Sei sulla pagina 1di 8

On the Influence of Viscosity on

ESP Performance
Gilmar Amaral and Valdir Estevam, SPE, Petrobras, and Fernando A. Franca, SPE,
the State University of Campinas

Summary
This paper presents experimental data and a preliminary analysis
of the influence of fluid viscosity on centrifugal-pump performance. Two centrifugal pumps, a conventional radial (specific
speed Nq=8 rev/min) and a semi axial electrical submersible pump
(ESP) (Nq=28 rev/min) were tested with 1-cp water and clear
glycerin. Adjusting and controlling the fluid temperature in a
closed test loop, it was possible to vary the glycerin viscosity from
67 to 1,020 cp within the range of light and heavy crude oils.
The main purpose of these tests, in addition to appraising the
influence of viscosity on the pumps overall performance through
the measurement of the derating factors for head, flow, and power,
was to supply detailed information on the energy-transfer processes taking place in the pumps internal components. To accomplish
this, the pressure distribution along the flow path from the pump
inlet eye to the discharge section, including detailed pressure
difference across impellers and diffusers, was measured. Thus, in
addition to measuring the flow rate, the overall pressure difference, the speed, the power and the mean operation temperature for
fluids with various viscosities within a full range of operational
conditions, detailed data on the energy-transfer processes performed by impellers and diffusers were also taken.
Later analyses indicated that, in addition to the physical
dimensions, operational conditions, and fluid properties, the pump
performance is set by the strong flow interactions that exist between impellers and diffusers. In other words, these succeeding
internal blade rows influence each other in terms of the head gain
and the viscous dissipation effects. Thus, any generalizing approach dealing with the influence of viscosity on the pump performance must account for those interactions to give a proper
measure of the derating factors over an extended range of operational conditions. Unfortunately, this is not true for the procedures
available in the open literature. They lack representation and do
not deliver proper correction factors for pumps that are not similar
to those that generate the correlation database or for pumps working under operational conditions other than at the best-efficiency
point (BEP). The data presented herein can be a launching point
for a deeper analysis aimed to tackle these limitations.
Introduction
ESPs have been used increasingly as an artificial-lift method for
producing medium-to-heavy oils in deep offshore fields. The
pump may be positioned inside the well or, depending on technical requirements, even above the seabed to diminish intervention
costs. When the oil viscosity is too high to enable an efficient
pump operation, injecting light oil (usually referred to as a downgrading technique) or solvents upstream of the pump suction line
or acting on the mixture temperature to reduce the oil viscosity are
some of the strategies to be considered. When the pumping system
is installed on the seabed, series-parallel arrangements may be
used to reduce the power usage of each single pump. Thus, new
operation strategies and different equipment allocation and
arrangements have been used to exploit heavy oils in these harsh
environments. This is precisely the case in Jubarte, a pioneering
Copyright 2009 Society of Petroleum Engineers
This paper (SPE 110661) was accepted for presentation at the SPE Annual Technical
Conference and Exhibition, Anaheim, California, 1114 November 2007, and revised for
publication. Original manuscript received for review 1 August 2007. Revised manuscript
received for review 3 April 2008. Paper peer approved 7 May 2008.

May 2009 SPE Production & Operations

heavy-oil field in the Campos basin, Brazil. After careful technical and economic analyses, a seabed-installed ESP system was the
technique chosen to lift the heavy-oil/low-gas-content mixture to
the platform without any previous phase separation. Thus, two
parallel connected ESPs were mounted on a skid that was placed
on the seabed, close to a wellhead.
Sizing the ESP accurately is very important in this offshorefield scenario. When the ESP is installed on the seabed, the produced-mixture temperature at the pump intake is lower than the
reservoir temperature. Moreover, some of the heat generated by
the electrical motor and the viscous dissipation inside the ESP,
which contributes to the reduction of the fluid viscosity, may be
lost partially to the surrounding cold water. Thus, in seabed installations, the pumped mixture attains a lower temperature and a
higher viscosity, degrading the pump performance. Furthermore,
compared to an in-well pump installation, the reduced pressure at
the pump intake may increase the free-gas content (GOR), additionally degrading the pump performance. These are some of the
reasons that the knowledge of the fluid viscosity is a key technical
issue and that special care is required when sizing seabed-installed
ESPs to lift medium-to-heavy oils in offshore fields.
The influence of fluid viscosity on centrifugal-pump performance has received attention for many decades. The approaches
found in literature to predict the pump performance with viscous
liquids can be summarized into two groups: (a) empirical methods
and (b) deterministic methods. The empirical methods usually try to
adjust the pump performance with the water baseline curves as a
reference. Viscosity deratingor correctionfactors adjust head,
flow rate, and efficiency in regard to the water baseline conditions,
providing the pumps actual performance. They deliver quick
results of practical use when the derating-factors database was taken
from similar pumps, which means pumps with the same specific
speed over a broad range of operational conditions. Unfortunately,
this is not always true because most of the publicly available charts
and correlations apply for pumps with narrow specific speed limits,
far from the range that applies to ESPs. Moreover, ESPs used for
crude-oil lifting usually present a challenging production history
because the flow rate and fluid properties may vary within a short
period of time. Many of the viscosity derating databases were produced for a fixed operational condition, the pumps BEP. Thus, the
lack of agreement on specific-speed limits and restrictions on operational conditions make the estimation of viscosity derating factors
an open issue as far as the ESPs are concerned.
A typical empirical method was proposed by Stepanoff (1957),
who correlated data to find the viscosity correcting factors for head
and efficiency in terms of properly defined centrifugal pump
Reynolds number. Another empirical approach appears in the
charts provided by the Hydraulic Institute-USA (HI-USA) (1983).
In this case, the pumps were tested with 1-to-3,000 cp fluids. The
correctness of the factors provided by the Hydraulic Institute
charts has been questioned when used for oil pumps, in general,
and ESPs, specifically. According to Gulich (1999) and Li (2002),
the HI-USA conducted tests within narrow limits of the pump
specific speed. The former found that his own experimental data
best agreed with the HI-USA viscosity factors for pumps with the
specific speed Nq=5 rev/min (Ns=600 rev/min, when N is given in
min-1, Q is given in gal/min, and H is given in ft), defined as
Nq

p
N Q
3

H =4

; min1 ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)

303

with N given in min-1, Q given in m3/s, and H is given in m, water


performance at BEP condition. This is far from the usual operational range of oil pumps.
Paciga (1967) used a procedure quite similar to Stepanoffs,
proposing a new formulation for the Reynolds number. His approach accounted for the pump specific speed, with the author
claiming that this gave rise to a generalizing viscosity data-reduction procedure. Gulichs (1999) procedure relies on the estimation
of the overall viscous dissipation inside the pump, with the author
claiming that they are the most important factors when determining pump performance. Turzo et al. (2000), after a critical comparison of available approaches, used the HI-USA database to
propose a computational procedure to estimate the head and efficiency derating factors for a given flow rate. Li (2002) collected
data with an API610 Standard pump operating with water and
more-viscous oils up to 240 cp. His viscosity correcting factors
for flow rate, head, and efficiency are always less than the ones
disclosed by the HI-USA charts: -10, -5, and -9.7%, respectively. Sun (2002) proposed a more general approach to calculate
the pump performance for viscous-fluids operation that was based
on the 1D formulation of the conservation equations. With no
viscous flow solution as a starting point, the authors added the
viscous dissipation effects taking place in curved and rotating
impeller channels. The paper does not include, however, the
results for viscous fluids other than water.
ESP manufacturers are also making major efforts to develop
reliable procedures to obtain viscosity derating factors for head,
flow rate, and power. A comprehensive work presenting the influence of crude oils and produced fluids in general on ESPs performance was conducted by Patterson at al. (2002). The authors
conducted tests at different viscosities and investigated the waterin-oil-emulsion formation caused by the pumping action. They
learned that the API gravity alone is not a good indicator of crude
viscosity; knowledge of oil composition, especially the paraffin or
asphaltene content, is important because crude oils with the same
API grade may have a different impact on the pump performance.
For this reason, they claimed that the widely used derating factors
usually obtained from extensive tests with mineral oils may not
correct an ESP or even a conventional-oil-pump performance for
crude-oil operations. This is important because the same conclusion can be drawn when a homogeneous fluid such as glycerin is
used in tests, as in the ones presented herein.
Deterministic methods use computational fluid dynamics
(CFD) to simulate the flow through the pump and to calculate the
head and efficiency. Some CFD methods are designed specifically
for flow calculations in rotating fluid machines, such as pumps
and turbines. They rely on constitutive relationships to calculate
the turbulence terms and to predict the action of strong centrifugal
fields on the turbulence suppression, phenomena that carry open
questions. Their main disadvantages are that (1) they are costly;

(2) the CFD flow calculations inside pumps are time consuming;
(3) the CFD algorithm may not converge in every circumstance;
(4) they usually rely on constitutive relations obtained for restricted flow conditions under simplifying assumptions; and (5) occasionally, they do not give good pump-performance estimations if
they are not able to capture the flow complexities inside the pump
channels or simulate properly the flow interactions that exist between the pumps internal components. In this respect, it is worthwhile to know about the jet-wake-type flow complexities that
exist between the two interacting blade rows in pumps, the impeller and the diffuser. The experimental work conducted by Akhras
et al. (2004) gives a very detailed description of these complexities because the experimentations were performed at different
operating points besides the BEP, and this work points out that
the extension of the wake seems to be affected by the flow rate.
The present work is a first step toward the development of a
generalizing approach aimed at prediction of the viscosity influence on ESP performance. It is part of a broader research project
that focuses on the influence of fluid viscosity and inlet GOR on
ESP and conventional centrifugal-pump performance. To accomplish this, a semi axial ESP and a conventional two-impeller radial
pump were tested with water (1 cp) and double-filtered clear
glycerin in a very broad viscosity range (from 60 cp at 60 C to
1,020 cp at 20 C). The pressure distribution and head gain along
the flow path, through impellers and diffusers, from the pump
inlet eye to the discharge section, were measured for a full range
of operational conditions. As a result, the newly formed database
not only disclosed the pump overall performance but also revealed
details of the energy-transfer processes taking place in the pump
internal channels. The measurements at different flow rates, aside
from the BEP condition, revealed the strong interactions that exist
between the succeeding pump blade rows, as diffusers and impellers, in terms of head gain and viscous dissipation. Additionally,
deviations between the actual pump curve and the similarity
conditions, resulting from changes in fluid viscosity, were found
for the limiting values of viscosity and pump speed applied in
the tests.
The Experimental Setup
A schematic of the water and clear-glycerin (density r = 1260
kg/m3) pumps testing loop is shown in Fig. 1. The series-parallel
pumping system was formed by a booster pump and two test
pumps. Because two different pumps were tested, an ESP and a
conventional radial two-impeller pump, they were mounted parallel in the test loop. Each electrical-pump motor (15 kW for the
conventional radial pump and 40 kW for the ESP) was driven by
its own dedicated variable-speed driver. The test pumps, one at a
time, suctioned from and discharged to the same reservoirs, in a
series association with the booster pump. Part of the system head
loss was supplied by the booster pump, allowing for pressure

Fig. 1Schematic of the test loop. TQ: torque.


304

May 2009 SPE Production & Operations

changes at the test-pump intake. Besides, the test pumps could


operate within a full range of operational conditions, including
the high-flow-rate/low-head region. Thus, the test-pump operational conditions could be varied greatly and their performances measured over a broad range of flow rate, pressure or head,
and efficiency. To vary the glycerin temperature between 20 and
60 C, which allowed for a change in viscosity between 60 and
1,020 cp, a heat exchanger and a heat pump [20 tons of refrigeration (tr)] were connected to the system. In addition to adjusting
the fluid temperature, the heat pump controlled it within the uncertainty range of +/-1 C. This is quite important because the
glycerin has an exponential-decay temperature-dependent viscosity. In the end, simultaneous changes in the pump speed and fluid
viscosity allowed for data acquisition focusing on deviations from
the similarity operation.
Thirteen OmegaTM capacitance pressure sensors, one Omega
torque cell, one LebowTM torque cell, a Micro MotionTM mass
flow meter (0.075-m connecting flanges), one optical speedometer,
thermocouples, four Advantech ADAMTM data-acquisition modules and serial interface, one PC, the GeniDaqTM software, and the
ExcelTM spreadsheet program working under Windows XP in dynamic data link (DDL) mode made up the instrumentation and
data-acquisition system. The same flow-metering section was used
when testing both pumps. The ESP was a semi axial GN-7000/
540TM model, with specific speed Nq=28 rev/min (Ns=38,50 rev/
min when N is given in min-1; Q is given in gal/min; H is given
in ft), and flow rate and head at the BEP condition taken from the
water-baseline-performance curve. The three-stage GN-7000 was
mounted with eight-vane impellers and 12-vane diffusers. It delivered a water-flow rate in the range of 0 to 75 m3/h; the head was
up to 43 m when in shutoff condition (0 to 11,250 B/D, head up to
140 ft), the speed varying from 1,800 to 3,500 rev/min. The waterbaseline-performance curves were verified carefully and showed
good agreement with the manufacturers published ones. As far as
the head is concerned, the maximum deviation was +6% at the
maximum flow rate, and there was absolutely no deviation at the
BEP; for efficiency, the deviation increased when the flow rate
increased, reaching +5% at the highest flow rate. There were no
significant deviations for power. A schematic of the GN-7000
partial internal view, showing the pressure measuring points along
the second stage: (1) diffuser entrance, (2) diffuser exit (impeller
entrance), and (3) impeller exit (diffuser entrance) appears in
Fig. 2. The two additional pressure taps used to measure the overall pressure-difference data for head calculation were located just
before and after the pump inlet and exit eye, respectively.
The conventional radial two-impeller test pump was the ITATM
65-330/2 model. It was mounted with a 205-mm-outside diameter
(OD) (first stage) impeller and a 260-mm-OD (second stage) impeller, both with six vanes, and a regular six-vane diffuser between them. A similarity analysis showed that this test pump and
a low flow rate in-well pump (DN-280TM model, radial impeller,
Nq=8 rev/min at BEP or Ns=1,000, when N is given in min-1; Q is
given in gal/min; H is given in ft, flow rate and head taken from
the water-performance curve), in addition to being similar geometrically, have acceptable dynamic similarity when the former
operates at 800 rev/min and the latter operates at 3,600 rev/min
(nominal speed for in-well oil pumps). The dynamic-similarity
criteria followed the Lazarkiewicz et al. proposition (1965) in
terms of the Froude-number/Reynolds-numbers ratio: RFr=1 and
RNr=4, within the suggested range of 0.07<RNr<15. Further
details on pump dimensions are given by Estevam (2002) and
Amaral and Franca (2006).

The radial pump delivered a water-flow rate in the range of


0 to 70 m3/h, the head was up to 22 m (shutoff condition at
nominal speed), and the speed was varied from 612 to 1,150 rev/
min. The water baseline curves were measured and checked
against the published curves, showing acceptable agreement: The
head deviation near the BEP condition was within 3%, the
efficiency had a -2% slight and constant decrease within the full
operational range, and there were no noticeable deviations between the published and measured power curves. The suction line
was mounted in the vertical-downward position (in regard to the
intake flow) to allow for future tests with varying inflow gas
contentvoid fraction.
Setting up pressure sensors and inspection windows in a large
conventional radial pump is far easier than in a radial ESP. Pressure taps could be located on both sides of the ITATM 65-330/
2 pump case, accessing every internal flow channel. However, this
radial pump had two impellers of different sizes, a standard diffuser connecting the two impellers, and a volute casing-type vaneless diffuser after the second impeller. Thus, the conventional
radial pump had two stages of distinct shape and size. Because of
this, aside from measuring the pressure at the pump inlet and exit
sections, six other pressure taps have been installed to measure the
pressure along the flow path through the impellers and diffusers.
A schematic of the ITA 65-330/2 sectional view is given in Fig. 3;
the numbers 1 to 6 indicate the locations of pressure-measuring
points along the flow path.
When glycerin was used as the working fluid, the viscosity was
established on the basis of temperature measurements. The glycerin-viscosity-vs.-temperature relationship was determined previously with a Fann viscometer in the 10 C-to-60 C temperature
range. It is an exponential-decay temperature-dependent relationship, as expected. However, because of the glycerin hygroscopic
nature, a daily viscosity check was performed. Additionally, special care was taken to keep the glycerin in the closed loop without
any mixing with fresh air, properly sealing the supply reservoir.

Fig. 2GN-7000 internal view: Points 1, 2, and 3 indicate the


pressure-measuring points along the second pumping unit
(impeller plus diffuser).

Fig. 3ITA 65-330/2 sectional view: 1 through 6 indicate the


pressure-measuring points inside the pump (2 impellers plus
2 diffusers).

May 2009 SPE Production & Operations

Results and Analyses


The results presented herein emphasize some questions concerning
the influence of fluid viscosity on the pumps overall performance.
These questions are (1) the viscosity-dependent processes that degrade the pump performance in regard to the water operation conditions and produce the viscosity derating factors, (2) the deviations

305

Fig. 5GN-7000 ESP, head vs. flow rate: influence of viscosity,


1-cp water to 1,020-cp glycerin, 3,500-rev/min nominal speed.
Fig. 4Radial ITA 65-330/2, head-vs.-flow-rate curve: influence
of viscosity, 1-cp water to 720-cp glycerin. Pump speed is approximately 910 rev/min. Single crosses indicate the BEP on
each curve.

upon the similarity relationships in a wide range of speed and flow


rate and fluid viscosity, (3) the energy-transfer processes in specific
devices such as the impeller and diffuser and, in the end, (4) the
identification of the limiting operational conditions owing to viscous-fluid operations.
The conventional radial ITA 65-330/2 pump head-vs.-flowrate curves for water and 87-, 350-, and 720-cp glycerin are
shown in Fig. 4. The closed dots are experimental points that were
fitted by a second-degree polynomial curve to display continuity.
Because of power limitations, the curve for 720-cp glycerin was
obtained for 800 rev/min and reduced, by similarity relationships,
to 912 rev/min. This small speed difference did not cause significant deviations between the actual curve for 912 rev/min and
the reduced curve. This data reduction was processed following
the usual similarity relationships: If N is the pump speed, the
similarity relationships for flow rate (Q), head (H), and power
(P) are
 2
 3
N2
N2
N2
Q2 Q1 ; H2 H1
; P2 P1
: . . . . (2)
N1
N1
N1
These results showed clearly the influence of fluid viscosity on the
head-vs.-flow-rate (H-vs.-Q) curve at a constant pump speed. On
each curve, the BEP is identified by a single cross. The expected
and found trend was toward the simultaneous head and flow-rate
reduction when the operational conditions at the BEP were taken
for comparisons. However, this trend was not verified over an
extended range of operational conditions when the results for different fluid viscosities were analyzed. When the curves for water,
1 cp, and the 87-cp glycerin were compared, the head deviations
diminished when the flow rate increased above the BEP. The
water baseline H-vs.-Q curve trend indicated that, at flow rates
greater than 50 m3/h, the head can be even slightly higher for
the 87-cp glycerin operation than for water. For fluids heavier than
the 87-cp glycerin, the head always decreased comparison to the
lighter fluids in the full range of tested flow rates. Another interesting aspect in these data is concerned with the head developed at
the shutoff condition: 87-cp and heavier glycerin operations produced the same head at shutoff. These results posed questions
concerning the role played by singularities, at the impeller and
diffuser inlet region, on the energy-transfer processes taking place
inside the pump. Therefore, careful pressure measurements were
taken inside the pump, along the blade-row channels forming a
pump stage.
306

In Fig. 5, the GN-7000 ESP H-vs.-Q performance curves (N =


3,500 rev/min) are shown for fluid viscosity in the range of 1 cp
(water) to 1,020 cp (glycerin at 60 C). Similarly, as in Fig. 4, the
BEP operational condition is identified by a single cross. Taking
the BEP as a reference condition, when the fluid viscosity changed,
the largest deviations in regard to head and flow rate occurred
in the lowest-viscosity range. For instance, the 1- to 60-cp fluidviscosity change produced a larger head and flow rate reduction
than the 270- to 720-cp fluid-viscosity change in the full range of
operational conditions. Thus, when the curve trends in Figs. 4 and
5 were compared, it became clear that the influence of fluid viscosity upon the pump performance depended on the pump characteristics, whether it is radial and semi axial, as well as on the
operational condition. These results supported the conjectures previously stated by Gulich (1999) and Li et al. (2002) upon the
extensive noncritical use of the HI-USA charts as a pump-performance-correction tool. Radial, semi axial, or axial pumps may
behave differently in terms of the influence of fluid viscosity on
performance; the charts and correlations used widely today do not
give necessarily a good pump degrading performance estimation
because they usually refer to a limited specific-speed range.
In actual operations, when it is quite common that a pump is
driven by a frequency inverter to enhance operational flexibility,
the estimation of pump performance under similarity conditions is
a key issue. The similarity conditions exist if a performance curve
can be estimated from a baseline curve precisely when the pump
speed changes. If this is true, the similarity relationships, Eq. 2,
merge every head-vs.-flow-rate curve generated at a given speed
into a single curve. The ITA 65-330/2 head-vs.-water-flow-rate
similarity reduced curves appear in Fig. 6. The head deviation was
no greater than 5% to the 1,150-rev/min baseline curve within the
whole range of water flow rate when the speed ratio varied between 1 and 1.9 (1,150 to 612 rev/min). Fig. 7 shows a similar
graph for the first impeller pressure-gain-vs.-water-flow rate
curves. A limiting 2% head deviation was found in the whole
range of operational conditions, encompassing the 0- to 68-m3/h
water-flow-rate range. If the similarity relationships produced
such good results for the pump, they must produce even better
results for the rotating device, the impeller. The smaller pressuredifference deviations disclosed by the impeller in regard to the
pump head, when the similarity curves in Figs. 6 and 7 are compared, supports this conclusion. When operating with water, similarity conditions also existed for the GN-7000 ESP. In this case,
the speed varied from 1,800 to 3,500 rev/min and the flow rate
was in the range spanning from 0 to 80 m3/h. As for the conventional ITA 65-330/2, when the similarity relationships were applied to reduce the performance curves to the design speed, 3,500
rev/min, the impeller pressure difference and pump head deviations were always below 2% and 5%, respectively.
May 2009 SPE Production & Operations

Fig. 6Radial ITA 65-330/2, head vs. water flow rate: data reduction according to the similarity relationships. Speed range
from 612 to 1,150 rev/min.

The operational similarity conditions require dynamic similarity


and impose constant pump efficiency. Whenever the fluid viscosity
rises, the previously attained force balances that are important in
rotating flow machines alter, deviations among similarity reduced
performance curves appear, and efficiency degradation takes place.
Fig. 8 shows the 60-cp glycerin GN-7000-ESP-impeller pressuredifference-vs.-flow-rate curves for pump speeds ranging from
1,800 to 3,500 rev/min. When these curves were reduced according
to the similarity relationships (Fig. 9), they merged into a single
curve from a practical standpoint. A reasonable agreement was
found, with limiting deviations at less than 5% at the highest
flow rates if a single trend could be claimed to exist. It is important
to emphasize that these data refer to the light fluid 60-cp-glycerin
GN-7000-ESP operation. Fig. 10 shows the heavier fluid 720-cpglycerin operation GN-7000-ESP similarity reduced head-vs.-flowrate curves. In this case, a large change in fluid viscosity led to 40%
head deviations at the highest flow rate, 38 m3/h, for a 1.9 speed
ratio variation, from 3,500 to 1,800 rev/min. The growing viscous
forces and energy dissipation in the impeller and diffuser caused by
an increasing flow-Reynolds number are based on the similaritycondition deviations and performance degradation.
The performance of the ITA 65-330/2 intermediate diffuser
can be seen in Figs. 11 and 12, in terms of the no dimensional
pressure coefficient and flow Reynolds number. It is worthwhile
to mention that a radial-pump intermediate diffuser, such as the

Fig. 8GN-7000 ESP, first-impeller pressure difference vs. 60cp-glycerin flow rate. Speed range from 1,800 to 3,500 rev/min.
May 2009 SPE Production & Operations

Fig. 7Radial ITA 65-330/2, first-impeller pressure gain vs.


water flow rate: data reduction according to the similarity relationships. Speed range from 612 to 1,150 rev/min.

one used in the ITA 65-330/2, is formed by three pieces: (1) a


radial-outward flow-path-diverging nozzle that receives the flow
leaving the impeller; (2) a sudden section reduction (e.g., a flow
singularity) in which a wide direction change creates a radial
inflow; and (3) a converging nozzle that guides the radial inflow
to the entrance of the second impeller.
Thus, the pressure coefficient, Cp, is defined by
Cp

po  pi
; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
1
2
2 r Vo

where p is the pressure, r is the fluid density, and V is the flow


velocity. The subscripts i and o indicate the device inlet and outlet
flow section, respectively. The Reynolds number, Re, is defined by
Re

r Vs D
; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
m

where Vs and D are the flow mean velocity and pipe diameter at
the pump-inlet eye, respectively, and m is the fluid viscosity.
Fig. 11 shows the Cp-vs.-Re curves for the ITA 65-330/2 firstdiffuser diverging section (e.g., the diverging nozzle), the whole
set of experimental data encompassing the full range of fluid viscosity (1 to 720 cp) and pump speed (from 612 to 1,150 rev/min).
The diffuser diverging section acted as a real diffuser, Cp>0,
which meant that the pressure difference across this diverging

Fig. 9GN-7000 ESP, first-impeller pressure difference vs. 60cp-glycerin flow rate: data reduction according to the similarity
relationships. Speed range from 1,800 to 3,500 rev/min.
307

Fig. 10GN-7000 ESP, head vs. flow rate: data reduction


according to the similarity relationships, 720-cp glycerin, speed
range from 1,800 to 3,500 rev/min.

piece was always positive in the full range of flow rate applied in the
tests, even when the 720-cp glycerin was used in the tests. Fig. 11
revealed that the Cp-vs.-Re curves are similar for every fluid viscosity when the Reynolds number and the pressure coefficient are in
the range of Re=101 to 105 and Cp=1 to 2,000, respectively. Power
equations and exponential equations in the lowest- and highestReynolds-number ranges, respectively, fitted the fluid-viscosityrelated curves. The fact that similar relationships existed for each
fluid viscosity has produced expectations that a viscosity-ratio term
could merge these curves into a single curve. If this is true, the
radial-pump diverging-section diffuser performance for various fluid viscosities and operational conditions, including the pump speed,
could be expressed by a single generalizing correlation, thus simplifying the pump overall-performance modeling.
In Fig. 12, the data and the fitted curves represent the flowpressure drop across the restriction and the converging nozzle.

Fig. 12Radial ITA 65-330/2, diffuser converging section Cp vs.


Reynolds number, 1-cp water to 720-cp glycerin. Pump speed
range from 612 to 1,150 rev/min.
308

Fig. 11Radial ITA 65-330/2, first-diffuser diverging section


Cp vs. Reynolds number. 1-cp water to 720-cp glycerin. Speed
range from 612 to 1,150 rev/min.

When the fluid reaches the converging nozzle after flowing through
the diverging nozzle and restriction, any upstream influences caused
by the flow inside the impeller and positive or negative incidenceflow-angle variations occurring at the diffuser-vane entrance section
are dampened. Thus, the restriction-converging nozzle pressuredrop-vs.-flow-rate relationship is a classical one: a log function
fitted the CpRe data for Re<2000, and the pressure coefficient
became constant for Reynolds number greater than 5104.
It is important to emphasize that this curve fitting encompassed
the whole data set, including fluid-viscosity and pump-speed variations. There were no available data in the 2,000<Re<5104
range, and additional tests performed with aqueous glycerin solutions viscosity lower than 87 cp) must be performed to fill the gap.
The data scatter for Re of approximately 6104 (1 cp water) was
caused mainly by the pressure measurement uncertainties because
the pressure difference was very low for these conditions at the low
end of the pressure-sensor measuring range. For higher flow rates
and Re>105, a constant Cp=1.25 was found, as expected.
The pressure differences across the GN-7000 ESP diffuser for
60-cp-glycerin flow rates within the 0- to 65-m3/s limits are presented in Fig. 13. The diffuser in this semi axial ESP acted as a

Fig. 13GN-7000 ESP, pressure difference across diffuser


vs. 60-cp-glycerin flow rate. Pump speed from 1,800 to 3,500
rev/min.
May 2009 SPE Production & Operations

Fig. 14GN-7000 ESP diffuser pressure difference vs. flow


rate: data reduction according to the similarity relationships.
Pump speed from 612 to 1,150 rev/min.

pressure-gain device in low flow rates, far below the BEP condition. An energy balance accounting for the regular friction dissipation, the dissipation at the vane-entrance region, and the flow
deceleration produces this particular behavior. Even at the 60-cpglycerin operation BEP condition (40 m3/h; 6,200 bbl), a small
negative pressure difference of approximately -15,000 Pa was
measured. The pressure difference decreased steadily at increasing flow rates, reaching -50,000 Pa at 65 m3/h. At this high
flow rate, 50% of the pressure gain across the preceding impeller was lost across the diffuser, -50,000 Pa out of +98,000 Pa.
More-viscous-fluid flows created a higher pressure drop. Even the
water operation at the BEP condition (50 m3/h; 7,500 bbl) caused
a pressure drop across the diffuser of -10,000 Pa. This value is
slightly lower than that produced by the 60-cp-glycerin flow.
An important finding concerning the semi axial GN-7000 ESP
diffuser operation was that the similarity relationships could be
applied successfully for data reduction for various pump speeds. It
is worthwhile to remember that the diffuser is a stationary device
that receives the flow discharged by the impeller. Nevertheless,
when the pressure-difference-vs.-flow rate data were reduced
according to the similarity relationships (Eq. 2) the set of curves
for various speeds merged into a single curve with minor deviations in terms of practical applications (Fig. 14). This was caused
by the strong flow interactions that exist between these two succeeding blade rows, the impeller and the diffuser, and the positive

May 2009 SPE Production & Operations

or negative flow deviations at the diffuser entrance vane and the


related energy-dissipation processes, both related to the flow rate.
Again, the previous expectations concerning the conventionalradial-pump diffuser-performance modeling can be widened to the
semi axial GN-7000 diffuser: If the use of similarity relationships
merges each speed-related diffuser performance curve into a single
generalizing curve with small deviations, the diffuser modeling
would be greatly simplified. It was verified that more-viscousfluid flows produced similar results, though with larger deviations
at the highest flow rates. The limiting fluid viscosity in these tests
was 1,020 cp. Thus, the analysis of data, curves, and trends appearing in Figs. 11 through 14 revealed important issues regarding the
diffuser performance, in both the radial conventional pump and the
semi axial ESP, in a wide range of operational conditions and fluid
viscosity. The variable flow incidence angle in the diffuser blades
resulting from operational conditions other than those at BEP has
caused fluid-viscosity and flow-rate dependence on both the radial
and the semi axial diffuser performance. In other words, the strong
flow interactions that exist between the successive blade rows,
impeller, and the diffuser, depend on flow rate and fluid viscosity
and, as such, play an important role when the viscosity derating
factors are to be set.
In the end, it is worthwhile to disclose the GN-7000 ESP performance deviations that have been found when the experimental
data were compared with the data provided by the HI-USA charts,
as far as the influence of viscosity on the performance is
concerned. Table 1 reveals that when the fluid viscosity changed
from 60 to 1,020 cp, the flow-rate deviation at the BEP condition
was lower than 10% for the highest-fluid-viscosity operation,
1,020 cp. However, operations at the BEP with less-viscous fluids,
from 60- to 720-cp glycerin, have produced deviations between
the measured and the HI-USA-predicted flow rates that are always
greater than 20%. Minor deviations were found between the
measured and the HI-USA-predicted head at BEP, always less
than 5%. When one looks at the pump efficiency at the BEP,
greater deviations were found once more: It decreased from a
14.3% positive deviation for a 60-cp-fluid operation to only 5.5%
for a 270-cp-fluid viscosity, but increased again to -25% for a
720-cp fluid and exceeded 35% for a 1,020-cp-fluid viscosity
operation.
Conclusions
Experimental data were obtained, and preliminary analyses on the
viscosity influence upon an ESP and a conventional-radial pump
performance were performed. These two pumps were tested with
1-cp water and 67- to 1,020-cp clear glycerin, within the light- to
heavy-oil viscosity range. The pumps overall performance was
measured, and detailed information on the energy-transfer processes taking place in diffusers and impellers was disclosed.

309

Additionally, the influences of fluid viscosity and operational


conditions on the strong flow interactions that exist between the
successive pump devices in terms of the head gain, pressure difference, and viscous dissipation were assessed. The deviations on
similarity operations caused by simultaneous changes in the pump
speed and fluid viscosity were revealed, within the full range of
operational conditions.
A preliminary analysis has led to the supposition that the usual
black-box approaches, such as the HI-USA charts for viscosity
pump-performance correction, may not be able to give proper
viscosity correction factors for different pumps because they do
not account for the pump operational conditions and the fluidviscosity effect upon the strong flow interactions that exist between the rotating and the fixed internal blade rows, the impeller,
and the diffuser. In fact, comparisons between the measured data
and the HI-USA-charts predicted data have shown flow-rate
deviations as high as 25% and pump-efficiency deviations as high
as 35% for viscosity changes in the range of 1 to 1,020 cp, as far
as the GN-7000 ESP tests are considered.
Nomenclature
Cp = pressure coefficient
H = head, m
N = pump speed, rev/min
Nq = specific speed, rev/min
Ns = specific speed, rev/min
p = pressure, Pa
P = power, W
Q = flow rate, m3/h
V = velocity, m/s
r = density, kg/m3
m = viscosity
Z = efficiency
Subscripts
i, o = indicate inlet and outlet, respectively
1,2 = indicate operational conditions
Acknowledgments
The authors gratefully acknowledge Petrobras S.A. for the financial
support as well as the leave of absence granted to MSc. Gilmar
Amaral during the period the research was carried out at LabPetroUnicamp. The authors gratefully acknowledge Schlumberger-Reda
for the ESPs, skid, and torque-sensor donation.
References
Akhras, A., El Hajem, M., Champagne, J.-Y., and Morel, R. 2004. The
flow rate influence on the interaction of a radial pump impeller and the
diffuser. Int. J. of Rotating Machinery 10 (4): 309317. DOI:10.1155/
S1023621X04000326.
Amaral, G. and Franca, F.A. 2006. An experimental investigation of ESP
performance with viscous fluids (in Portuguese). Proc., SPE/IBP

310

Workshop On Artificial-Lift Heavy Oil Offshore, Buzios City, Rio de


Janeiro, Brazil, 2830 May.
Estevam, V. 2002. Uma Analise Fenomenologica da Operacao de Bomba
Centrfuga com Escoamento Bifasico (A Mechanistic Approach on
Gas-Liquid Flow Centrifugal Pump Performance Prediction). PhD
thesis, State University of Campinas, Campinas, Brazil (October 2002).
Gulich, J.F. 1999. Pumping highly viscous fluids with centrifugal pump.
World Pumps: 395/6 Aug/Sept 1999.
Hydraulic Institute Standards for Centrifugal, Rotary & Reciprocating
Pumps, fourteenth edition. 1983. Parsippany, NJ: Hydraulic Institute.
Li, W.G. 2002. Experimental investigation of performance of commercial
centrifugal oil pump. World Pumps: 26-28, February 2002.
Paciga, A. 1967. Projektovanie Zariadeni Cerpacej Techniky. Bratislava,
Slovakia: Slovenske Vydavatelstvo Technickej Literatury.
Patterson, J.C., Henry, J., and Dinkins, W. 2002. Emulsion viscosity testing with ESPs. Proc., Electrical Submersible Pump Workshop, Houston, 13 May.
Stepanoff, A.J. 1957. Centrifugal end Axial Flow PumpsTheory, Design
and Application, second edition. New York: John Wiley & Sons.
Sun, D. 2002. Modeling gas-liquid head performance of electric submersible pumps. PhD dissertation, University of Tulsa, Tulsa, Oklahoma
(April 2003).
Turzo, Z., Takacs, G., and Zsuga, J. 2000. Equations correct centrifugal
pump curves for viscosity. Oil & Gas Journal 98 (22): 57.

SI Metric Conversion Factors


atm  1.013 250*
cp  1.0*

F ( F 32)/1.8
ft  3.048*
m3/s  4.416 314

E + 05 = Pa
E 03 = Pas
= C
E 01 = m
E 05 = bbl/hr

*Conversion factor is exact.

Gilmar Amaral is a petroleum engineer at Petrobras, now


working on artificial-lift projects with ESPs in deepwater fields.
He holds a BS degree in mechanical engineering from the
Federal University of Rio de Janeiro and an MSc degree
in petroleum engineering from the University of Campinas.
Dr. Valdir Estevam is a senior consultant at Petrobras working
on R&D projects related to flow separation and artificial lift. He
holds a BS degree in mechanical engineering from the Federal University of Minas Gerais, an MSc degree in mechanical
engineering from the Federal University of Ouro Preto, and a
PhD degree in petroleum engineering from the University of
Campinas-Unicamp. Fernando A. Franca is a professor of
petroleum engineering in the Department of Petroleum Engineering at the State University of Campinas, Sao Paulo, Brazil.
He holds a BS degree in mechanical engineering from the
University of Braslia, and MSc and PhD degrees in mechanical
engineering from the University of Campinas-Unicamp, Brazil.
Francas main research interests are in the fields of multiphase
flows, multiphase instrumentation gas/liquid separators, and
multiphase pumping.

May 2009 SPE Production & Operations

Potrebbero piacerti anche