You are on page 1of 27

SPE-176780-MS

Converting ESP Real-Time Data to Flow Rate and Reservoir Information for
a Remote Oil Well
L. Camilleri, M. El Gindy, A. Rusakov, and S. Adoghe, Schlumberger

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Middle East Intelligent Oil & Gas Conference & Exhibition held in Abu Dhabi, UAE, 15–16 September 2015.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Flow-rate testing is the most fundamental and essential form of reservoir surveillance and is typically
performed on a monthly basis using a test separator. This case study reviews the results from modelling
a well located in a jungle for which poor testing quality precluded inflow analysis. The main causes were
logistical difficulties, large flow transients, and phase segregation in the flowline between the wellhead
and test separator.
A virtual flowmetering technique was investigated to determine a trend of reservoir pressure and skin
and identify production optimisation opportunities. The algorithms used real-time data from permanent
downhole gauges and electric submersible pump (ESP) surface controllers, which provided the necessary
measurement frequency, resolution, and repeatability to capture well performance transients and provide
a full production history. The liquid rate calculation used the principle that the power absorbed by the
pump is equal to that generated by the motor. Water cut was calculated by modelling the production tubing
pressure drop with a multiphase correlation.
Virtual flowmetering is demonstrated to provide a credible alternative where test separators and
multiphase meters cannot be deployed owing to logistical difficulties such as in swampy terrain or on
unmanned platforms and subsea. Furthermore, the high frequency, resolution, and repeatability of this
particular flowmetering technique are greater than those obtained with traditional monthly, or even
weekly, test separator data. This case study illustrates how these three attributes enhance nodal analysis,
pressure transient analysis (PTA), and superposition, which enabled monitoring the well’s reservoir
pressure and well interference without the need to conduct pressure buildups.
Additional value comes from extending the application of this virtual flowmetering technique to wells
that do not have testing difficulties, especially those that are under producing and require detailed
diagnostics (e.g., PTA). The technique is also well suited to application in true real time because it is
analytical, valid across the full pump range, and able to capture transients without requiring an iterative
process. Is real-time flow rate just around the corner?
Introduction
Traditional production testing provides periodic measurements of oil, water, and gas rates. In most cases,
these are obtained on a monthly basis using a test separator. These data, combined with downhole flowing
2 SPE-176780-MS

pressure measurements, are essential to monitoring both outflow and inflow performance of a well, which
is the reason that this type of surveillance is maintained by most operators even in the most marginal
oilfields where other types of surveillance are no longer conducted for economic reasons, notwithstanding
the fact that it is often a legal requirement in some countries. As will be shown further on in this paper,
the production historical trend allows the practicing petroleum engineer to monitor the performance of the
lift mechanism, as well as measure the well productivity and monitor reservoir depletion. Therefore, when
the production data are either inaccurate and/or of insufficient frequency to provide a meaningful trend,
the production and/or reservoir engineer can be at a loss as to how to optimise both the method of artificial
lift and the well production. In such cases, the only solution is to resort to historical flow-rate recon-
struction utilizing a virtual flowmeter, which calculates the liquid rate and water cut using real-time data,
if available. The following discussion is the case study of such an oil well equipped with an ESP and
real-time data and located in the jungle in West Africa, where logistical challenges precluded obtaining
meaningful production test trends over a 16-month production period. The analysis demonstrates how a
virtual flowmeter can provide the necessary data to both characterize the well’s inflow as well as optimise
the ESP design and operation.
The completion for the well which is the subject of this case study is shown in Fig. 1. The key points
to note for the subsequent analysis are
SPE-176780-MS 3

Figure 1—Completion diagram complete with key depths


4 SPE-176780-MS

● Two layers are open to production and commingled through the ESP using the bypass tubing.
● Both layers are in the same sandstone reservoir with similar fluid properties.
● Both sandface completions include gravel packs.
● Although a production packer has been included in the completion, it is shallow set, and the
annulus includes a gas vent valve; therefore wellbore storage is not eliminated.
Problem Statement
The problem statement is best explained by plotting concurrently the downhole permanent gauge data
with the test separator data, as shown Fig. 2. One can readily see that there are insufficient points to obtain
a trend because

Figure 2—During the 16 month production period, there are only 10 physical test separator measurements (top), of which several are
invalid, and it is difficult to establish a trend.

● The production period from date of electric submersible pump (ESP) installation to failure spans
16 months; however, there are only 10 tests, and, in actual fact, during the first 3 months and last
4 months, no tests were performed.
● One test point should be discarded as it was performed while the pump was actually turned off,
which is why the flow-rate measurement is so low. It is highlighted with a red circle in Fig. 2.
● A further two wells tests (shown in orange circles in Fig. 2), were performed while the well was
being drawn down so the well was not in pseudosteady-state condition (i.e., was in transient
condition), which is why the flow rate was higher than expected on one of these points.
● Even after removal of the above three points (see Fig. 3), there is s till a lack of repeatability in
the test points. One flow-rate measurement of 5,500 BPD is most probably incorrect because, at
SPE-176780-MS 5

this flow rate, the pump curve (see Fig. 4) produces an insignificant amount of head, which is
insufficient to lift the well and is therefore not a valid ESP operating point.

Figure 3—Trend of seven remaining test points after elimination of invalid tests.
6 SPE-176780-MS

Figure 4 —Pump test curve for 108 stage SN36000, showing best fit though factory test curves at five flow rates at 60 Hz, which is used
in liquid calculation algorithm to remove error caused by difference between catalogue and actual installed pump curves. There is an
additional point as efficiency is zero at zero flow rate irrespective of the pump condition.

● There is a lack of repeatability in the water cut testing, and it is difficult to identify a trend because
the water cut measurements both increase and decrease with time. There was potentially an
additional problem associated with the test measurements due to the test separator being located
more than 3 km from the wellhead and connected with a 4-in. inside diameter (ID) flowline. Such
an arrangement creates phase segregation, which can distort water cut measurements, however
water cut was monitored with fluid samples taken at the wellhead and therefore not effected by
phase segregation. In actual fact, the test separator was operated in two-phase mode.
Fig. 3 demonstrates that the key problem is a lack of test data with which to provide a reliable trend
to estimate depletion, changing productivity index (PI) and/or skin, and recoverable incremental reserves
and to monitor ESP performance. Furthermore, the lack of repeatability sheds doubt on the accuracy of
the tests.
The test data required to monitor this well performance are simply the liquid rate and water cut; gas/oil
ratio (GOR) is not critical because there are good reasons to believe that this is stable and equal to the
original solution GOR (Rs) for the following reasons:
● Observation of the openhole log suggests that there is no gas cap communicating with the two sets
of perforations because, although the neutron-density log does show a gas-bearing zone, this is
more than 150 m above the top oil-bearing zone and separated from it by three shale layers.
● A secondary gas cap is not expected because both reservoir and production pressures are higher
than the bubblepoint, which is approximately 1,500 psia. The reservoir pressure at the time of ESP
installation was 3,350 psia and therefore substantially greater than the bubblepoint even after any
SPE-176780-MS 7

depletion. One can see from Fig. 2 that the lowest flowing pressure at pump intake depth was 1,200
psia. This therefore represents a pressure greater than 1,300 psia at perforation depth for the upper
layer and approximately 1,550 psi for the lower layer, which, combined with a water cut of
approximately 50%, suggests that the free gas at the reservoir is negligible.
These two observations corroborate the hypothesis that this is an undersaturated reservoir, and one
should not expect free gas breakout in the reservoir on this well. Based on this conclusion, it is sufficient
to analyse liquid rate and water cut trends to extract well inflow properties.
One of the reasons that a trend of liquid rate is so important is because the well suffers from an
unusually high number of stops and starts associated with unreliable power provided by a generator
located in the West African jungle. This means that for a large proportion of the 16-month production
period, the well is in transient condition (i.e., the time during which the fluid level is drawn down from
static level, which is close to hydrostatic, to the dynamic level) because there is no packer to isolate the
pressure transient drawdown from wellbore effects (the annulus is completely open). The well is therefore
rarely in pseudosteady-state condition meaning that superposition is required to measure depletion, which
requires a high-frequency continuous trend of liquid flow rates. Fig. 5 demonstrates that the ESP totaled
182 stops over the 16-month production period from installation to pullout of well, which is equivalent
to 11 stops per month. As a point of reference, the usual number of stops/starts where power supply is
reliable is in the order of fewer than three stops per month.

Figure 5—Number of stops and starts during 16-month production period with uptime analysis.

Method for Historical Flow-Rate Reconstruction


Virtual flowmetering was used to reconstruct the liquid rates (oil and water) over the full electric
submersible pump (ESP) production period of 16 months to take advantage of the real-time data which
were stored in a data historian and readily available for electronic download.
8 SPE-176780-MS

The available ESP instrumentation is shown in Fig. 6. All the data shown were available in real time
with the exception of tubinghead pressure (Pth), which was manually recorded during each test. The data
frequency was on average 1 point every 6 minutes, but in actual fact was at a higher frequency of 1 point
per minute during most of the ESP startups and then reduced to 1 every 30 minutes once the ESP stabilized
as the polling rate was being managed remotely by a 24/7 surveillance center. This was deemed sufficient
to capture transients due to both wellbore storage and reservoir drawdown. Fig. 7 illustrates this point with
a typical drawdown. As a noteworthy point, the downhole gauge had a pressure resolution of 1 psi, which
did impair pressure transient analysis. More recent ESP gauges have an improved resolution of 0.1 psi,
which would enhance the workflows in the future.

Figure 6 —ESP instrumentation available. Only the six highlighted readings were used by the virtual flow-rate meter. All data were
available in real time and stored in a data historian with the exception of tubinghead pressure.
SPE-176780-MS 9

Figure 7—Pressure transient analysis of typical drawdown illustrating constant wellbore storage (unit slope), phase segregation, and
effect of two layers, but also showing that the polling frequency was sufficient to capture transients, which were quite long. Well
requires approximately 2 days to stabilize and reach infinite acting radial flow.

Two separate algorithms were selected to calculate liquid (oil plus water) rates and water cuts to avoid
any interdependency and possible iterative solution.
Liquid Rate Calculation
Liquid rate calculation was based on the power equation described by Camilleri et al. (2010) and Camilleri
and Zhou (2011). An explanation of the equation is provided in Appendix A. The technique was chosen
because it offers the four benefits listed below:
1. The use of an analytical equation derived from first principles as opposed to a correlation or neural
networks ensured that any change in measured data was translated into a change in flow rate. The
main benefit of this is that the trend in liquid rate is correctly captured even when the model is
uncalibrated. Furthermore, because the calculation respects the physics at all times, once the model
is calibrated, the validity can be confidently extrapolated to other pump operating points and other
times as long as the pump efficiency curve remains unchanged, in which case recalibration is
required.
2. The solution provides a unique solution across the full ESP operating range from zero flow up to
run-out (e.g., over 5,300 BPD at 60 Hz, as illustrated by Fig. 4) whereas other methods were found
to be sometimes limited to the pump’s recommended operating range. This feature, combined with
the inherent low inertia of an ESP, means that transient flow rates can be captured.
3. The power equation shown in Appendix A relies on the fact that measured current varies linearly
as a function of flow rate. Another way of expressing this concept is that pump-required power is
the “action” and current is the “reaction” maintaining the equilibrium between required and
10 SPE-176780-MS

generated power. This property is well suited to ESPs due to their inherent low inertia associated
with their small diameter, which means that current reacts quasi-instantaneously to any changes in
flow rate.
4. The power equation method is independent of fluid specific gravity, which was also an important
feature as changes in water cut were unknown both during startup due to phase segregation as well
as during steady-state conditions due to measurement uncertainties. Furthermore, as the ESP is
commingling production from two layers with potential differences in fluid properties despite
being in the same reservoir, there was no way of knowing what percentage of the total rate
emanates from each layer and how this changes with time and affects the average commingled
fluid specific gravity. The independence of liquid rate calculation on specific gravity also enables
decoupling of the liquid rate and water cut algorithms.
The combination of the above four properties enables capturing transient liquid rates, which are
primarily caused by slugging and startup phenomena because the calculation is valid at any pump flow
rate, has negligible inertia, and can handle rapidly changing specific gravity.
Water Cut Calculation
Water cut calculation was based on the pressure drop in the production tubing and also tested by Camilleri
and Zhou (2011). This provides a measurement of the average mixture fluid density, which can then be
translated to a water cut based on the in-situ oil and water densities if one knows the average liquid
hold-up in the tubing. It is this last factor that makes calibration important because the average liquid
holdup is usually less than 1 due to fact that tubinghead pressure is substantially less than the bubblepoint
and there is no way of knowing the holdup. Even multiphase correlations require calibration to a physical
measurement and are therefore not an alternative. The equations governing this method are detailed in
Appendix B, which shows that there is a dependency on flow rate for the frictional pressure drop.
Although this has been considered in the calculations, it could have been neglected because, even at the
maximum pump capacity of 4,500 BPD, the frictional pressure drop in 3.5-in. tubing is less than 10% of
the gravitational pressure drop, and therefore even a large error in the frictional pressure loss calculations
represents only a small error in the average density. Note that, for most of the production period, the
electric submersible pump (ESP) is operating at approximately 3,000 BPD, at which level the frictional
pressure drop represents less than 6% of the gravitational pressure drop. The main challenge for this well
was the lack of real-time high-frequency, high-resolution tubinghead pressure data, which, under normal
circumstances, would have precluded the calculation of water cut utilizing this method. The workaround
was to use manual measurements of tubinghead pressure, which are shown in Fig. 2. This was initially
345 psia after ESP startup and then dropped to between 150 and 200 psia for the rest of the production
period. Using manual episodic measurements was deemed an acceptable approximation for the following
two reasons:
1. Tubinghead pressure was known to be relatively stable and approximately equal to the production
separator, with the exception of a small choke differential pressure for a few months after the
initial ESP startup. This is common when the surface production system hydraulic inertia is
substantially greater than that of the well (i.e., when changes in well performance such as flow rate
do not have an impact on the tubinghead pressure, which is dominated by the backpressure from
the production facility). In the absence of real-time data, this can easily be verified empirically in
the field by visually observing if the tubinghead pressure varies when there is a change in ESP
frequency, which was the case in the well being analysed.
2. As discussed above, gas/oil ratio (GOR) changes are not expected on this well. Moreover, GOR
is quite low and equal to the solution GOR of ~200 scf/bbl, resulting in low compressibility of the
fluid in the tubing. This was also seen during pressure transient analysis as there is a unit slope in
SPE-176780-MS 11

the log-log plot during early time, suggesting that there is no change in fluid (see Fig. 7).
For these reasons, it was a reasonable assumption to assume that any step change in pump measured
discharge pressure could be attributed to a change in wellhead pressure caused by a choke movement and
not due to a change in fluid compressibility, which could be caused by a change in water cut or free gas.
This argument is corroborated by the fact that reservoir water cut rarely changes in “steps” and hardly ever
drops suddenly, especially in clastic waterflooded reservoirs, such as the one being analysed. Changes in
water cut are usually associated with a gradual change in discharge pressure, usually increasing. Fig. 8
illustrates how one can identify choke movements in the discharge pressure trend if one selects the
appropriate scale for the plot. The “steps” in the discharge pressure can therefore be removed to obtain
a trend in discharge pressure for use in water cut calculations as per the formula in Appendix B.

Figure 8 —Discharge pressure trend and identifying choke movements to be removed for water cut analysis.

It should be noted that the above was essentially a workaround for calculating water cut without
real-time measurement of the tubinghead pressure; however, if monitoring is planned correctly, one
should always install an electronic gauge at the wellhead with a digital real-time connection to the data
historian to provide a more reliable water cut reconstruction process. These devices are relatively
inexpensive and can be retrofitted to most wellheads.
Capturing Transient Flow Rates and Calibrating the Virtual Flowmeter
Using the above virtual flowmeters, it is possible to reproduce the rate history over a 16-month period and
also to capture the transient rates, as illustrated in Fig. 9, which provides the calculated flow rate during
drawdown following two ESP restarts. The initial rate of the ESP is much higher than the stabilized rate
because the reservoir pressure is nearly hydrostatic; therefore, at startup, the differential pressure required
to lift the well is relatively small, and the pump operates to the right of the curve. As the well is drawn
12 SPE-176780-MS

down, the fluid level drops, and therefore the lift requirement increases and the pump operates at a lower
flow rate. This is, of course, a gradual process, which takes approximately 3 to 4 hours for the rate to drop
from an initial 4,600 RBPD to a stabilized rate of 3,500 RBPD (see Fig. 9).

Figure 9 —Capturing liquid rate transients based on measured pressure and current transients.

Note the high current at startup is a reflection of the high water cut caused by phase segregation, which
enables the power equation to correctly capture the high flow rate at startup despite a dynamically
changing specific gravity. This is confirmed by the fact that the drop in power required due to the pump
properties is less than 4% (see pump curve in Fig. 4) whereas the current drops by more than 10%. The
difference of 6% can only be attributed to the change in specific gravity.
This transient analysis not only provides the confidence that the power equation provides liquid rate
across the full operating range of the ESP, but is essential to correctly calibrating the flow-rate model
using tests that are carried out during pseudosteady state condition, as shown in Fig. 10. Calibrating the
liquid rate calculation is a process that involves matching the calculated rate to the measured rate by
correcting the absorbed power calculation with a constant (i.e., the calibration factor); however, this must
be done on a like-for-like basis. Whereas the calculated rate provides an instantaneous flowrate, the test
separator reports an average over a given time period, which is 8 hours in this case. Therefore, the
comparison is only valid when the well is in pseudosteady-state condition and the variation of actual flow
rate over the 8-hour test period is negligible.
SPE-176780-MS 13

Figure 10 —Explanation of calibration process. Red dots show test separator and sampling results for liquid rate and water cut,
respectively. Points with black circles have been used to calibrate the virtual flowmetering model.

The calibration shown in Fig. 10 demonstrates the complementary nature between physical testing in
the field and this particular method of virtual flow metering. The calculated rates (oil and water) provide
the repeatability, resolution, and high frequency, which complements the accuracy of test separator
physical metering. In essence, the calculated flow rate provides greater accuracy in capturing the change
in rates as opposed to the absolute rates. It is for this reason that the final calibration shown in Fig. 10 does
not seek to match all the physical measurement points, but only those that
● are confirmed to be during steady-state conditions (i.e., not during the drawdown transient)
● match the change in rate calculated by the virtual meter, which captures true change in rate due
to the analytical nature of the equations used
Finally, a single calibration was used for the full 16-month production period (i.e., a unique calibration
factor was applied to the analytical models for both liquid rate and water cut for the whole production
period). This is key to using the virtual flowmeter as a forecasting tool and obtaining reliable rate trends
for the 4 months prior to the start of physical testing and the 5 months after physical testing stopped (see
Fig. 3). The final calibrated full production period reconstruction is shown in Fig. 11.
14 SPE-176780-MS

Figure 11—Complete production composite log.

The only caveat to using a single calibration factor for the liquid flow-rate model is the assumption that
the pump efficiency curve remains unchanged, which may not be the case if there is pump mechanical
wear, free-gas breakout, or a change in viscosity. It was the need to verify this assumption that was the
main driver for developing an algorithm that can detect changes in pump efficiency without any flow-rate
measurements.
ESP Health Quality Flag and Flow-Rate Calibration Enhancements
The potential Achilles heel of the virtual flowmeter is that recalibration is required if there is a change in
pump efficiency over the 16-month period as one assumes that prior to calibration, the efficiency is equal
to the “as new” condition of the pump. Although it was possible in this case to rule out pump efficiency
degradation due to gas and viscosity based on fluid property knowledge as explained above, it was not
possible to rule out the possible effect of wear, especially since the ESP was producing from a sandstone.
Despite the gravel pack completion, wear due to fines could occur, and the potential impact on the
pump needed to be considered as the efficiency used was the factory five-point test, which provides the
“as new” condition of the pump (see data in Fig. 4). To enable this analysis, a new pump health indicator
(PHI) monitoring algorithm was developed, which did not require flow-rate measurement. This last
property was important in view of the scarce production test data and the uncertainty associated with the
few test points available.
Traditional pump condition monitoring is based on comparing the measured pump differential head at
a given flow rate with the “as new” head from the reference curve, which is typically the pump factory
test curve. The weakness of this process is that it requires an accurate flow-rate measurement, and so it
is difficult to use where there is a lack of test data, as in the case being discussed.
SPE-176780-MS 15

The new method also compares “actual” to “as new” condition, but instead of performing the analysis
at a given flow rate, it is done at a given differential pressure across the pump, which is readily available
from real-time data. Furthermore, the value compared is the ratio of differential pressure to pump absorbed
power using the new PHI, as expressed in Eq. 1:
(1)

This formulation is useful because one can demonstrate that it is a function of the three pump
degradation factors shown in Eqs. 2, 3, and 4. A more detailed explanation of the workflow can be found
in the patent by Camilleri (2015).
Head degradation (Ch) is
(2)

Efficiency degradation (C␩) is


(3)

Flow degradation (Cq) is


(4)

The pump absorbed power is calculated by the same method used for liquid rate calculation (i.e.,
utilizing voltage, current, and frequency real-time measurements).
The comparison is performed in two ways:
● The first comparison method is shown in Fig. 12. This is particularly useful for calibrating the
model during the few weeks or months following an installation when one can safely assume that
the pump is operating in its “as new” condition.

Figure 12—Ratio of differential pressure to absorbed power plotted independently of flow rate showing comparison of “as new” with
actual field measurements after power calibration. Measurements that deviate from the reference curve show imminent failure.
16 SPE-176780-MS

● The second comparison method is shown Fig. 13 and plots the PHI shown in Eq. 1. Once the
model is calibrated, a PHI of 1.0 indicates that the pump is operating in its “as new” condition
without any degradation in performance. Any deviation from 1.0 implies degradation as the actual
measurements are no longer equal to the reference.

Figure 13—Plot of PHI versus time, confirming that the pump is operated in “as new” condition for the full 16 months except for a few
days prior to motor failure.

Three interesting properties of plotting the ratio of DP/P as shown in Fig. 12 are listed below and are
used in the subsequent interpretation:
● The points always go through the origin by definition irrespective of the pump condition.
● DP/P is independent of specific gravity as it is implicit in both the numerator and denominator and
therefore cancels itself out.
● The reference curve is quasilinear when operating at flow rates greater than the pump best
efficiency point (BEP).
Calibration of the model is required to remove the errors in measuring and calculating the pump
absorbed power. In other applications, one would also need to remove errors due to pump degradation
caused by gas or viscosity, but these are not present in the well being studied. A further calibration is
required to correct downhole pressure at pump in-situ conditions of absolute pressure and temperature to
the reference curve specific gravity, which, in this case is 1.0 as the factory test curves are conducted with
fresh water. In Fig. 12, both these calibrations are orthogonal and decoupled because
● Calibration shift along the ordinate axis is inversely proportional to the pump absorbed power and
is independent of specific gravity, as explained above.
● Calibration shift along the abscissa axis is required to correct the downhole measured pressure to
an equivalent surface water pressure caused by the specific gravity differences. This is usually a
SPE-176780-MS 17

two-step process. The first step is to correct between downhole and surface at a reference time
(e.g., the installation date). The second step corrects for the change in downhole pump specific
gravity over time usually due to a change in water cut, which is obtained from the change in
pressure drop in the production tubing as used for water cut estimation.
This decoupling of the calibration process in the ordinate and abscissa directions ensures that there is
a unique calibration for power and specific gravity and not an infinite number of combinations of the two.
Note that for the purpose of detecting a change in pump efficiency, the power calibration along the
ordinate axis is not strictly required. If not performed, the PHI would simply trend at a value other than
1, and a deviation from the base line would indicate a change in efficiency. There is, nevertheless, value
in performing this calibration, especially in cases where there is no gas and viscosity degradation as the
theory indicates that the same power calculation calibration factor applies to both the PHI and the liquid
rate algorithm. The diagnostic plot in Fig. 12 can therefore be used to calibrate both of the virtual
flowmeter models for liquid rate and water cut where the user does not expect any degradation due to gas
or viscosity without even needing a physical flow-rate measurement. This is particularly useful on wells
where individual well tests are not possible, which is often the case in subsea applications and remote
wells that have commingled multiphase production. It is also beneficial in wells such as the one discussed
in this paper where physical test separator frequency is low. For this case study, the pump absorbed power
correction factor utilized to calibrate the diagnostic plot in Fig. 12 is the exact same value used in the
liquid rate calculation technique. Furthermore, it is the only calibration factor utilized in the liquid rate
model for the whole 16-month production period.
The time-dependent plot of the PHI shown in Fig. 13 serves to identify if and when there is a need to
recalibrate the virtual flow-rate models, which was not required in the well discussed and confirmed the
use of the virtual flowmeters as forecasting tools. This PHI plot can also be used as an early indicator of
imminent ESP failure, as shown by the green points in Fig. 13 and the zoom-in on the PHI prior to failure
in Fig. 14. In actual fact, a dismantle inspection and failure analysis (DIFA) was conducted after the ESP
was pulled out of the well, and the failure root cause analysis indicated that the stator insulation was
damaged by rotor strike. One of the symptoms observed while the pump was still operational was a rapid
rise in variable speed drive (VSD) output current from 300 to 330 amps within a single day following a
startup. Although the mechanical load had not changed substantially during this period, the PHI calculates
the pump absorbed power assuming that the motor is in an “as new” condition and therefore calculated
a higher power, which is why the PHI is less than 1 when there is insulation loss. In general, PHI is greater
than 1 in most other pump-related conditions such as degradation due to gas, viscosity, and wear.
18 SPE-176780-MS

Figure 14 —Early failure detection utilizing the PHI.

Optimising Power Consumption


One of the natural by-products of utilizing the power equation for the liquid rate calculation and the above
pump health indicator (PHI) is that a detailed calibrated motor model is required to calculate the pump
absorbed power; the model must take into consideration the changes in motor saturation as voltage is
modified and thereby provide an estimate of the product of PF (power factor) and ␩m (motor efficiency)
at all times. This means that it is possible to calculate the voltage that would maximize the PF ⫻ ␩m. Fig.
15 illustrates this workflow for the present case study. Rv is plotted instead of voltage, where Rv is the
normalized voltage obtained by dividing the voltage by the motor nameplate voltage after correcting for
frequency, as shown in Eq. 5.
(5)
SPE-176780-MS 19

Figure 15—Optimising motor power consumption with a detailed motor model. KVA is expressed at the wellhead and includes the cable
losses.

This graph illustrates how a reduction in voltage can increase the PF ⫻ ␩m product when the motor is
underloaded, thereby reducing the consumed KVA as well as reducing the motor operating temperature.
Rv was in actual fact reduced from approximately 1.0 to 0.95 in October 2011, and one can see the actual
KVA is nearly equal to the ideal KVA. The reduction in power consumed is relatively small on this well
at approximately 5 KVA on a total of 180 KVA i.e. approximately 3% saving, but the workflow is
noteworthy as it can be substantially greater on other wells.

Monitoring Pump Operating Point


Most pump operators keep a close eye on where the pump is operating relative to the best efficiency point
(BEP). This is because there is less mechanical wear on the pump, and power consumption is minimized
when operating within the recommended range (i.e., as close to BEP as possible). Interestingly, calcu-
lating the liquid rate using the power equation virtual flowmeter provides the downhole average flow rate
through the pump without the need to correct from the surface measured rate to downhole conditions
utilizing formation volume factor effects. To illustrate this monitoring workflow, the downhole flow rate
has been plotted in Fig. 16 for this case study after normalizing by the flow rate at BEP to remove the
effects of frequency. The ESP was operated at near BEP throughout the 16 months of production, with
the exception of the few hours following each startup, and therefore there was no need to change pump
frequency or choke position on this well for this purpose. This technique is nevertheless noteworthy as the
liquid rate effectively automates an often laborious and repetitive task for many field production engineers
who carry out nodal analysis on a regular basis (e.g., monthly) to monitor pump operating point.
20 SPE-176780-MS

Figure 16 —Monitoring pump operating point relative to BEP and recommended operating range limits.

Obtaining Reservoir Pressure and Well Productivity Index


The key to optimising production in a well equipped with artificial lift is to obtain an inflow performance
relationship (IPR) curve for the well. Although the previous analysis provided the necessary dynamic flow
rates and pressures, the reservoir pressure was unknown for the following reasons:
● There are no recorded buildups because the downhole permanent gauge was unfortunately not
operational during the numerous buildups. This is a common problem where the gauge power is
connected to the main power source, which is unavailable during most of the shut-downs.
● Pore pressure measurements from drilling could not be used, either, as the well and reservoir had
been on production for some time prior to this 16-month production period.
Fortunately, there was one recorded buildup immediately prior to the 16-month production period
being discussed, which recorded a pressure at gauge depth of 3,300 psia. This buildup was performed prior
to acid stimulation of the formation, and therefore the subsequent drawdown could not be used to measure
the productivity index (PI) as the skin has changed as a result of the stimulation. This therefore led to the
question of whether the well was suffering from depletion and if any variation of skin was present due to
the gravel packs.
A simple observation of the pressure trend led to the hypothesis that the well was indeed seeing
boundary-dominated flow, especially during the first 6 months (see Fig. 11). Plotting the rate normalized
drawdown as expressed in Eq. 6 (Fig. 17) corroborated that there was indeed depletion as well as
confirming a constant PI (i.e., unchanging skin) during this same initial time period:
(6)

where A is the inverse of the PI, and B is the derivative of reservoir pressure with respect to time.
SPE-176780-MS 21

Figure 17—Identifying pseudosteady-state (boundary-dominated) flow regime and constant depletion and constant PI during first 6
months of production as well as subsequent reduction in PI. Vertical lines on DP/Q plot are ignored as this is for well in transient
condition.

The plot of the rate normalized drawdown (Fig. 17) is essentially a representation of the pseudosteady-
state flow regime (also known as boundary dominated), and the full derivation can be found Dake (1978).
Traditionally, this is plotted assuming a constant rate drawdown; however, the rate normalization takes
advantage of the fact that the full liquid rate trend has been reconstructed, and this representation allows
multirate production to be analysed on a single graph. The benefit of this analysis is that where the data
matches the linear model expressed in Eq. 6, one can conclude that the PI is constant, which is very useful
in any simulation. Fig. 17 illustrates the time periods when PI is constant and where a decrease in PI can
be observed, which can be assumed to be due to a skin increase.
With this information in hand, it was possible to build a single-phase well simulation assuming constant
PI, and the results are shown in Fig. 18. The difference between Fig. 17 and Fig. 18 is the quantification
of the depletion and skin increase, with Fig. 18 providing greater confidence that the depletion seen during
the first 6 months was indeed approximately 280 psi (or 1.6 psi/day) and that subsequent to this period,
reservoir pressure decline reduced to 0.3 psi/day. Note that in order to evaluate the uncertainty in such
analysis, another match was obtained and shown in dotted line for a lower initial decline and then higher
skin, which means a lower PI during the second period of production. This second simulation shown in
dotted line has the advantage that it is corroborated by the peak pressures captured at the beginning of each
drawdown (see Fig. 17). In this case, the key was not necessarily the absolute pressure or PI but the trend
in order to manage drawdown. In terms of well production gain estimation, to be conservative, one should
22 SPE-176780-MS

assume the worst case PI, which is associated with the highest reservoir pressure. This uncertainty could
be resolved by capturing a build-up and confirming the absolute reservoir pressure during the second time
period. For this case study, the value would have been to confirm the skin and whether further stimulation
would have been justified.

Figure 18 —Single-phase simulation of pressure as a function of flowrate trend showing initial depletion of 300 psi over 175 days and
subsequent stabilization of reservoir pressure.

The trend in reservoir pressure is of value to the reservoir engineer in many ways, but confirming low
depletion or stable pressure is a proxy that there is active pressure support at the well level, which, in this
case, comes from aquifer waterflooding. The well could, therefore, support additional drawdown if no
other wells are drilled in close proximity. Furthermore, the water cut was stable over the 16-month period
and the well has not seen water coning, which suggests that additional drawdown would increase recovery
factor and not just additional oil rate.
Based on this, the pump speed could have been increased from 54 to 60 Hz and delivered a production
net gain of 160 BOPD, as shown in Fig. 19. This was not done at the time as the ESP failed; however,
the analysis illustrates the value of rate reconstruction.
SPE-176780-MS 23

Figure 19 —Potential production optimisation of well by increasing pump speed from 54 to 60 Hz.

Conclusion
Although this study is representative of one well only, the use of analytical methods with a solid
foundation in physics means that the same workflows can be applied to other waterflooded wells, which
also suffer from a dearth of testing data. It also reinforces the case for equipping wells with downhole
permanent gauges that provide both intake and discharge pump pressures which are stored on a
high-frequency data historian. Interestingly, many ESP wells do have such data sets but lack real-time
wellhead pressure gauges, which are relatively simple to retrofit without needing a workover, and would
greatly enhance the workflows described. Another lesson learnt is the recommendation to install unin-
terrupted power supply for the downhole gauges to capture build-ups when the main power supply is not
available for the gauges. Finally, are we seeing the dawn of a virtual flowmeter that does not require
calibration against a physical meter? The jury may still be out on this point, however, it is possible to
confirm that the pump health indicator provides real-time feedback on when the flow-rate virtual meter
requires recalibration, thereby removing the need for regular testing and recalibration and enhancing
automation. The operator can therefore test on demand when the virtual flowmeter falls out of calibration
allowing production operators to optimise testing logistics and potentially reduce operating costs while
also improving the overall quality of reservoir surveillance. Finally but not least, high frequency and high
rate reconstruction enables identification of depletion rate and PI changes even where the well is often in
transient state.
24 SPE-176780-MS

Nomenclature

Abbreviation Example Unit of Measure Meaning:

B nondimensional Formation volume factor


Ch nondimensional Pump head degradation
Cq nondimensional Pump flow degradation
C␩ nondimensional Pump efficiency degrdaation
DP psi Differential pressure across the pump
DH ft or m Differential head
F Hz Power frequency
g ft/s2 Gravitational acceleration
H ft Pump differential head
h ft Vertical depth of ESP relative to wellhead
Power hp Power absorbed by the pump
HL nondimensional Average tubing liquid Hold-up
Pi psia Initial reservoir pressure
Pd psia Pump discharge pressure
PIP psia Pump intake pressure
Pwf psia Flowing pressure
Pth psia Tubing head pressure
PI Bpd/psi Productivity index
␩p nondimensional Pump efficiency
␩m nondimensional Motor efficiency
I amps Current, subscripts are
– d for current measured at variable speed drive output
– m for downhole current at motor

V volts Motor voltage (volts) and subscripts define where the voltage is measured:
– m for downhole voltage at motor terminals.
– s for surface voltage on secondary of step-up transformer where this exists.
– d for voltage measured at variable speed drive output.

PF nondimensional Motor power factor


Q bpd Flowrate
Subscript a Actual measurement
Subscript r Reference measurement
Subscript np Nameplate
␳ nondimensional Density and with m, w, HC subscripts for mixture, water, and hydrocarbon

References
API RP86m (2005) Recommended Practice for Measurement of Multiphase Flow, first edition. 2005.
Washington, D.C.: API.
Camilleri, L.A.P., Banciu, T., and Ditoiu, G. 2010. First Installation of Five ESPs Offshore Roma-
nia—A Case Study and Lessons Learned. Presented at the SPE Intelligent Energy Conference and
Exhibition, Utrecht, The Netherlands, 23–25 March. SPE-127593-MS. http://dx.doi.org/10.2118/
127593-MS.
Camilleri, L.A.P. and Zhou, W. 2011. Obtaining Real-Time Flowrate, Water Cut and Reservoir
Diagnostics from ESP Gauge Data. Presented at the SPE Offshore Europe Oil and Gas Conference
and Exhibition held in Aberdeen, UK, 6 – 8 September 2011. SPE-145542-MS. http://dx.doi.org/
10.2118/145542-MS.
SPE-176780-MS 25

Camilleri, L. A. P., inventor; Schlumberger Technology Corporation, assignee. System, Method, and
Computer Readable Medium for Calculating Well Flow Rates Produced with Electric Submersible
Pumps. U.S. Pat. No. 8,527,219. Sept. 1, 2013.
Camilleri, L.A.P., inventor; Schlumberger Technology Corporation, assignee. Centrifugal Pump
Degradation Monitoring Without Flow Rate Measurement. International Application No. PCT/
US15/44241, Aug. 7, 2015.
Dake, L.P. 1978. Fundamentals of Reservoir Engineering. Elsevier.
26 SPE-176780-MS

Appendix A
Liquid Rate Calculation

The method for calculating liquid rate is based on the electric submersible pump (ESP) “power equilibrium equation.” This is
based on the principle that the torque and speed of the pump absorbed power and motor-generated power are equal at all times
in an ESP, as shown in Eq. A-1 in which pump absorbed power is on the left-hand side and motor-generated power on the
right-hand side:
(A-1)

The only unknown is the flow rate through the pump (i.e., Qp), which can therefore be calculated using real-time data. A
more detailed procedure for resolving this equation is contained in the patent by Camilleri (2013). The following is a review
of the meaning of each term in this equation, and, more importantly, the source of the data and assumptions made. Eq. A-1
includes the constants that should be used in conjunction with oilfield units to calculate the power in horsepower:
DP This is the differential pressure across the pump in psi. The downhole gauge measures the
intake (PIP) and discharge (Pd) pressures in real time; this is therefore just the difference
between the two. Note that it is important to run a gauge that measures discharge pressure
in addition to intake pressure to enable flow-rate calculation.
␩p This is pump efficiency. Because the pump efficiency cannot be measured directly, the
solution requires that the pump efficiency be based on that of a new pump taken from either
the catalogue or test curves.
I This is the motor current (amps) and is usually measured directly by the variable speed drive
(VSD) or switchboard controller. Often the current is measured at the VSD and needs to be
multiplied by the transformer ratio to obtain downhole current.
Vm This is the downhole motor voltage (volts). Because only surface voltage can be measured,
it is necessary to subtract the voltage loss in the power cable. This can be estimated from
cable resistance properties and the measured current. Note that although it is important to
make the correction, the estimate does not need to be accurate, because the voltage drop
typically represents less than 10% of the total voltage; therefore, even a 20% error in
voltage drop estimate has only a 2% error impact on the flow-rate calculation. As with the
current measurement, the surface voltage is measured by the VSD controller and therefore
needs to be divided by the transformer ratio to obtain downhole voltage.
␩m⫻PF This is the motor efficiency and power factor. For the current case study, these values could
not be measured. However, it is possible to take direct measurements of power factor with
additional surface electrical instrumentation, which can be considered on other wells.
Because variably rated motors are used, this factor can be assumed to be constant for load
factors greater than 50 to 60%, and therefore liquid rate trends are not affected when the
motor is loaded more than 60%, although calibration is required. For the present case study,
a motor model based on motor laboratory test data calculated the product of motor
efficiency and power factor for all loads and voltages, thereby ensuring that the liquid rate
was valid at low startup frequencies when the load factor is low.
SPE-176780-MS 27

Appendix B
Water Cut Calculation

The proposed method for calculating water is based on calculating the average mixture density in the production tubing above
the electric submersible pump (ESP) as a function of the pressure difference between the pump discharge and wellhead as
expressed by Eq. B-1. This technique is analogous to a gradiometer with the difference being that because of the length of the
tubing, friction must be taken into consideration to achieve a single calibration over a wide range of flow rates. In environments
where the gas/oil ratio (GOR) is either constant or insignificant, the change in density can be related to a water cut change using
Eq. B-2. Once such a model is calibrated against measured water cut, ideally measured with a multiphase flowmeter, trending
water cut with time becomes possible. This technique requires both pump-discharge pressure and tubinghead pressure
measurements in real time, and when slugging is present in the tubing, the data are required at high frequency to capture
pressure oscillations.
The relationship between mixture density and outflow pressures is
(B-1)

The relationship between mixture density and water cut for zero slip is
(B-2)

You might also like