Geothermics 89 (2021) 101930

Contents lists available at ScienceDirect

Geothermics
journal homepage: www.elsevier.com/locate/geothermics

Real-time enthalpy measurement of two-phase geothermal fluid flow using T

load cell sensors: Field testing results
Mohamad Husni Mubaroka,b,*, Sadiq J. Zarrouka, John E. Catera, Ali Mundakirb,
Eko Agung Bramantyob, Yoong Wei Limc
a
Department of Engineering Science, The University of Auckland, Private Bag 92019, Auckland, New Zealand
b
PT Pertamina Geothermal Energy, Jakarta, Cakrawala Building, MH Thamrin 9, Jakarta, Indonesia
c
Contact Energy Ltd., Wairakei Power Station, Private Bag 2001, Taupo, 3352, New Zealand

A R T I C LE I N FO A B S T R A C T

Keywords: Field testing data from five different wells using load cell sensors were analysed to investigate their possible
Measuring enthalpy application for estimating geothermal two-phase fluid enthalpy. Two different data analysis approaches -
Real-time measuring the liquid phase velocity and estimating the dryness fraction - were examined. The liquid velocity was
Two-phase predicted successfully using the cross-correlation technique. However, liquid velocity cannot be used alone to
Geothermal fluid
measure the fluid enthalpy without slip ratio and void fraction. It was only possible to calculate the dryness
Load cell sensor
fraction estimate from the weight of the pipeline for one data set due to measurement noise and the results were
Field testing
found to be of low quality. The remaining technical challenges and recommendations are discussed.

1. Introduction pressure methods). Pressure differential flow meters are the most in-
vestigated alternative to two-phase mass flow rate measurement and
Most conventional geothermal wells produce a two-phase mixture the sharp-edge orifice plate has been widely used since the 1980s (Lin,
of steam and water at the wellhead (Grant and Bixley, 2011; Zarrouk 1982), but with variable reported accuracy (Helbig and Zarrouk, 2012;
and McLean, 2019). Measurement of two-phase mass flow rate and Mubarok et al., 2020; Zarrouk and McLean, 2019). Zarrouk et al. (2019)
enthalpy from each production well after drilling is used to calculate presented field testing results from using a modified sharp-edge orifice
total available power (DiPippo, 2007; Zarrouk and McLean, 2019). It is plate for the accurate measurement (prediction) of both mass flow rate
also crucial for day-to-day field management and, at the same time, it is and enthalpy in two geothermal wells.
the legal obligation of the field developers to satisfy consented fluid Other techniques that have been considered for measuring the
take conditions. Monitoring of these parameters is also important for flowing geothermal fluid enthalpy, including radio frequency (Sisler,
the early detection of technical problems in the wells, which allows for 2018; Sisler and Zarrouk, 2015), micro lip pressure (James, 1987;
early planning and execution of appropriate intervention (Mubarok and Sulaiman and Freeston, 1986), resistive enthalpy meters (Spielman,
Zarrouk, 2016). Enthalpy and mass flow rate of a geothermal well 2003), and compression load cells (Sisler, 2018; Sisler et al., 2017,
change over time during operation (Watson, 2013). A continuous 2016) have been investigated by several researchers (Zarrouk and
measurement will capture the enthalpy and mass flow rate changes and McLean, 2019).
help the geothermal fields operators optimise their fields output and
detect potential problems in wells (Spielman, 2003). 1.1. Radio frequency
There are several existing two-phase flow measurement techniques
in geothermal applications - the total flow calorimeter, lip pressure A radio frequency (RF) experiment to measure the void fraction of
method, separator method, and tracer dilution (tracer flow test) method the two-phase fluid flow was undertaken by Sisler et al. (2015, 2018).
(Helbig and Zarrouk, 2012; Mubarok et al., 2019; Zarrouk and McLean, The transmitter sends an RF wave to the receiver through a fluid in the
2019). However, all the methods are either high in cost (separator and pipeline at a specific amplitude and frequency (Fig. 1). The attenuation
tracer dilution), not in real-time (tracer dilution, calorimeter) or require of the signal (amplitude and frequency) is then converted to the local
the geothermal well to be out of production (calorimeter and lip void fraction which can possible be used to calculate the dryness

⁎
Corresponding author at: Department of Engineering Science, The University of Auckland, Private Bag 92019, Auckland, New Zealand.
E-mail addresses: [email protected], [email protected] (M.H. Mubarok).

https://doi.org/10.1016/j.geothermics.2020.101930
Received 12 May 2020; Received in revised form 20 July 2020; Accepted 23 July 2020
0375-6505/ © 2020 Elsevier Ltd. All rights reserved.
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Nomenclature Δw Change of measured liquid weight (kg)

ΔX Distance between sensor 1 and sensor 2 (m)
A Cross sectional area (m2) ρ Density (kg/m3)
A′ Signal of load cell sensor (kg) τ Time-delay (s)
A Mean of signal A’ (kg)
ATF Acoustic transfer function Subscripts
B′ Signal of load cell sensor 2 (kg)
B Mean of signal B’ (kg) f Liquid phase
h Enthalpy (kJ/kg) g Steam phase
LMS Least mean square i Initial
ṁ Mass flow rate (kg/s) new New value
MDP Maximum discharge pressure (barg) 0 The pipe is empty
plip Lip pressure (barg) 1 The pipe with the two-phase fluid
RAB Normalised cross-correlation coefficient 2 The pipe full of liquid
S Slip ratio f1 The pipe in two-phase flow condition
ν Velocity (m/s) f2 The pipe in the full liquid condition
w Liquid phase measured weight (kg)
w1 Weight of a one percent liquid change in the measured Superscripts
pipe length (kg)
WHP Wellhead pressure (barg) n Correction factor for the effect of the flow restriction on
x Dryness fraction of the two-phase fluid slip ratio

Greek letters

α Void fraction

Fig. 1. Schematic diagram of radio frequency flow meter (modified from Sisler and Zarrouk, 2015).

fraction inside the pipeline and hence the enthalpy of the two-phase 1.3. Resistive enthalpy meter
fluid. To prevent RF energy reflection to the transmitter, a protection of
RF power needed to be improved. The reflected RF to the transmitter Spielman (2003) laboratory tested a sharp-edge office plate in-
causes a significant challenge in the inaccuracy of this measurement stalled with a water resistivity sensor to measure the two-phase geo-
technique (Sisler and Zarrouk, 2015). At the same time measuring the thermal fluid enthalpy in real-time. The device was called the ‘resistive
void fraction is not enough to calculate the dryness fraction because the enthalpy meter’, and it was installed on a flow line of the well test
different phases (liquid and steam) travel at different velocities. facility (Fig. 3). The enthalpy is quantified from the liquid volume
fraction, which is converted from the measured resistivity. However,
this technique required further improvement in equipment design, and
1.2. Micro lip pressure more experiments are required to reduce the error, through the ad-
justment of the correlation factors and liquid hold-up corrections.
Sulaiman and Freeston (1986) and James (1987) tested a micro lip
pressure tube to determine the mixture of steam-water enthalpy 1.4. Load cell
through field experiments in several geothermal wells. A pressure
vessel/pipe of 100 mm diameter was connected to the wellhead or two- A load cell is a real-time sensor that is commonly used for measuring
phase pipeline (Fig. 2). The two-phase fluid from the vessel then dis- stress, using the piezoelectric principle where an electrostatic signal is
charged to the atmosphere through a lip pressure tube with a 10.7 mm obtained by converting a mechanical force from the measured object
throat diameter and the lip pressure is recorded using a pressure gauge. together with a strain gauge (Karaus and Paul, 1992; Money and Sims,
The field test results showed that micro lip pressure was useful in de- 1988; Wiercigroch et al., 1998). To reduce measurement error, me-
termining the two-phase fluid enthalpy; however, further field testing is chanical and electrical calibrations are recommended before the mea-
necessary to ensure this technique can be applied with a wide range of surement. Money and Sims (1988) explained that in general load cell
pipe sizes (James, 1987; Sulaiman and Freeston, 1986). sensors can be divided into two different types - compression and ten-
sion. In the compression type, an object is placed above a platform and

2
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 2. Schematic diagram of micro lip pressure (modified from James, 1987).

this is connected directly to the sensor (Fig. 4a). A tension sensor is An initial value of dryness is needed in Eq. (1). This is measured
installed above the object, and in this case the sensor is used as a sus- using a discharge test performed in the field and can contribute to
pending hanger (Fig. 4b). uncertainty in the result when the initial dryness has relatively large
The implementation of a load cell sensor for measuring the dryness errors. Moreover, obtaining a pipe weight measurement is challenging,
fraction of geothermal two-phase flow has been investigated by Sisler due to thermal expansion that can change the measured weight of the
et al. (2016, 2017, 2018) through field-testing at geothermal well sites. pipeline in hot and cold conditions. The cold condition weight is ne-
Two compression load cells were installed separately under the pipe cessary to know the empty pipe weight as a baseline ( x = 1). The hot
and load cells were attached to the existing sliding pipe supports, as condition weight consists of the pipe and the fluid, either full of liquid
shown in Fig. 5. The measured weight captured by the load cell sensors ( x = 0) or two-phase (0 < x < 1). These tests should be performed in
was assumed to be the combined weight of liquid and the pipeline. the field and will have to be repeated at every location where the load
Sisler et al. (2016) concluded that a 1% change in measured liquid cells are to be deployed.
weight by the load cell was similar to a change in dryness of 1%, ex- In this present study, the enthalpy measurement focusses on using
pressed by: load cell sensors with different installation and data analysis methods
for field-testing at five geothermal wells in Indonesia and New Zealand.
Δw
x new = x i − ,
w1 (1)
2. Field testing
where x new is the new value of dryness, x i is the initial value of dryness,
Δw is the change of measured liquid weight (kg), and w1 is the weight The first field-testing was performed at four geothermal wells in
of a 1% liquid change in the measured pipe length (kg). Indonesia - UBL-K/3 in the Ulubelu field (Lampung province), SBY-3 in

Fig. 3. Schematic diagram of resistive enthalpy meter (modified from Spielman, 2003).

3
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 4. Schematic diagram for installation of (a) compression and (b) tension load cell sensors (modified from Money and Sims, 1988).

the Sibayak field (North Sumatera province), LHD-37/2 in the a data logger. Fig. 8c shows an example of load cells (1 and 2) installed
Lahendong field (North Sulawesi province), and the BDN-A/1 in the in well WK-124 at a distance of 7.57 m apart. It was essential to make
Bukit Daun field (Jambi province). The second set of field-testing was sure the pipeline is seated properly on the load cell cap (Fig. 8b) so that
conducted at well WK-124 in the Wairakei field, New Zealand. The the signals record the entire load/weight of the pipeline. In this work, a
schematic installation of the load cells at the five geothermal wells in minimum 3 mm gap was specified between the bottom of the pipeline
Indonesia and New Zealand are shown in Figs. 6 and 7 respectively. and the existing support (Fig. 9). During installation, the pipeline was
The load cells and their accessories are explicitly designed to ac- lifted using a chain block (crane) or a hydraulic jack to install the
commodate a large diameter two-phase pipeline (≥ 10 inch). A com- support and the load cell sensor below the pipeline.
bination of tension and compression load cell types were selected for During field-testing, the load cell data were validated using a hor-
field testing in Indonesia (Fig. 6), while two compression load cells izontal discharge lip pressure test (Fig. 10d) or separator test (Fig. 10c)
were used for the field testing in New Zealand (Fig. 7). The reason for (Mubarok et al., 2019). In the horizontal discharge test (Fig. 10d), the
using two different load cells in Indonesia was to investigate the ap- two-phase enthalpy and mass flow rate data were recorded from a lip
propriate type (compression or tension types) for geothermal pipeline pressure (Fig. 10d) along with the water level in the weir box (Fig. 10e).
applications. While having the orifice plate in-between (Fig. 6) is in- In the separator test (Fig. 10c), the separated steam (ṁg ) and liquid flow
tended to allow the measurement of the mass flow rate (Mubarok et al., rate (ṁf ) were obtained from the separator discharge (Fig. 10c). The
2019). At the same time to insure that, at the orifice plate, both steam data were recorded continuously at the same frequency as the orifice
and liquid travel at the same velocity. data readings. The digital pressure gauges in Figs. 10a and 10d were
Two sets of load cells were purchased. The load cells had been la- used to measure wellhead pressure (WHP) (Fig. 10a) and plip (Fig. 10d),
boratory tested and calibrated at the manufacturer facility before de- while the water level in the weir box was measured and recorded using
livery to the warehouse (Fig. 8a). The load cells and accessories were a digital infra-red sensor (Fig. 10e).
set up by connecting cables, interface and computer for the no-load Several challenges were encountered during the experiments at
tests in the workshop, as shown in Fig. 8b. The test was performed using some locations, including the pipe falling from its support due to
five measurements set with various known weights. thermal stress/shock (Fig. 11b), a broken load cell platform (Fig. 11b),
The two load cells were installed in the two-phase pipeline of the load cell software that did not work, equipment transportation (do-
geothermal well for field-testing (Fig. 8c). The load cells were placed at mestic and international shipment delays), a broken sensor, and im-
a known distance from each other and the two signals were recorded on proper cable connections. These happened as the load cells could not be

Fig. 5. Schematic of a compression load cell sensor installation in a geothermal two-phase pipeline (adopted from Sisler et al., 2016).

4
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 6. Schematic of load cell installation for field-testing in Indonesia for wells UBL-K/3, SBY-3, LHD-37/2 and BDN-A/1.

Fig. 7. Schematic of load cell installation for field-testing in New Zealand for well WK-124.

adapted in geothermal environments, including high temperature and Geothermal two-phase annular flow is not totally smooth as the flow
corrosive gases which can damage the electronic (Copper and Silver) of the steam is much faster than the flow of water/liquid (Austrheim
parts of the load cell sensor and connecting cables. et al., 2007). The steam flow will create small waves (this is not as
visible as wave flow regime) in the liquid that is flowing at the bottom
3. Measurement approaches of the pipeline, which creates changes in the weight of the pipeline on
the supports.
The basic principle for testing the load cell sensors, is to measure the When choosing the wells for field-testing, slugging wells are avoided
weight change of the pipeline with the flowing two-phase fluid inside because it can highly induced vibration, which is detrimental to the
during well discharge/operation (Sisler, 2018; Sisler et al., 2017, 2016). operation of the load cells in the long term. For some cases slug flow
The magnitude of weight change is expected to be converted to the also causes the pipeline to jump from its support (Beuthe, 1997),which
velocity, dryness or enthalpy of two-phase geothermal fluid. Real-time will make these measurements not very useful and can cause high
discharge data were used to compare and validate the results. Pro- sudden loads beyond the load cell design limit.
duction testing (Fig. 10) was performed at the same time as the load cell It is important to note that the use of two-phase flow regime map to
experiment, to validate the results. Two different approaches were help predict the type flow regime before considering the well is not
performed in this work - a time-delay and liquid phase weight. accurate. This is because, most of these maps have been developed for
water and air using much smaller diameter pipelines, which makes
3.1. Two-phase flow regime them not very accurate in large diameter geothermal pipelines. For
example, The test data from the five wells used in this study are plotted
Considering liquid phase wave is based on the field observation. on the Mandhane map as shown in Fig. 12 (Mandhane et al., 1974).
Even though the well is not slugging (the fluid is of annular flow re- These wells have the same visible behaviour from field observation and
gime), there is always small level vibration which can be felt by hand as demonstrated by the load cell data. Three different flow regimes –
placed on the two-phase pipeline connected to a well under long term annular, dispersed and slug flow – are predicted by the Mandhane map
steady-state operation (Beuthe, 1997). This could also be due to the (Fig. 12). We believe that a dedicated geothermal two-phase flow re-
change in the flow regime from the two-phase geothermal fluid flowing gime map should be developed from field test data.
vertically in the well then change in direction to horizontal flow at the
ground surface (Chadha et al., 1993). These waves maintain their shape
for some distance in the straight pipeline with no obstruction which we 3.2. Liquid velocity through two-phase wave time-delay
have tried capture in this work to calculate the velocity of the liquid/
water flowing at the bottom of the pipeline. However, even if it possible The first approach is to determine the time-delay (τ ) between two
to calculate the velocity of water, we will not be able to measure the signals from two load cell sensors installed at a fixed distance (ΔX ), as
mass flow rate/velocity of steam as it is not possible to estimate/mea- shown in Fig. 13. With a known distance between two load cell sensors
sure the slip ratio. Therefore, this technique will not be suitable for and an estimated signal time-delay, then the liquid velocity can be
measuring the mass flow rate and the work is focused on using one load calculated as:
cell to calculate the enthalpy.

5
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 8. Images of the load cell system during (a) calibration, (b) equipment set up and no-load test at the workshop and (c) the installation in the two-phase pipeline
at well WK-124 (images acquired by author).

νf =
ΔX
, phase fluid can be considered homogeneous at the vena-contracta zone
τ (2) (Fig. 6) of the orifice plate so that n is zero, and S becomes unity (S = 1)
where νf is the liquid velocity (m/s), ΔX is the distance between sensor or very close to unity. The void fraction (α ) is also an important para-
1 and sensor 2 (m), and τ is the time-delay between sensor 1 and sensor meter that needs to be known for characterising the two-phase fluid
2 (s). In this work, it was not possible to measure the steam velocity (νg ) flow, which is calculated from the ratio of steam ( Ag ) and liquid ( Af )
using load cells because the steam is much less dense and travels at a cross-section area (Eq. 5), as shown in Fig. 14.
higher velocity above the liquid phase. Ag
The first field-tests were performed in Indonesian geothermal fields α= .
(Ag + Af ) (5)
using the two different Load Cell types. The tension (Sensor 1) and
compression (Sensor 2) sensors were installed upstream and down- Campos et al. (2014) explained that the two parameters, void
stream of a two-phase orifice plate respectively (Fig. 6). The placement fraction (α ) and slip ratio (S ), need to be known to determine the
of the two sensors between the orifice plate is intended to ensure that dryness fraction ( x ):
the liquid and steam velocity slip ratio (S ) is close to unity (Eq. 3) near −1
the orifice plate. ⎡ 1 − x ⎛ ρg ⎞ ⎤
α = ⎢1 + ⎛ ⎞ S⎥ .
νg ⎝ x ⎠ ⎜⎝ ρf ⎟⎠ (6)
S= ≡1 ⎣ ⎦
νf (3)
Abdul-Majeed (1996) proposed a void fraction model without the
Another expression for S comes from the combination of the dryness fraction. This void fraction is a function of liquid and steam
Chisholm (1983) and Kojasoy et al. (1997) correlation: velocity. In the present study, the liquid velocity can be estimated using
n load cell method. However, the slip ratio is still necessary to estimate
⎡ ⎛ ρf ⎞⎤ the steam velocity (Eq. 3).
S = ⎢ 1 + x ⎜ − 1⎟ ⎥ ,
⎢ ρ In the second field tests at well WK-124, two compression load cells
⎣ ⎝ g ⎠⎥⎦ (4)
were placed under the pipeline between existing sliding supports in a
3
where x is the dryness, ρf is the liquid density (kg/m ), ρg is the steam straight two-phase pipeline, as shown in Fig. 7. The liquid and pipe
density (kg/m3), and n is the correction factor for the effect of the flow weight were measured by the two sensors to estimate the time-delay
restriction on S . Roul and Dash (2012) noted that the mixture of two- and calculate the liquid velocity. In this case, the two-phase fluid is

6
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

flowing without the restriction of an orifice plate (Fig. 7).

The pipeline weights from the tension (Fig. 15a) and compression
(Fig. 15b) load cells are received at an interface through a fibre optic
cable. The electrical signal from the sensors is then converted to the
weight in the interface and a computer records signals in real-time, as
shown in Fig. 15c.
Sun et al. (2016) showed that the time-delay between two sensors
can be calculated using cross-correlation, least mean squares (LMS),
acoustic transfer function (ATF), or time-frequency energy methods. In
the present study, the cross-correlation method was used because it is
the most accurate to estimate the time-delay (Shi et al., 2015). The
time-delay is estimated by detecting the lag for the highest peaks
(caused by the passage of liquid water waves inside the pipeline) of the
cross-correlation coefficient (RAB ) from the data.
The cross-correlation function between two sensor signals - A′ (from
sensor 1) and B′ (from sensor 2) - is expressed as (Nelson-Wong et al.,
2009):
1 N
∑ (Ai
N i=1
− A )(Bi − B )
RAB = ,
1 N N
N
∑i = 1 (Ai − A )2∑i = 1 (Bi − B )2 (7)

where RAB is the normalised cross-correlation coefficient (between

0–1), N is the number of data points in the record of signal input, Ai is
signal A′ (kg), A is the mean of signal A′ (kg), Bi is signal B′ (kg), and B
is the mean of signal B′ (kg). In the present work, the cross-correlation
was carried out using the built-in function (CORREL) in Microsoft®
Excel®. The relationship criteria based on the value of RAB introduced
by Cohen (1988) has been adopted, as shown in Table 1.

3.3. Liquid phase weight

Fig. 9. Image of compression load cell showing the gap (more than 3 mm)
between the bottom of the two-phase pipeline shoe and the existing support A second approach to estimate the enthalpy is estimating the dry-
(image taken by author).
ness ( x ) based on the liquid phase measured weight of the pipe in empty

Fig. 10. Photographs of (a) a geothermal well, (b) a silencer, (c) a separator, (d) a lip pressure recorder and (e) weir box with digital water level recorder (images
taken by author).

7
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 11. Photographs of pipe-load cell support conditions, (a) before and (b) after well discharge of UBL-K/3 (images taken by author).

Fig. 12. Mandhane map with data from field-testing results.

Fig. 13. Concept of time-delay (τ ) estimation from the profile of

measured pipe and fluid weight from two load cell sensors.

8
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Table 1
Cross-correlation. relationship criteria (Cohen, 1988).
RAB Criteria

< 0.3 Weak relationship

0.3 – 0.5 Moderate relationship
0.5 – 1.0 Strong relationship

(w0 ), two-phase fluid (w1) and full of liquid (w2 ) conditions, as shown in
Fig. 16. In this method, the measured weight recorded by the load cell is
assumed to be the liquid phase weight while steam is neglected. The
pipe in the cold empty condition is measured where the fluid is not
flowing. The measured weight (w0) is used as a baseline. If the empty
pipe is not adjusted to zero, the liquid phase (w1 or w2 ) is effectively the
measured weight minus the pipeline weight (w0 ). The correlation be-
Fig. 14. Schematic of a two-phase pipeline with the cross-section areas of steam tween three parameters (w0 , w1 and w2 ) and x can be expressed as:
and liquid.
wf 1 = (w1 − w0), (8)

wf 2 = (w2 − w0), (9)

Fig. 15. Photographs of (a) tension, (b) compression load cells and (c) the real-time data visualisation of the weight measured pipe during data acquisition (images
taken by author).

9
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 16. Schematic of liquid phase to estimate the dryness by measuring three conditions of weight: (a) empty (only air), (b) two-phase fluid and (c) full of liquid.

Fig. 17. Simplified of schematic testing facility diagram and the operation process to get the full liquid condition during the field-testing at well UBL-K/3.

wg = (wf 2 − wf 1), (10) practically is not easy to achieve the full liquid condition using fluid
from a two-phase flowing wells and cold water is safer to handle.
wg (wf 2 − wf 1 ) When the well is closed, the wellhead pressure (WHP) will increase
x= = , to the maximum discharge pressure (MDP) (Zarrouk and McLean,
wf 2 (w2 − w0 ) (11)
2019). The pressure beyond MDP will drop gradually to zero, and in
where x is the dryness fraction, wf 1 is the liquid phase weight in the this case the fluid will stop discharging from the well and the well may
two-phase flow condition (kg), wf 2 is the liquid phase weight in the full require discharge stimulation (Mubarok and Zarrouk, 2017).
liquid condition (kg), and wg is the steam weight (kg). Filling the pipeline with cold water was practically difficult to carry
Theoretically, the full liquid condition in a pipeline can be obtained out on remote sites, moreover cold water is not exactly representative
by filling the pipe with cold water or by closing the throttle valve in the of the geothermal fluid due to the difference in density. At the same
downstream location to 100 %, as shown in Fig. 17. When the valve is time cold water will not allow the pipeline to heat up (thermally ex-
fully closed, the line pressure will increase, and the steam phase will pand) as in normal flow conditions. Therefore, we attempted the ap-
change to liquid ( x = 0) in the mixture of two-phase flow. However, proach given in Fig. 17, which is not ideal, because filling the pipeline

10
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 18. Load cell data from field-testing at well UBL-K/3.

Fig. 19. Load cell data from field-testing at well SBY-3.

Fig. 20. Load cell data from field-testing at well LHD-37/2.

with hot geothermal fluid can cause the brine to drain back into the 1 could not be calculated.
well leaving the pipeline full of steam, which can be a weakness in the The load cell signals for each well show different trends
calibration method. Possibly adding the equivalent weight (of pipe full (Figs. 18–22), due to various factors including pipeline arrangement,
of liquid) to the outside of the pipeline would be a better option for pipe size, flow regime, operating pressure, and load cell installation
future testing. (Table 2). In Figs. 18–20 and Fig. 22, the cross-correlation coefficient
(RAB ) is estimated using Eq. (7). The summary of the cross-correlation
4. Results delay for wells UBL-K/3, SBY-3, LHD-37/2 and WK-124 are shown in
Figs. 23–26 respectively.
The results of the real-time signal 1 and 2 data acquired using the The entire original signals were cross-correlated in Figs. 23–26 to
load cells are shown in Figs. 18–22 respectively. Two load cells were find the relationship between the two signals. The cross-correlation
installed with different separation distances for each well, as summar- coefficient RAB is presented in Table 3. From Figs. 23–26, it is apparent
ized in Table 2. From Fig. 21, ‘signal 1′ at well BDN-A/1 was not re- that only WK-124 test data has a peak RAB value greater than 0.5. The
corded, as the measured pipeline weight was over the maximum ca- other wells (UBL-K/3, SBY-3 and LHD-37/2) have a poor correlation
pacity of the sensor (> 5000 kg). Thus, the time-delay for well BDN-A/ due to the noise that resulted from having the orifice plate between the

11
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 21. Load cell data from field-testing at well BDN-A/1.

Fig. 22. Load cell data results from the field-testing at well WK-124.

Table 2 two sensors, causing a major disturbance to the flow or having a low
Load Cells Installed. data frequency setting (1 Hz), as shown in Table 3. The peak RAB values
Well Load Cell Type Orifice Plate Distance between
in the overall cross-correlation of the wells are summarised in Table 3.
between Load Load Cells Table 3 shows that; the time-delay of the two load cell signals can be
Sensor 1 Sensor 2 Cells (m) estimated for well WK-124, which is indicated by the strong relation-
ship (Peak RAB = 0.99) for the correlation. Unfortunately, the other
UBL-K/3 Tension Compression Installed 7.00
SBY-3 Tension Compression Installed 4.94
three wells (UBL-K/3, SBY-3 and LHD-37/2) do not allow the time-
LHD-37/2 Tension Compression Installed 15.58 delay to be calculated using cross-correlation, because the three have a
BDN-A/1 Tension Compression Installed 16.00 weak correlation relationship (Peak RAB < 0.3).
WK-124 Compression Compression None 7.57 From the load cell field-testing in five wells, the test for liquid phase
weight was only measured in well UBL-K/3 due to technical problems
and test time limitations on the other wells. An average measured liquid

Fig. 23. Cross-correlation of signal 1 and signal 2 from well UBL-K/3.

12
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 24. Cross-correlation of signal 1 and signal 2 from well SBY-3.

Fig. 25. Cross-correlation of signal 1 and signal 2 from well LHD-37/2.

Fig. 26. Cross-correlation of signal 1 and signal 2 from well WK-124.

Table 3 weight is 813 kg. In this case, the average value is defined as wf 2, be-
Cross-correlation Peak Value from Wells UBL-K/3, SBY-3, LHD-37/2 and WK- cause the fluid in the pipeline is assumed to be full of liquid. The pi-
124. peline weight in the empty condition is reset to zero, so the w0 = 0 kg. A
Well Data Frequency Peak RAB Correlation Criteria real-time profile for wf 1 is provided in Fig. 27. The data shown in Fig. 27
(Hz) will be used in Eq. (11) to calculate the dryness fraction which is used to
calculate the enthalpy at the pipeline pressure.
UBL-K/3 1 0.292 < 0.3, weak relationship
From the observation of the overall cross-correlation profile of well
SBY-3 1 0.298 < 0.3, weak relationship
LHD-37/2 1 0.268 < 0.3, weak relationship WK-124 in Fig. 28, the profile was partitioned into shorter samples, in
WK-124 50 0.990 0.5−1.0, strong relationship this case every 60 s (sec) of measurement to gain an averaged τ in every
minute of the total measurement. Fig. 28 shows an example for ob-
taining the τ in the first minute of measurement from the original

13
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 27. The measured weight profile for wf 1 in a real-time mea-

surement at well UBL-K/3.

Fig. 28. (a) Real-time load cell data and (b) cross-correlation in the first minute of measurement from well WK-124, Wairakei, New Zealand.

Fig. 29. Real-time profiles for: time-delay, peak RAB and liquid
velocity of well WK-124.

signals (Fig. 28a) and the cross-correlation (Fig. 28b). From Fig. 28b, phase flow. However, after 23 min of measurement, the velocity de-
the peak RAB value occurs at τ = 0.16 s. By using the information of τ clines rapidly and becomes stable around 4 m/s.
and distance between load cell sensors (ΔX = 7.57 m), the velocity The high liquid velocity in the early 23 min measurement could be
(ν = ΔX / τ ) will be 47.31 m/s. related to well discharge transient effects and pipeline thermal expan-
sion. The transient effects usually occur in the beginning of the well
discharge due to the two-phase fluid flow from the feed zone in the well
5. Discussion
bore (Grant and Bixley, 2011). Pipeline expansion takes place because
of the heat transfer (convective and conductive) from the geothermal
Fig. 29 shows the peak RAB , τ and v in real-time measurement. From
two-phase fluid. However, the pipe weight remains steady after the
Fig. 29, the velocity (solid green line) over the 23 min of measurement
transient region. This phenomenon is shown in the load cells data
decreases from 43 to 8 m/s. These velocities are very high for liquid

14
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

Fig. 30. The real-time estimated velocities from the load cells with comparison to field data (liquid velocity, steam velocity, two-phase velocity and slip ratio).

Fig. 31. The real-time enthalpy measurements, comparison between load cell and actual field data for UBL-K/3.

including the radio frequency sensor (Sisler and Zarrouk, 2015).

However, those option was not investigated in this work.
Therefore, on its own, the load cell method cannot provide accurate
measurement of two-phase mass flow rate or enthalpy. At the same
time, the technical challenges encountered during field-testing show
that more work is needed to improve the reliability of the sensor
system.
From the liquid phase weight at well UBL-K/3, the real-time profile
comparison between load cell and field data is shown in Fig. 31. The
trend of estimated h (red solid line) is higher than the field data (black
dots). The prediction of h using liquid phase weight over-predicts the
measured h from field data by 13.7 %, as shown in Fig. 32.
The high relative error of the enthalpy prediction results (Figs. 31
and 32) could be caused by two factors. Firstly, the measurement of the
empty pipe is carried out in cold condition, while w1 and w2 are mea-
sured in the hot condition, so the thermal stress of the pipe due to heat
transfer will contribute to the measured pipe weight. Secondly, the
measured wf 1 is assumed to be liquid weight only, but the steam, which
Fig. 32. The predicted and field data of enthalpy at UBL-K/3 using load cell-
has some weight (approximately 1–5 % of the total fluid weight), is
liquid phase weight. neglected.
Some technical problems were faced during the field-testing due to
the complexity of the equipment setup and remote site operation.
(Fig. 22) and the estimated velocity (Fig. 30).
Future work is necessary to enhance the accuracy, simplify the in-
From Fig. 30, the estimated velocity (load cell) after 23 min follows
stallation and enhance the repeatability of the load cell method. It is
the liquid velocity (field data) trend with an average slip ratio of 10.6
recommended that future field-testing consider including a longer
(S = νg / νf ). This result confirmed that the load cell method is capable of
straight pipe length (> 20 m) and a longer testing duration (at least one
estimating the liquid velocity by measuring the time-delay of two sig-
day), avoiding the installation between orifice plates, valves or other
nals from load cell sensors. However, the load cell method requires a
pipeline components and use a minimum frequency of 50 Hz for data
slip ratio (S ) and a void fraction (α ) to calculate the dryness ( x ) of two-
acquisition. Furthermore, selecting the appropriate maximum load cell
phase fluid (Eq. 6), which, can possibly be measured by other devices,

15
M.H. Mubarok, et al. Geothermics 89 (2021) 101930

capacity is also important to accommodate the heavy pipe system in the Acknowledgements
field to prevent malfunctions or sensor damage.
The authors would like to thank New Zealand ASEAN Scholarship
and PT Pertamina Geothermal Energy for their financial and technical
6. Conclusions
supports. We would also like to thank Contact Energy Ltd for facil-
itating the field work at Wairakei geothermal field.
A pair of load cells were installed in geothermal two-phase pipelines
with two installation setups during field-testing. First, two different
References
load cell types were placed at a particular distance either side of an
orifice plate; these were tension and compression sensors. The second
Abdul-Majeed, G.H., 1996. Liquid holdup in horizontal two-phase gas-liquid flow. J. Pet. Sci.
installation scheme was carried out using two identical load cells Eng. 15, 271–280.
(compression) with no orifice plate in-between. Austrheim, T., Gjertsen, L.H., Hoffmann, A.C., 2007. Re-entrainment correlations for demisting
cyclones acting at elevated pressure on a range of fluids. Energy Fuels 21, 2969–2976.
The cross-correlation results show that the data from the two sen- Beuthe, T.G., 1997. Review of two-phase water hammer. In: Proceedings of the 18th Canadian
sors in the first scheme are not correlated, thus this scheme cannot Nuclear Society Conference. Toronto, Canada. pp. 20.
Campos, S.R.V., Baliño, J.L., Slobodcicov, I., Filho, D.F., Paz, E.F., 2014. Orifice plate meter
estimate the time delay of the load cell signals. In contrast, the cross- field performance: formulation and validation in multiphase flow conditions. Exp. Therm.
correlation results from the second scheme (two compression load cells Fluid Sci. 58, 93–104.
with no restriction between them) have a better cross-correlation re- Chadha, P.K., Malin, M.R., Palacio-Perez, A., 1993. Modelling of two-phase flow inside geo-
thermal wells. Appl. Math. Model. 17, 236–245.
lationship (RAB ) between the two sensors signals. Chisholm, D., 1983. Two-phase Flow in Pipelines and Heat Exchangers, 1st ed. George Godwin,
From the field-testing results, estimating enthalpy using load cell Michigan.
Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Lawrence
sensors is difficult to achieve. Measuring the liquid velocity is possible Erlbaum.
with two parallel load cells; however, the input parameters of slip ratio DiPippo, R., 2007. Geothermal power plants: principles, applications. Case Studies and
Environmental Impact, 2nd ed. Butterworth-Heinemann, USA.
and void fraction from other measurement/devices is necessary to find Grant, M.A., Bixley, P.F., 2011. Geothermal Reservoir Engineering, 2nd ed. Elsevier, United
the enthalpy. Estimating the dryness fraction to obtain enthalpy is Kingdom.
possible, but the result shows low accuracy (the mean relative error is Helbig, S., Zarrouk, S.J., 2012. Measuring two-phase flow in geothermal pipelines using sharp
edge orifice plates. Geothermics 44, 52–64.
13.7 %). James, R., 1987. Discharging through an orifice determines steam-water enthalpy. Proceedings
Further field testing of load cells should be carried out to optimise of the 12th Workshop on Geothermal Rerservoir Engineering. Stanford University,
California, USA, pp. 301–304.
the equipment setting to improve the enthalpy measurement technique. Karaus, A., Paul, H., 1992. Load cells with small nominal load based on strain gauges using thin-
It would be also interesting to combine the load cell with other mea- film techniques. Measurement 10, 133–139.
Kojasoy, G., Landis, F., Kwame-Mensah, P., Chang, C.T., 1997. Two-phase pressure drop in
surement techniques, such as the radio frequency, micro lip pressure, or
multiple thick-and thin-orifice plates. Exp. Therm. Fluid Sci. 15, 347–358.
resistive enthalpy meter. Lin, Z.H., 1982. Two-phase flow measurements with sharp-edged orifices. Int. J. Multiph. Flow
All persons who meet authorship criteria are listed as authors, and 8, 681–693.
Mandhane, J.M., Gregory, G.A., Aziz, K., 1974. A flow pattern map for gas-liquid flow in
all authors certify that they have participated sufficiently in the work to horizontal pipes. Int. J. Multiph. Flow 1, 537–553.
take public responsibility for the content, including participation in the Money, M.W., Sims, G.D., 1988. Calibration of quartz load cells; An in-situ procedure for in-
strumented falling weight impact machines. Polym. Test. 8, 429–442.
concept, design, analysis, writing, or revision of the manuscript. Mubarok, M.H., Zarrouk, S.J., Cater, J.E., 2019. Two-phase flow measurement of geothermal
Furthermore, each author certifies that this material or similar material fluid using orifice plate: field testing and CFD validation. Renew. Energy 134, 927–946.
has not been and will not be submitted to or published in any other Mubarok, M.H., Cater, J.E., Zarrouk, S.J., 2020. Comparative CFD modelling of pressure dif-
ferential flow meters for measuring two-phase geothermal fluid flow. Geothermics 86,
publication before its appearance in the Geothermics. 1–18.
Mubarok, M.H., Zarrouk, S.J., 2016. Steam-field design overview of the ulubelu geothermal
project, Indonesia. Proceedings of the 38th New Zealand Geothermal Workshop.
Authorship contributions Mubarok, M.H., Zarrouk, S.J., 2017. Discharge stimulation of geothermal wells: overview and
analysis. Geothermics 70, 17–37.
Nelson-Wong, E., Howarth, S., Winter, D.A., Callacghan, Jack, P., 2009. Application of
Please indicate the specific contributions made by each author (list Autocorrelation and Crosscorrelation Analyses in Human Movement and Rehabilitation
Research. ournal Orthop. N. Am. J. Sports Phys. Ther. 39, 287–295.
the authors’ initials followed by their surnames, e.g., Y.L. Cheung). The Roul, M.K., Dash, S.K., 2012. Single phase and two-phase flow through thin and thick orifices in
name of each author must appear at least once in each of the three horizontal pipes. J. Fluids Eng. 134, 1–14.
Shi, B., Wei, J., Pang, M., 2015. A modified cross-correlation algorithm for PIV image proces-
categories below. sing of particle-fluid two-phase flow. Flow Meas. Instrum. 45, 105–117.
Category 1 Sisler, J.R., 2018. Real-time sensors for multi-phase flows: results of Field trials and further
Conception and design of study: M.H. Mubarok, S.J. Zarrouk, J.E. study. In: Proceedings of the 43rd Workshop on Geothermal Reservoir Engineering.
Stanford University. pp. 1–4.
Cater; acquisition of data: M.H. Mubarok; analysis and/or interpreta- Sisler, J.R., Zarrouk, S.J., Urgel, A., Lim, Y.W., Adams, R., Martin, S., 2016. Measurement of two
tion of data: M.H. Mubarok, S.J. Zarrouk, J.E. Cater phase flows in geothermal pipelines using load-cells: Field trial results. In: Proceedings of
the 41st Workshop on Geothermal Reservoir Engineering. Stanford University, California,
Category 2 USA. pp. 1–10.
Drafting the manuscript: M.H. Mubarok, J.E. Cater, S.J. Zarrouk; Sisler, J.R., Boetzer, K.D., Zarrouk, S.J., 2017. Weight-based Phase Composition Ratio
Determination. US, pp. 1–16.
revising the manuscript critically for important intellectual content:
Sisler, J.R., Zarrouk, S.J., 2015. Measurement of two phase flows in geothermal pipelines using
M.H. Mubarok, S.J. Zarrouk, J.E. Cater radio frequency (RF) power measurements: experimental results. In: Proceedings of the
Category 3 37th New Zealand Geothermal Worksop. International Geothermal Association, Taupo,
New Zealand. pp. 1–7.
Approval of the version of the manuscript to be published (the Spielman, P., 2003. Continuous enthalpy measurement of two-phase flow from a geothermal
names of all authors must be listed): M.H. Mubarok, S.J. Zarrouk, J.E. well. In: Geothermal Resources Council. California, USA. pp. 413–416.
Sulaiman, S., Freeston, D.H., 1986. Geothermal Two-Phase Field and Laboratory Measurements.
Cater, A. Mundakir, E.A.B. Bramantyo, Y.W. Lim Mech. Eng. The University of Auckland, Auckland, New Zealand.
Sun, H.M., Jia, R.S., Du, Q.Q., Fu, Y., 2016. Cross-correlation analysis and time delay estimation
of a homologous micro-seismic signal based on the Hilbert-Huang transform. Comput.
Declaration of Competing Interest Geosci. 91, 98–104.
Watson, A., 2013. Geothermal Engineering: Fundamental and Applications, 1st ed. Springer,
New York, USA.
The New Zealand ASEAN Scholarship (NZAS) and PT Pertamina Wiercigroch, M., Neilson, R.D., Player, M.A., 1998. Load cell for dynamic measurement of
cutting forces generated during ultrasonic drilling and milling of hard materials.
Geothermal Energy (PGE) are providing financial support and field- Proceedings of the Instrumentation Mechanical Engineers 263–270.
testing to our research but had no involvement in the conduct of the Zarrouk, S.J., McLean, K., 2019. Geothermal Well Test Analysis : Fundamentals, Applications
and Advanced Techniques, 1st ed. ed. Elsevier, London, UK.
research or preparation of the article and there is no restriction is im- Zarrouk, S.J., Mubarok, M.H., Cater, J.E., 2019. The geothermal Two-phase orifice. ULUSAL
posed on publishing the work, thus, there are no interests to declare. TESISAT MHENDISLIGI KONGRESI. Izmir, Turkey.

16