWSKI results showed poor matching at the shallower zones

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1634

Table 2
Wellbore pressure measurements and simulation results obtained by GEOWELLS and ORKISZEWSKI
Wellbore
depth

Measured
pressure (bar)

GEOWELLS
Simulated
pressure (bar)

Squared residuals
(psim  pmeas)2

vi residuals

Simulated
pressure (bar)

Squared residuals
(psim  pmeas)2

vi residuals

H-1
20
100
200
300
400
500
600
700
800
900
1000
1100
1200

20.4
22.1
24.5
27.0
29.9
32.6
36.4
39.7
43.4
48.0
53.5
59.6
67.8

25.6
27.0
29.0
31.1
33.3
35.8
38.5
41.7
44.9
48.5
53.5
60.2
67.8

26.420
24.010
20.070
16.565
11.560
9.922
4.410
4.080
2.132
0.281
0.004
0.384
0.000
119.838a

1.294
1.084
0.819
0.613
0.386
0.304
0.121
0.103
0.049
0.006
0.000
0.006
0.000
4.786a

30.7
32.1
33.7
35.4
37.2
39.3
41.1
42.8
45.0
48.4
53.2
59.9
67.8

105.678
99.202
84.456
70.560
52.708
44.489
22.468
9.486
2.434
0.168
0.102
0.109
0.000
491.859a

5.175
4.481
3.446
2.613
1.760
1.363
0.618
0.239
0.056
0.004
0.002
0.002
0.000
19.759a

Az-18
0
100
200
300
400
500
600
700
800
900
1000
1050
1100
1150
1200
1250

30.7
31.9
32.9
33.8
34.9
35.9
36.9
38.1
39.1
40.2
41.5
42.6
43.6
44.5
45.7
47.2

29.8
31.0
32.2
33.4
34.6
35.9
37.2
38.5
39.8
41.1
42.6
43.5
44.4
45.3
46.3
47.2

0.846
0.865
0.533
0.168
0.073
0.000
0.073
0.130
0.462
0.846
1.232
0.846
0.689
0.723
0.336
0.000
7.823a

0.028
0.027
0.016
0.005
0.002
0.000
0.002
0.003
0.012
0.021
0.030
0.020
0.016
0.016
0.007
0.000
0.205a

33.5
34.3
35.1
35.9
36.8
37.6
38.5
39.3
40.2
41.1
42.4
43.4
44.3
45.3
46.3
47.2

7.840
5.760
4.840
4.410
3.610
2.890
2.560
1.440
1.210
0.810
0.810
0.640
0.490
0.640
0.360
0.000
38.310a

0.255
0.181
0.147
0.130
0.103
0.081
0.069
0.038
0.031
0.020
0.020
0.015
0.011
0.014
0.008
0.000
1.124a

H-37
20
100
200
300
400
500
600
700
800
900
1000
1100
1200

30.0
30.6
31.1
31.2
31.4
32.0
32.4
32.7
33.0
33.3
33.6
34.1
34.3

30.5
30.7
31.1
31.4
31.7
32.0
32.4
32.7
33.0
33.4
33.7
34.0
34.3

0.230
0.029
0.005
0.036
0.068
0.005
0.000
0.000
0.001
0.002
0.012
0.005
0.000
0.393a

0.008
0.001
0.000
0.001
0.002
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.013a

29.5
29.7
30.0
30.3
30.6
30.8
31.1
31.4
31.7
32.2
32.9
33.6
34.3

0.250
0.740
1.277
0.792
0.706
1.346
1.638
1.664
1.822
1.232
0.449
0.230
0.000
12.146a

0.008
0.024
0.041
0.025
0.022
0.042
0.051
0.051
0.055
0.037
0.013
0.007
0.000
0.377a

ORKISZEWSKI

Sums of residuals (square and Chi-square).

(from 959 m to the surface): Table 2. This observation was also veried by the matching analysis conducted
between the simulated and measured data. RSS values upto 1.2 and 7.8 were found for the GEOWELLS

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1635

Table 3
Wellbore temperature measurements and simulation results obtained by GEOWELLS and ORKISZEWSKI
Wellbore
depth

Measured
temperature (C)

GEOWELLS

H-1
20
100
200
300
400
500
600
700
800
900
1000
1100
1200

221.3
223.5
228.8
234.3
238.9
244.4
250.8
256.1
260.7
265.1
268.9
272.6
274.2

225.1
228.1
232.0
235.8
239.8
243.9
248.1
252.6
257.4
262.1
268.1
273.9
274.2

Az-18
0
100
200
300
400
500
600
700
800
900
1000
1050
1100
1150
1200
1250

238.0
242.5
244.0
246.0
248.0
249.0
251.0
253.0
254.0
256.0
257.5
258.8
260.2
261.5
263.0
267.0

H-37
20
100
200
300
400
500
600
700
800
900
1000
1100
1200

227.1
230.0
231.0
231.7
232.6
233.5
233.9
234.6
235.4
236.5
236.8
237.4
238.3

ORKISZEWSKI
vi residuals

Simulated
temperature (C)

Squared residuals
(Tsim  Tmeas)2

vi residuals

14.900
21.902
10.498
2.341
0.980
0.240
7.236
12.888
11.022
8.762
0.624
1.563
0.000
92.955a

0.067
0.098
0.046
0.010
0.004
0.001
0.029
0.050
0.042
0.033
0.002
0.006
0.000
0.389a

235.2
237.7
240.4
243.1
246.1
249.3
252.1
254.4
257.5
261.9
267.9
274.2
274.2

193.210
203.918
135.490
77.969
52.418
23.717
1.588
3.028
10.176
9.986
1.082
2.250
0.000
714.829a

0.873
0.913
0.592
0.333
0.219
0.097
0.006
0.012
0.039
0.038
0.004
0.008
0.000
3.135a

233.4
235.6
237.8
239.9
241.9
244.0
246.0
248.0
250.0
252.0
254.1
255.4
256.6
257.9
259.1
260.4

20.885
47.472
39.063
37.577
36.724
25.000
24.701
24.602
15.761
16.080
11.424
11.492
12.603
12.960
14.900
44.090
395.331a

0.088
0.196
0.160
0.153
0.148
0.100
0.098
0.097
0.062
0.063
0.044
0.044
0.048
0.050
0.057
0.165
1.574a

240.0
241.4
242.8
244.1
245.5
246.8
248.1
249.4
250.7
252.0
253.8
255.2
256.6
257.9
259.1
260.3

4.162
1.166
1.488
3.497
6.401
4.884
8.410
12.960
10.956
16.241
13.323
12.674
13.177
13.177
14.823
44.489
181.826a

0.017
0.005
0.006
0.014
0.026
0.020
0.034
0.051
0.043
0.063
0.052
0.049
0.051
0.050
0.056
0.167
0.704a

234.7
235.2
235.8
236.3
236.9
237.5
238.1
238.6
239.2
239.8
240.3
240.9
241.5

58.064
26.420
23.040
21.622
18.662
15.682
17.389
16.000
14.746
10.956
12.320
12.320
9.860
257.081a

0.256
0.115
0.100
0.093
0.080
0.067
0.074
0.068
0.063
0.046
0.052
0.052
0.041
1.108a

232.9
233.3
233.9
234.4
234.9
235.4
235.9
236.4
236.9
237.8
239.1
240.3
241.4

33.872
10.890
8.468
7.236
5.198
3.572
4.080
3.168
2.403
1.769
4.884
8.526
9.797
103.864a

0.149
0.047
0.037
0.031
0.022
0.015
0.017
0.014
0.010
0.007
0.021
0.036
0.041
0.448a

Simulated
temperature (C)

Squared residuals
(Tsim  Tmeas)2

Sums of residuals (square and Chi-square).

and ORKISZEWSKI simulations, respectively; whereas for the vi evaluation, values upto 0.03 and 0.26
were obtained. The sum of residuals vi also showed lower values for GEOWELLS (0.2) in comparison with

1636

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

those corresponding to ORKISZEWSKI (1.1). According to these statistical parameters, the best agreement
results were provided by GEOWELLS.
The temperature proles predicted either by GEOWELLS or ORKISZEWSKI show some dierences
(upto 7 C at the bottom conditions) against the measured data (Table 3, Az-18). Such dierences are mainly
due to the use of water thermophysical properties, which are not suitable to describe the thermodynamic
behavior of the uid produced in this well, which is characterized by a high content of salts (6949 ppm as
NaCl) and gases (7.5 wt% as CO2) [31]. RSS values upto 47.4 and 44.5 were computed for the
GEOWELLS and ORKISZEWSKI simulations, respectively; while for the vi estimation, similar values (upto
0.2) were found. The sum of residuals vi for GEOWELLS and ORKISZEWSKI showed values of 1.6 and
0.7, respectively. For these cases, similar matching results were predicted by both GEOWELLS and
ORKISZEWSKI.
4.2.3. Los Humeros geothermal well (H-37)
The geothermal uid produced by this well enters at the bottom feed zone with a wet steam quality of 0.88.
Such a thermodynamic condition, together with the other ow characteristics (Table 1), were used as input
data for running the simulations with GEOWELLS (using 1198 CVs and a similar mesh size Dz = 1 m)
and ORKISZEWSKI. In both cases, the pressure and temperature proles were successfully obtained (Tables
2 and 3, respectively). The total pressure drop computed by GEOWELLS showed that 93.6% was attributable
to the gravitational pressure term, 6.3% by the frictional pressure gradient and only 0.1% by the acceleration
pressure term.
The pressure proles predicted by GEOWELLS show a better agreement with the measured data in comparison with those obtained for ORKISZEWSKI, which underestimate the pressure along the whole wellbore
depth. The matching analysis between the simulated and measured pressure show that the RSS values computed for GEOWELLS (upto 0.2) were less than those obtained for ORKISZEWSKI (upto 1.8); whereas
the vi values were upto 0.008 and 0.055, respectively.
The sum of vi residuals also showed lower values for GEOWELLS (0.013) than those related to
ORKISZEWSKI (0.38). This analysis conrms that the best agreement results were obtained by the
GEOWELLS simulations.
The temperature proles predicted either by GEOWELLS or ORKISZEWSKI show some dierences (upto
7 C at the bottom conditions) against the measured data (Table 3, H-37). Such dierences are also due to the
salinity (1370 ppm as NaCl) and the gas content (1.4 wt% as CO2) of the uid produced in this well. The
matching analysis of these proles showed RSS values upto 58 and 34 for the GEOWELLS and
ORKISZEWSKI simulations, respectively; while for the vi estimation, values upto 0.25 and 0.15 were
found. The sum of residuals vi for the GEOWELLS and ORKISZEWSKI simulations show values of 1.1
and 0.45, respectively. For these simulation cases, similar matching results were provided either by
GEOWELLS or ORKISZEWSKI.
Adiabatic and non-adiabatic (considering the heat losses between the well and the surrounding formation)
heat ow conditions were also analyzed for this well. Very small dierences between the adiabatic and nonadiabatic simulation results were found. This behavior is totally expected due to the mass ow rate produced
in this well (7.8 kg s1), which allows the heat conduction between the well and the formation to be ignored
[33].
4.2.4. Global matching analysis
A global matching analysis of the RSS calculated for all the standard simulations was additionally performed. Four box whisker plots were drawn for presenting these comparative results (Fig. 4). The rst two
box whisker plots correspond to the RSS values computed for the pressure proles predicted by each simulator (GEOWELLS and ORKISZEWSKI), whereas the other two plots are related to the temperature proles. Each whisker box plot used a sample of 42 RSS values. This statistical analysis shows that the best
matching results between the simulated and measured data were obtained for the GEOWELLS simulations,
which are clearly evidenced by the box size and the number of extreme and outlier values identied for each
RSS sample.

Residual sum of squares (RSS)

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

200

Mean
MeanSE
Extremes

Mean1.96*SE

1637

Outliers

150

100

50

0
GEOWELLS (P)

GEOWELLS (T)

ORKISZEWSKI (P)

ORKISZEWSKI (T)

Fig. 4. RSS box whisker plots of pressure and temperature proles computed by GEOWELLS and ORKISZEWSKI simulations. (P) and
(T) represent the pressure and temperature proles, respectively.

4.3. Results of the sensitivity and uncertainty analysis


To perform this analysis, input data such as the bottomhole pressure and temperature, the uid enthalpy
and the void fraction were selected as sensitivity parameters. The sensitivity and uncertainty analysis was conducted simply by giving one of these parameters a high and low value (uncertainties), while the other input
data were kept constant [34].
The uncertainties of the input pressure and temperature were assumed by the minimum and maximum
measurement errors (pressure: 0.4 bar and 2 bar, and temperature: 1 C and 5 C). The smaller errors
represent a careful and well controlled measurement using high precision equipment, while the larger errors
correspond to typical values present in logging operations [7,13,16,33]. The enthalpy uncertainty was calculated by the error propagation equation (Eq. (13)) using lip pressure and uid mass velocity measurements
together with their experimental errors (Table 1). The uncertainties associated with the calculation of the void
fraction (Eq. (10)) were represented by the minimum and maximum errors that are within the accuracy of the
correlation (1% and 5%) [24]. All the uncertainties assumed for the sensitivity parameters are compiled in
Table 4.
The sensitivity and uncertainty analysis was conducted to investigate how the uncertainties assumed in
input parameters aect the matching results predicted by GEOWELLS in the initial standard simulations
(Tables 2 and 3, which are represented as solid-lines in Figs. 57AF). Minimum (MIN) and maximum
(MAX) prediction limits for pressure and temperature were computed for all cases. The simulated pressure
and temperature proles obtained for the wells H-1, Az-18 and H-37 are shown in Figs. 57AF, respectively.
The MIN prediction limits obtained either for the pressure or temperature proles are shown as dashed lines,
whereas the MAX prediction limits are denoted by dotted lines. A summary of the sensitivity and uncertainty
analysis is also described in Table 4, where the MINMAX simulations are linked to their corresponding gures to evaluate the eect of the sensitivity parameters.
4.3.1. Sensitivity and uncertainty analysis: H-1, Az-18 and H-37
4.3.1.1. Well H-1. The calculated pressure and temperature proles (represented by the standard simulations
and the MINMAX prediction limits) are compared with eld measurements in Fig. 5AF. For the input
pressure variations, it was found that this parameter does not have an important eect on the simulated pressure and temperature proles (Fig. 5A and B). On the other hand, the input temperature variations improved
the matching between the simulated and measured proles, especially for those results related to the MAX
prediction limits (Fig. 5C and D). For the void fraction variations, a negligible eect on the simulated pressure
proles was found (Fig. 5E), while for the temperature proles such variations produced a better agreement
against the eld measurements (mainly for the MAX prediction limits: Fig. 5F).

1638

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

Table 4
Numerical simulation cases executed by GEOWELLS for the sensitivity and uncertainty analysis
Input data (sensitivity
parameters) uncertainties

Geothermal
well

Simulated proles
p  T (prediction limits)

Figures (simulated prole)

pin 0.4 bar

H-1

MIN

Fig.
Fig.
Fig.
Fig.

5A (pressure)
5B (temperature)
5A (pressure)
5B (temperature)

Fig.
Fig.
Fig.
Fig.

5C (pressure)
5D (temperature)
5C (pressure)
5D (temperature)

Fig.
Fig.
Fig.
Fig.

5E
5F
5E
5F

Fig.
Fig.
Fig.
Fig.

6A (pressure)
6B (temperature)
6A (pressure)
6B (temperature)

Fig.
Fig.
Fig.
Fig.

6C (pressure)
6D (temperature)
6C (pressure)
6D (temperature)

Fig.
Fig.
Fig.
Fig.

6E
6F
6E
6F

Fig.
Fig.
Fig.
Fig.

7A (pressure)
7B (temperature)
7A (pressure)
7B (temperature)

Fig.
Fig.
Fig.
Fig.

7C (pressure)
7D (temperature)
7C (pressure)
7D (temperature)

Fig.
Fig.
Fig.
Fig.

7E
7F
7E
7F

pin 2.0 bar


Tin 1 C

MAX
H-1

Tin 5 C
eg 1%

MAX
H-1

eg 5%
pin 0.4 bar

Az-18

Az-18

Az-18

H-37

eg 5%

MIN
MAX

H-37

hin 162.9 kJ/kg


eg 1%

MIN
MAX

pin 2.0 bar


hin 32.6 kJ/kg

MIN
MAX

eg 5%
pin 0.4 bar

MIN
MAX

hin 101.9 kJ/kg


eg 1%

MIN
MAX

pin 2.0 bar


hin 20.4 kJ/kg

MIN

MIN
MAX

H-37

MIN
MAX

(pressure)
(temperature)
(pressure)
(temperature)

(pressure)
(temperature)
(pressure)
(temperature)

(pressure)
(temperature)
(pressure)
(temperature)

The basis of the sensitivity analyses was simply to give one of the input parameters a high and low value (uncertainties), while the others
input data were kept constant (see Table 1). The minimum (MIN) prediction limits obtained either for pressure or temperature proles are
shown in all the gures as dashed-curves, whereas the maximum (MAX) prediction limits are depicted by dotted-curves.

4.3.1.2. Well Az-18. The computed pressure and temperature proles (represented by the standard simulations
and the MINMAX prediction limits) are also compared with measured data in Fig. 6AF. Fig. 6A shows
that the pressure proles resulted from minimum variations in input pressure (0.4 bar) provide an excellent
matching with the measured data. These minimum variations were not enough to match correctly with the
measured temperature data (Fig. 6B), although for pressure variations up to 2 bar, this comparison was
slightly improved (Fig. 6B). However, underestimation problems of temperature, probably due to the use
of water thermophysical properties, are still present in this well. A similar behavior was also observed when
the uid enthalpy was varied (1620 20 and 1620 102 kJ kg1: Fig. 6C and D). For the void fraction
changes, a strong eect on both the simulated pressure and temperature proles was systematically observed

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

20

30

Pressure [bar]
40
50

60

70

210

220

230

Temperature [oC]
240
250

pin (Table 1)
pin 0.4 bar (MIN)
pin 2.0 bar (MAX)
experimental data

200

260

1639

270

280

pin (Table 1)
pin 0.4 bar (MIN)
pin 2.0 bar (MAX)
experimental data

Depth [m]

400

600

800

1000
Flashing
Point

1200

[B] H-1

[A] H-1

Tin (Table 1)
o
Tin 1 C (MIN)
o
Tin 5 C (MAX)
experimental data

200

Tin (Table 1)
o
Tin 1 C (MIN)
o
Tin 5 C (MAX)
experimental data

Depth [m]

400

600

800

1000
Flashing
Point

1200

[C] H-1
[D] H-1

0
g
g 1 % (MIN)
g 5 % (MAX)
experimental data

200

g
g 1 % (MIN)
g 5 % (MAX)
experimental data

Depth [m]

400

600

800

1000
Flashing
Point

1200

[E] H-1
20

[F] H-1
30

40
50
Pressure [bar]

60

70

210

220

230

240
250
Temperature [oC]

260

270

280

Fig. 5. Pressure and temperature proles measured in discharging well H-1. Error bars represent the measurement errors of pressure and
temperature (according to the specications of transducers 0.4 bar and thermocouples 1 C). Standard pressure and temperature
proles computed by GEOWELLS (without an error transport) are denoted by solid lines (these data are also compiled in Tables 2 and 3).
The MINMAX prediction limits predicted from the sensitivity and uncertainty analysis are represented by dashed lines and dotted lines,
respectively. (A, B) and (C, D) show the eect of the bottomhole pressure and temperature uncertainties, respectively. The eects of the
void fraction uncertainties are represented in E and F.

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1640

28

32

Pressure [bar]
36
40

44

48

230

Temperature [oC]
240
250

260

270

0
pin (Table 1)
pin 0.4 bar (MIN)
pin 2.0 bar (MAX)
experimental data

200

pin (Table 1)
p in 0.4 bar (MIN)
p in 2.0 bar (MAX)
experimental data

Depth [m]

400

600

800

1000

1200

[B] Az-18

[A] Az-18

0
hin (Table 1)
hin 20.4 kJ/kg (MIN)
hin 101.9 kJ/kg (MAX)
experimental data

hin (Table 1)
hin 20.4 kJ/kg (MIN)
hin 101.9 kJ/kg (MAX)
experimental data

200

Depth [m]

400

600

800

1000

1200

[C] Az-18

[D] Az-18

0
g
g 1 % (MIN)
g 5 % (MAX)
experimental data

200

g
g 1 % (MIN)
g 5 % (MAX)
experimental data

Depth [m]

400

600

800

1000

1200

[E] Az-18
28

32

[F] Az-18
36
40
Pressure [bar]

44

48

230

240
250
Temperature [oC]

260

270

Fig. 6. Pressure and temperature proles measured in discharging well Az-18. Error bars represent the measurement errors of pressure and
temperature (according to the specications of transducers 0.4 bar and thermocouples 1 C). Standard pressure and temperature
proles computed by GEOWELLS (without an error transport) are denoted by solid lines (these data are also compiled in Tables 2 and 3).
The MINMAX prediction limits predicted from the sensitivity and uncertainty analysis are represented by dashed lines and dotted lines,
respectively. (A, B) and (C, D) show the eect of the bottomhole pressure and enthalpy uncertainties, respectively. The eects of the void
fraction uncertainties are represented in E and F.

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

26
0

28

30

Pressure [bar]
32

34

36

225

230

Temperature [oC]
235
240

pin (Table 1)
pin 0.4 bar (MIN)
pin 2.0 bar (MAX)
experimental
data

200

1641

245

250

pin (Table 1)
pin 0.4 bar (MIN)
pin 2.0 bar (MAX)
experimental
data

Depth [m]

400

600

800

1000

1200

[A] H-37

[B] H-37

0
hin (Table 1)
hin 32.6 kJ/kg (MIN)
hin 162.9 kJ/kg (MAX)
experimental
data

200

hin (Table 1)
hin 32.6 kJ/kg (MIN)
hin 162.9 kJ/kg (MAX)
experimental
data

Depth [m]

400

600

800

1000

1200

[C] H-37

[D] H-37

0
g
g 1 % (MIN)
g 5 % (MAX)
experimental

g
g 1 % (MIN)
g 5 % (MAX)
experimental

200

data

data

Depth [m]

400

600

800

1000

1200
26

[E] H-37
28

[F] H-37
30

32
Pressure [bar]

34

36

225

230

235
240
Temperature [oC]

245

250

Fig. 7. Pressure and temperature proles measured in discharging well H-37. Error bars represent the measurement errors of pressure and
temperature (according to the specications of transducers 0.4 bar and thermocouples 1 C). Standard pressure and temperature
proles computed by GEOWELLS (without an error transport) are denoted by solid lines (these data are also compiled in Tables 2 and 3).
The MINMAX prediction limits predicted from the sensitivity and uncertainty analysis are represented by dashed lines and dotted lines,
respectively. (A, B) and (C, D) show the eect of the bottomhole pressure and enthalpy uncertainties, respectively. The eects of the void
fraction uncertainties are represented in E and F.

1642

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

(Fig. 6E and F), which produced a better agreement against the measured data (especially for variations of
5%).
4.3.1.3. Well H-37. The calculated pressure and temperature proles (represented by the standard simulations
and the MINMAX prediction limits) are compared with actual eld data in Fig. 7AF. Fig. 7A shows that
the pressure proles that result from the minimum variations of the input pressure (0.4 bar) provide an excellent match with the eld measurements. These pressure variations were not sucient to match correctly the
simulated temperature proles with the measured data. For the maximum pressure variations (upto 2 bars),
this comparison was slightly improved (Fig. 7B), although small overestimation dierences were still observed
at the shallower depths. These dierences were also attributed to the water thermophysical properties that
were employed to approach the thermodynamic behavior of the geothermal uids produced. A very similar
behavior was observed when the uid enthalpy was varied (2592 33 and 2592 163 kJ kg1; Fig. 7C and
D). For the void fraction variations, a strong eect on both the simulated pressure and temperature proles
was systematically observed (Fig. 7E and F), which produced a better agreement against eld measurements
(especially for variations of 5%).
5. Concluding remarks
A new steady and transient one dimensional simulator (GEOWELLS) for analyzing the ow inside geothermal wells has been developed. The uid ow and heat conduction governing equations were eciently
solved. The simulator capability was successfully validated with analytical data and applied for the analysis
of production of some Mexican geothermal wells. In most of these applications, the standard simulation results provided by GEOWELLS were in good agreement with measured eld data and were supported by statistical analyses based on residuals (RSS and Chi-square).
From the sensitivity and uncertainty analysis, it was found that the void fraction is one of the most sensitivity parameters that aect the numerical capability of wellbore simulators to predict correctly the pressure
and temperature proles. Reliable and more accurate correlations to estimate this parameter are, therefore,
needed for geothermal wells. Additional experimental and computer work are also required to include an
appropriate thermophysical properties package for considering the eects of salts and non-condensable gases.
It is also important to point out that the sensitivity and uncertainty analysis performed in this work only
provided an approach of the eects that would result from the transport of the input data uncertainties on the
simulations. Monte Carlo techniques are required to be considered in the future for combining all these uncertainties to have a much better estimation of the total propagated error in these simulations. Of course, such
studies will require a comprehensive knowledge of the measurement errors, which will only be obtained if a
large number of measurements are collected.
Acknowledgements
We are grateful to the anonymous reviewers and the editor for constructive comments on an earlier version
of this manuscript. The second author acknowledges support from Posgrado de IngenieraEnerga
(UNAM). This work was partly supported by DGAPA-PAPIIT project IN104703-3.
References
[1] Garca A, Ascencio F, Espinosa G, Santoyo E, Gutierrez H, Arellano V. A numerical modeling study of high-temperature deep wells
in the Cerro Prieto geothermal eld, Mexico. Geos Int 1999;38:25160.
[2] Moya SL, Aragon A, Iglesias E, Santoyo E. Prediction of mass deliverability from a single wellhead measurement and geothermal
inow performance reference curves. Geothermics 1998;27:31729.
[3] Khasani, Itoi R, Tanaka T, Fukuda M. An analysis of pressure drops in wellbore under low ow rate conditions shown on a
deliverability curve. In: Proceedings of the 24th NZ geothermal workshop, New Zealand, 2002. p. 11520.
[4] Tian SF, Finger JT. Advanced geothermal wellbore hydraulics model. J Energy Resourc TechnolTrans ASME 2000;122:1426.
[5] Hadgu T, Freeston D. A multi-purpose wellbore simulator. Geothermal Resourc Council, Trans 1990;14:127986.

O. Garca-Valladares et al. / Energy Conversion and Management 47 (2006) 16211643

1643

[6] Hadgu T, Zimmerman RW, Bodvarsson GS. Coupled reservoir-wellbore simulation of geothermal reservoir behavior. Geothermics
1995;24:14566.
[7] Gunn CIM, Freeston DH, Hadgu T. Principles for wellbore simulator validation and calibration using matching analysisI.
Analytical techniques. Geothermics 1992;21:34161.
[8] Chadha PK, Malin MR, Palacio-Perez A. Modelling of two-phase ow inside geothermal wells. Appl Math Model 1993;17:23645.
[9] Gorine AV. Two phase ow governing equations for transient geothermal well and pipeline simulation. In: Proceedings of the 24th
NZ geothermal workshop, New Zealand, 2002. p. 2116.
[10] Sanchez-Upton P. The well simulator SIMU89. Geotermia 1990;6:14154.
[11] Karaalioglu H, Watson A. A comparison of two wellbore simulators using eld measurements. In: Proceedings of the 21st NZ
geothermal workshop, New Zealand, 1999. p. 21122.
[12] Kleinstreuer C. Two-phase ow. Theory and applications. New York-London: Taylor & Francis Group; 2003.
[13] Garg SK, Pritchett JW, Alexander JH. A new liquid hold-up correlation for geothermal wells. Geothermics 2005;33:795817.
[14] Michaelides EE, Shafaie FF. A numerical study of geothermal well ow with salt and non-condensibles present. J Energy Technol
1986;108:1405.
[15] Battistelli A, Calore C, Pruess K. The simulator TOUGH2/EWASG for modeling geothermal reservoirs with brines and noncondensible gas. Geothermics 1997;26:43764.
[16] Jung DB, Wai KW, Howard WT. Geothermal ow metering errors. Geothermal Resourc Council, Trans 2001;25:235.
[17] Verma SP. Optimisation of the exploration and evaluation of geothermal resources. In: Chandrasekharam D, Bundschuh J, editors.
Geothermal energy resources for developing countries. Rotterdam, The Netherlands: Swets & Zeitlinger B.V., A.A. Balkema
Publishers; 2002. p. 195224.
[18] Ekberg C. Sensitivity analysis and simulation uncertainties in predictive geochemical modelling. Freiberg Online Geosciences,
Germany, 1999. Available from: http://www.geo.tu-freiberg.de/fog/issues.html.
urdova E. Computer simulation of the uncertainty of analytical results. Fresenius J
[19] Mestek O, Suchanek M, Koplk R, Fingerova H, C
Anal Chem 1999;364:2037.
[20] Orkiszewski J. Predicting two-phase pressure drop in vertical pipes. J Petroleum Technol 1967;19:82938.
[21] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and channel ow. Int Chem Eng 1976;16:35968.
[22] Dobson MK, Chato JC. Condensation in smooth horizontal tubes. J Heat Transfer 1998;120(1):193213.
[23] Shah MM. Chart correlation for saturated boiling heat transfer: equations and further study. ASHRAE Trans 1982;88:18596.
[24] Rouhani Z, Axelsson E. Calculation of volume void fraction in the subcooled and quality region. Int J Heat Mass Transfer
1970;13:38393.
[25] Beattie DRH. A note on the calculation of two-phase pressure losses. Nucl Eng Des 1973;25:395402.
[26] James R. Steam-water critical ow through pipes. Proc Inst Mech Eng 1962;176:7418.
[27] NIST/ASME Steam v2.2, Formulation for General and Scientic Use, NIST Standard Reference Database Number 10, 1996.
[28] Santoyo E. Transient numerical simulation of heat transfer processes during drilling of geothermal wells. PhD thesis, University of
Salford, UK, 1997.
[29] Garca-Valladares O, Perez-Segarra CD, Rigola J. Numerical simulation of double-pipe condensers and evaporators. Int J Refrig
2004;27(6):65670.
[30] Carslaw HS, Jaeger JC. Conduction of heat in solids. 2nd ed. Oxford, UK: Clarendon Press; 1959.
[31] Sanchez-Upton P. Numerical model of the ascending vertical ow of a two-phase uid of three components (H2OCO2NaCl) in
geothermal wells (in Spanish). Doctoral thesis, UNAM, Mexico, in press.
[32] StatSoft, Inc. STATISTICA (data analysis software system), version 6, 2003. Available from: www.statsoft.com.
[33] Sanyal SK, Brown S, Fandriana L, Juprasert S. A sensitivity study of uid ow in a geothermal wellbore. In: Proceedings of the 5th
geothermal reservoir engineering workshop, Stanford University, Stanford, CA, USA, 1979. p. 17.
[34] Cacuci DG. Sensitivity and uncertainty analysis theory. London: Chapman & Hall, CRC; 2003.

ID

762385

Title
Numerical modeling of flow processes inside geothermal wells: An approach for predicting production
characteristics with uncertainties

http://fulltext.study/journal/737

http://FullText.Study

Pages

23

Vous aimerez peut-être aussi