SPE-174937-MS

Optimizing Separation Efficiency of Produced Water Tanks by Installing

CFD Designed Internals
Nikolas Trofaier, VTU; Alexander Weilhartner, OMV; Bernhard Puchner, and Jaroslaw Konieczny, VTU;
Nikolaus Friedl, CAE; Georg Kuemmel, OMV; Oscar Alberto Perez, OMV Petrom; Anne Steinbrugger, VTU;
Richard Markl, CAE

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Houston, Texas, USA, 28 –30 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
Efficient produced water management has been a growing concern in the oil and gas industry over the last
decades. Water cuts have risen and more oil fields are maturing. Therefore growing volumes of produced
water have to be handled every year. The most common disposal option is reinjection into the reservoir.
In order to reinject produced water it has to be treated. Produced water tanks (PWT) are usually applied
as first stage separators in water treatment plants. The oil-water separation efficiency of PWTs can be
improved by optimizing tank internals to reduce the load on downstream treatment units. This study shows
the results of a Computational Fluid Dynamics (CFD) based optimization approach for various water
tanks in mature oil field environments.
The concept centers on the induction of a tangential liquid movement in PWTs to maximize residence
time for a given volume of water and to reduce short-circuiting. In order to show the benefits of such a
setup, an existing tank was upgraded with CFD designed internals. The separation efficiency of the tank
was recorded prior to and after installing the internals. The results showed that the average residence time
could be extended eightfold.
Based on the initial experience and additional simulation work, the internals concept was adapted and
improved for several new tanks. These tanks are currently under construction and first field results will
become available in 2016.

Introduction
Gravity separation is the main principle used to separate oil from water in oil production operations. It
occurs between two liquids with different densities. (McLelland et al. 1998) The density difference causes
the water to settle through and out of the oil, since water droplets are heavier than the volume of oil they
replace. They are dragged down by gravitational force, which, on the other hand, is resisted by a drag
force.
The resulting settling velocity for laminar conditions is calculated by Stoke’s law:
2 SPE-174937-MS

(1)

Dd⫽droplet diameter, ␩⫽dynamic viscosity of oil, ␳w⫽density of water ␳o⫽density of oil;

g⫽gravitational force
As Arnold (2007) explains, this fundamental equation leads to several general conclusions for the
design of a separation process:
● Water droplets with larger diameters will settle faster, resulting in shorter treatment time.
● The greater the difference in density between water and oil, the faster the settling velocity.
● Higher temperatures result in lower oil viscosities, leading to higher downward velocities of water
droplets.
As larger diameter droplets of the light fraction (oil) will separate faster from the heavier fraction
(water) than smaller droplets, mixing effects should be avoided in any separation process. Mixing
increases the dispersion of oil droplets in the water phase and results in reduced droplet sizes. Conse-
quently oil -water separation can be improved by reducing mixing effects and ensuring sufficient time for
gravity separation to come into effect. Oil droplets will coagulate to bigger droplets and separate from the
water phase when sufficient time is given for the separation process. Therefore, oil- water separation
equipment that is based on gravity separation should be designed to ensure low speed laminar flow and
maximum residence time.
Pilot Tank Design
These principles were taken into consideration for the upgrade of a pilot tank (total volume: 3750 m3, flow
rate 900 m3/h) which serves as primary water treatment stage and buffer volume in an oil production park
in Matzen oil field in Austria. Upstream facilities collect the three-phase flow from several oil fields and
treat it in conventional three-phase separators. Produced water from the separators is gathered in the pilot
tank and sent downstream to a water treatment plant (WTP). Increased production and more stringent inlet
water quality criteria at the WTP required higher separation efficiencies at the production park. The
original setup of the tank internals is shown in Figure 1.

Figure 1—Original arrangement of internals (Angst 2015)

SPE-174937-MS 3

A single inlet pipe and two separate outlet pipes are installed in the bottom section of the tank. The
actual flow pattern could not be determined. The oil in water (OIW) concentrations, however, measured
at the inlet and the outlet of the tank suggested a non-ideal flow distribution and severe short-circuiting.
Thus, in order to obtain a better understanding of the hydraulic characteristics, a tracer test was performed
for the PWT. The concentration of tracer in the water phase leaving the vessel was measured photomet-
rically as function of time, resulting in the residence time distribution (RTD) curve. This method has been
used in the industry for decades and has helped to better understand the hydraulic behavior of various
separation vessels (Zemel et Bowman, 1978).
Figure 2 shows the results of the tracer test with the peak outlet concentration occurring after
approximately five minutes, providing insufficient time for oil-water gravity separation. Consequently the
separation performance of the pilot tank had to be improved for the altered operating requirements.

Figure 2—RTD curve prior to internals installation (Angst 2015)

The goal of the new concept was to approximate the theoretical liquid retention time of the tank (four
hours) by maximizing the average residence time of oil droplets in the water phase. This optimization
approach was done by using a CFD model of the tank to test variuous tank internals designs and
arrangements. Simulation results suggested that oil and water droplets have the longest residence time in
a circular tank when they travel along a tangential flow line from inlet to outlet.
Various internals designs and arrangements were simulated. The best results for the given tank nozzles
were obtained with an inlet diverter close to the tank wall and two collecting pipes for the water in the
center of the tank bottom (Figure 3a). The inlet diverter uses parallel plates to decelerate incoming liquid
and inject a low speed, laminar flow into the tank (outlet velocity approximately 5 cm/s).
4 SPE-174937-MS

Figure 3a and 3b—New concept with single inlet diverter (Angst 2015) and resulting CFD velocity pattern

With this setup a circular flow pattern could be achieved as shown inFigure 3b. The outlet pipes that
were finally installed used two collectors with hole patterns on the lower half instead of simple open pipes.
The hole pattern helped to further reduce local velocities and avoid vortex effects at the pipe inlets. Figure
4 shows the inlet diverter and the water outlet pipes fully installed in the pilot tank.

Figure 4 —Installed tank internals in completed pilot tank

Once the plant was completed and operations started a performance test was carried out for the PWT
to obtain the RTD of the new tank concept. In addition the tracer test was also simulated using the CFD
model by inserting mass-free liquid droplets with uniform diameters. The results of both tests are
summarized in the graph below (Figure 5). Compared to the old internals, the new arrangement showed
a much slower increase of tracer concentration on the outlet. The significantly lower peak concentration
was reached after approximately 40 minutes. The simulated tracer test results showed similar hydraulic
characteristics with a slow incline in outlet concentration and a peak at 50 minutes. Thus, the new tank
SPE-174937-MS 5

internals extended the average residence time of a liquid droplet by a factor of eight. With this
improvement the PWT could reduce the OIW content from approximately 1000 ppm to approximately
250 ppm.

Figure 5—RTD after internals installation (Angst 2015)

Concept Improvements
Based on the experience from the pilot tank in Matzen, new PWTs at the Videle oil field in Romania had
to be optimized to reduce the overall number of equipments, treatment steps and investment costs. CFD
simulations had shown that circular flow patterns in separation tanks are an efficient way to extend
average residence times and reduce short-circuiting. Nevertheless, Figure 3b also showed that the single
inlet device caused an assymetrical flow pattern creating undesirable mixing and vortex effects and even
short-circuits towards the outlet collectors. Thus, the concept had to be improved and the assymetric
characteristic was elimanted by adding an identical inlet diverter on the opposite side of the tank (Figure
6). Both diverters injected produced water in the same flow direction, creating a more uniform flow
pattern. In addition, the outlet collector pipes were replaced by a central outlet cylinder with a hole pattern
in the top plate.
6 SPE-174937-MS

Figure 6 —Internals arrangement for improved concept

CFD simulation
To predict the impact of the new setup, the existing CFD model was adapted, The model had been verified
with field data and had to be adjusted to the new process and equipment parameters.
Figure 7a shows the CFD results for the side and front view of the improved inlet diverters with the
reduced flow velocities at the plate outlets. The new diverter design was slightly changed from the pilot
tank which had used a tilted feed pipe for the diverter. The new design used a vertical inlet pipe where
water flowed through the vertical section of the pipe into the box section of the inlet distributor. Further
downwards, diverted plates gradually narrowed the width of vertical channel, cutting away a certain
percentage of the flow and directing the water into the horizontal channels. These channels are spaced to
distribute the entire volume flow homogenously and slow the velocities to a few millimeters per second
(Figure 7a).

Figure 7a and 7b—Velocity profile inside inlet diverter and outlet cylinder
SPE-174937-MS 7

By using a cylindrical outlet collector (Figure 7b) instead of straight pipes for the water outlet, some
of the deficiencies of the collector pipes could be eliminated. The outlet cylinder better matched the
induced spiral flow pattern. It allowed liquids to leave the tank only once they had travelled all the way
to the center. Thus, central volumes of the tank between the inlet and outlet internals could still be used
for phase separation without having liquids sucked into the collector pipes (Figure 8). Small outlet
openings would lead to higher local velocities which could suck oil droplets to the water outlet. Larger
outlet surfaces on the other hand reduced local velocities but also removed water at larger radial distance
from the center.

Figure 8 —CFD side view of new internals concept with vertical velocities

The top view CFD image of the new arrangement (Figure 9) showed the difference to the assymetrical
concept in the pilot tank (Figure 3b). The pattern was symmetrical with the flow lines spiraling to the
center of the tank. Most areas in the tank showed local velocities of one to two centimeters per second.
Only at the inlet diverters and above the outlet cylinder could five centimeters per second be observed.
8 SPE-174937-MS

Figure 9 —Velocity flow pattern for improved internals concept

Residence Time Distribution

In order to compare the theoretical performance of the new tank internals to the results of the pilot tank,
a digital tracer test was created using mass-free Fluorescein. The first signal at the tank outlet was
registered only after 1500 seconds (25min). The peak concentration could be seen after 8000 seconds
(133min). At this time approximately 15% uranin had left the tank. Thus the average residence time in the
actual tank could be expected at around 2 hours.

Figure 10 —RTD for new tank internals with various outlet cylinder designs

Simulating Separation Efficiency

Tracer test results from the pilot tank simulation showed sufficient consistency with field tests. Thus, the
CFD model was applied to the new PWT to predict the hydraulic performance. Even though the hydraulics
SPE-174937-MS 9

for a given tank might be predictable, the actual separation performance can only be estimated if real oil
droplets with a different density and varying droplet diameters are inserted in the model. In order to gain
theoretical results prior to the tank installation, this evaluation was performed using a method called
particle tracking. First, to predict the separation performance, different droplet size classes from 10 to 100
micrometers in 5 micron increments were created. Then 100 droplets for each class were injected
distributed via both inlet diverters. The three dimensional motion of each particle was tracked until it
either fully separated from the water phase and was trapped in the oil film at surface level or left the tank
through the outlet cylinder.
Simulation parameters of both liquid phases used for the analysis are listed Table 1 and Table 2. The
model only considered the behavior between oil and water phases and did not take any coagulation effects
between oil droplets into consideration.

Table 1—Water parameters for CFD model

Parameter Value

Density 1017 kg/m3

Volume flow 135 m3/h per inlet diverter
Dynamic viscosity 0,0012 Pa s
CFD-Simulation Pressure at outlet velocity dependent (zeta ⫽ 250)
Turbulence model realizable k-␧

Table 2—Oil parameters for CFD model

Parameter Value

Density 947 kg/m3

Kinematic viscosity 1665,8 ⫻ 10⫺6 m2/s
Number of particle per diameter 100
Diameters 10 up to 100 micrometer in
increments of 5 micrometers

Figure 11 shows the results of the oil droplet tracking for two different inlet diverter geometries. The
model predicted good separation efficiency for droplets with larger diameters (⬎ 40 ␮m). Over 90% of
oil droplets bigger than 80 ␮m are trapped or separated from the water phase. On the other hand 90% of
the droplets with diameters smaller than 30 ␮m were sucked with the water phase to the outlet cylinder.
This means that the actual separation efficiency mostly depends on the droplet size distribution of oil in
the incoming produced water stream. The distribution is hard to determine since it depends largely on
installed upstream equipment in the actual plant. Pumps, valves and flow elements such as baffle plates
all act as emulsion formers and thus impact the OIW dispersion.
10 SPE-174937-MS

Figure 11—Separation efficiency for two different diverter geometries

In order to test the behavior of the CFD model, existing oil droplet distribution analyses (Table 3) were
taken from similar oil fields and 100 droplets for each distribution were injected.

Table 3—Droplet size distributions for CFD model

Droplet size [␮m] Concentration fine [%] Concentration medium [%] Concentration coarse [%]

0 0.0000 0.0000 0.0000

10 73.8607 4.4544 0.2341
15 11.1316 6.7700 1.2865
20 5.5531 8.9942 2.5537
25 3.3798 9.7886 3.7680
30 2.2196 10.0340 4.3323
35 1.6543 12.5890 4.6150
40 0.7192 15.8880 5.1719
45 0.5243 11.5000 6.4490
50 0.4678 6.7770 7.4295
55 0.1727 4.3300 8.4560
60 0.1482 2.2220 9.3334
65 0.0792 2.1110 10.1390
70 0.0441 1.0144 10.8740
75 0.0254 0.8970 11.7840
80 0.0053 0.8100 9.0456
85 0.0093 0.7660 3.4456
90 0.0027 0.6022 0.6022
95 0.0013 0.3300 0.3580
100 0.0013 0.1222 0.1222

The final simulation step was to test the potential of predicting outlet concentrations with the CFD
models. Therefore an arbitrary OIW concentration of 155 mg/l with the same droplet size distributions as
before was injected at the tank inlet. The model was then used to record the separated oil concentration
and the concentration of oil in the water outlet. Figure 12 shows the oil concentration that is separated for
each of the droplet size distribution for two different skimmer systems. Finer distributions were poorly
SPE-174937-MS 11

separated since the residence time was too short and oil droplets remained entrained in the water phase.
Compared with the pilot tank results, these separation rates seemed to be realistic. However, a thorough
evaluation can only be performed once the tanks will be completed.

Figure 12—Separated oil flow rate considering the oil droplet distribution, inlet: 155 mg/l

Summary
Produced water management is a crucial topic in the oil and gas industry and water disposal is mostly done
be reinjection into the reservoir. To reinject produced water it needs to be treated in dedicated water
treatment plants which usually apply PWTs a first stage separators. This paper shows how CFD designed
tank internals can be used to optimize the oil-water separation efficiency of PWTs. The internals can
improve the RTD avoid short-circuiting and thereby help to reduce the overall number of equipments,
treatment steps and investment costs. by optimizing tank internals to reduce the load on downstream
treatment units.
The general concept centers on the induction of a tangential liquid movement in PWTs to maximize
residence time for a given volume of water and to reduce short-circuiting. In order to show the benefits
of such a setup, an existing tank was upgraded with CFD designed internals. The separation efficiency of
the tank was recorded prior to and after installing the internals. The results showed that the average
residence time could be extended eightfold.
Based on the initial experience and additional simulation work, the internals concept was adapted and
improved for several new tanks. These tanks are currently under construction and first field results will
become available in 2016.

References
McLelland G., Garcia M.: Investigation into Oil/Water Separation Problems Utilizing RTD Testing
and Fluid Properties. SPE 48994. 1998.
Arnold E. Kenneth: Facilities and Construction Engineering Vol III. SPE. 2007.
Arnold E. Kenneth: Surface Production Operations. 2007.
Angst J.: Fluid Dynamic Simulation of a Water Tank at GOSP Matzen OMV AUT2015.
Zemel B., Bowman R. W.: Residence Time Distribution in Gravity Oil-Water Separations. JPT 1978.