Permeate Recovery Rate Optimization at the Alicante Spain SWRO Plant
Authors:
Richard Stover, 1 Antonio Ordonez Fernandez, 2 Joan Galtes 3
Presenter: Richard Stover1
Abstract
The permeate recovery rate in a reverse osmosis (RO) process is generally defined as the permeate flow
rate divided by the membrane feed flow rate. An alternative definition is the permeate flow rate divided
by the process feed flow rate. Historically, the process and membrane feed flow rates have been equal.
A high recovery rate means a high process yield. However, in a desalination process, operation at high
recovery results in higher average concentrate salinities in the membrane elements, higher osmotic
pressures and higher membrane feed pressures compared to operation at low recovery. In addition,
supersaturation of the concentrate can result in more scaling and high membrane flux can result in more
fouling. On the other hand, low recovery rate operation directly reduces process yield and can result in
excess pretreatment and supply-pumping expenses. Permeate recovery rate optimization, therefore, is a
critical exercise for RO process design and operation.
In most seawater RO processes being built today, such as the seawater RO plant built and operated by
Inima (Grupo OHL) in Alicante Spain, isobaric energy recovery devices (ERDs) are applied to save
energy. The flow rates of the high- and low-pressure streams through the devices to be unequal or
unbalanced. Earlier turbine-based ERDs did not allow this flexibility. As a result, permeate recovery rate
and process recovery rate can be set separately in RO processes equipped with isobaric ERDs. This
feature adds a degree of freedom to recovery rate optimization and an opportunity for reducing energy
consumption and/or improving process yield.
The authors present a detailed consideration of permeate recovery rate optimization in seawater RO
processes equipped with centrifugal high-pressure pumps and PX Pressure Exchanger energy recovery
devices. Optimization models are developed using practical process controls as independent variables.
Modeling results are verified with process data collected at the Alicante seawater RO plant.
1
Energy Recovery, Inc., USA, stover@[Link]
2
Inima (Grupo OHL), Spain, ordonez@[Link]
3
Energy Recovery, Inc., Spain, jgaltes@[Link]
International Desalination Association World Congress: Dubai 2009 - DB09-083
I. INTRODUCTION
In a reverse osmosis desalination process, the recovery rate or conversion rate is the ratio of membrane
permeate flow rate to the membrane feed flow rate. Historically, these processes were operated at the
highest possible recovery rate to obtain the maximum amount of permeate possible from the pressurized
membrane feed water. However, the introduction of “isobaric” or pressure-equalizing energy recovery
devices (ERDs) changed this practice (1).
In a seawater reverse osmosis (SWRO) process operating at a 45% recovery rate, isobaric ERDs supply
55% of the membrane feed flow, reducing the load on the high-pressure pump by a corresponding
amount. Energy is consumed by a circulation pump that works in series with the ERDs, however,
because the circulation pump merely circulates and does not pressurize water, its energy consumption is
minimal. Therefore, more than half of the membrane feed flow is pressurized with almost no energy
input. This means that seawater RO processes with isobaric ERDs can operate affordably at lower
permeate recovery rates compared to processes operating with no energy recovery devices or with
turbine-based devices. As a result, in most seawater RO processes being built today, such as the SWRO
plant built and operated by Inima in Alicante, Spain (“the Alicante plant”), isobaric ERDs and a
recovery rate of between 40% and 45% are applied to save energy.
Isobaric ERDs allow the flow rates of the high- and low-pressure streams through the devices to be
unequal or unbalanced. As a result, the membrane recovery rate and the overall process recovery rate
can be adjusted independently. This feature adds a degree of freedom that can result in further
reductions in energy consumption and/or process yield improvement. This paper considers recovery rate
optimization for the Alicante plant.
II. ALICANTE PLANT OVERVIEW
Alicante is a city of approximately 350,000 people located in southeastern Spain south of Valencia. The
desalination plant is located on the coast just south of the city. Alicante II was commissioned in April
2008. It is the second membrane desalination facility built on the site, the first having been put in
operation in 2003. Alicante II was designed and built by Spanish original equipment manufacturer Inima
of Grupo OHL, Construcciones Alpi and Sampol..
The plant is fed from beach wells. The high-pressure portion of the plant consists of seven independent
SWRO trains with a combined permeate production capacity of 65,000 m3/day. Each train has 128
vessels of Dow Filmtec membranes, with six SW30HRLE-400i elements plus one SW30XLE-400i
element per vessel. The trains are fed with Flowserve 8 x 10 x 13 DMX axially-split double-volute
pumps driven by a Siemens 1100 kW, 2,980 rpm, 6000 volt motors. Each train is also equipped with a
Flowserve 8HHPX15C horizontal circulation/booster pump with a 90 kW, maximum maximum 1,475
rpm, 400 volt motor equipped with a variable speed drive (VSD). A photograph of one of the SWRO
trains is given in Figure 1.
International Desalination Association World Congress: Dubai 2009 - DB09-083
Figure 1 – High Pressure Pump, Circulation Pump and Membrane Array
Energy recovery is achieved with arrays of twelve ERI PX-220 devices dedicated to each SWRO train.
In Alicante, the PX devices are installed in the piping run below the membrane arrays, as shown in
Figure 2.
Figure 2 – PX Energy Recovery Device Array
PX energy recovery devices (ERDs) are positive displacement isobaric devices commonly used in
SWRO processes built since 2003 (2). Pressure transfer occurs through direct contact between the high-
pressure concentrate and pressurized seawater inside the devices. Because there are no pistons or
International Desalination Association World Congress: Dubai 2009 - DB09-083
barriers in the flow paths, high- and low-pressure flow rates through the devices can be manipulated
freely.
The Alicante plant started production in April 2008. The plant was designed to operate at a 45%
recovery rate with sufficient capacity in the energy recovery device arrays and circulation pumps to
operate at a recovery rate of as low as 39% if desired. At 45% recovery, each train is designed to
produce 411 m3/hr of permeate for a nominal plant production capacity of 65,000 m3/day.
Shortly after startup, fouling struck the membranes. This increased the membrane feed pressure by about
4 bar and decreased permeate production. When it was realized that the fouling was persistent and
unavoidable, the recovery rate of the SWRO process was lowered to approximately 40%. Lowering
recovery lowered the membrane feed pressure resulting in an increase in permeate production to the
design flow rate.
After their success with recovery adjustment, the plant operations team was open to consider further
optimization of SWRO system flows. A recovery optimization model was developed and run over a
range of process conditions. The model results were verified with tests run on the system. A detailed
description of the analysis and results is given in the following sections.
III. RECOVERY OPTIMIZATION
To explain recovery and how it is adjusted, a simplified process flow diagram is shown in Figure 3.
Concentrate rejected by the membranes (stream G) flows to the ERDs, driven by a circulation pump.
The ERDs replace the concentrate with feedwater from the low-pressure supply system (streams A and
B). The pressurized feedwater (stream D) merges with the discharge of the high-pressure pump (stream
C) to feed the membranes (stream E). Water leaves the process as permeate from the membranes (stream
F) or as spent low-pressure concentrate from the ERDs (stream H). In these systems, the high-pressure
pump flow rate equals the permeate flow rate plus the leakage loss through the ERD. The leakage loss is
very small such that the permeate flow rate and the high-pressure pump flow rate are always nearly
equal (2).
C E F
HIGH CIRCULATION R.O.
PRESSURE PUMP MEMBRANES
PUMP
G
D
A B H
ISOBARIC ERD
Figure 3: Simplified Diagram of an RO Process with Isobaric ERDs
With reference to Figure 3, the following terms are defined:
Membrane Recovery Rate – Permeate flow rate divided by the membrane feed flow rate or F / E.
Overall Recovery Rate – Permeate flow rate divided by the system feed flow rate or F / A.
International Desalination Association World Congress: Dubai 2009 - DB09-083
Balanced ERD Flows – Flow rate of low pressure water fed to the ERD equals flow rate of high
pressure water taken from the ERD or B = D and G = H. At balanced flow, membrane recovery
and overall recovery are equal.
Lead Flow – Ratio of low-pressure flow rate to ERD divided by high-pressure flow rate from
ERD, set by adjustment of low-pressure flow rate. A positive lead flow occurs when B > D.
Lag Flow – Ratio of high-pressure flow from ERD divided by low-pressure rate to ERD, set by
adjustment of high-pressure flow rate. A positive lag flow occurs when D > B.
The membrane recovery rate, also known as the conversion rate, quantifies the amount of permeate
extracted from the membrane feed. This, in turn, determines the concentration of the dissolved solids in
the membrane reject stream. At high membrane recovery rates, the osmotic pressure of the concentrate
stream is high, resulting in a high membrane feed pressure. Membrane recovery rate can be manipulated
by altering the circulation pump speed. This, in turn, alters the membrane feed flow rate and changes the
denominator in equation that defines the membrane recovery rate.
Lead or lag flow can be imposed by adjusting the low-pressure flow rate through the ERDs or by
adjusting the speed of the circulation pump, respectively. However, for the sake of clarity in this
analysis, the term lead flow will be used to refer to adjustments made by changing just the low-pressure
flow rate through the ERDs. Therefore, positive or negative lead flows will be considered. Lag flow will
be used throughout this analysis to refer to flow adjustments made by changing just the circulation pump
speed. Positive and negative lag flows will be considered.
Some mixing occurs in the ERDs as a result of the direct contact between seawater and concentrate
inside the devices. The ratio of ERD flows has a distinct affect on mixing as illustrated in Figure 4. If
operated at positive lead flow, the excess seawater fed to the ERDs flushes the devices and reduces the
salinity of the high-pressure water flowing from the devices. Negative lead flow results in some
breakthrough of the concentrate to the high-pressure water flowing from the devices, evident as an
increase in mixing. Lag flow has a similar affect on ERD mixing with a negative lag flow resulting in
reduced mixing. Reduced mixing, in turn, results in lower salinity in the membrane elements and a
corresponding reduction in the osmotic and membrane feed pressures.
15%
ERD Volumetric Mixing
10%
5%
0%
-10% -5% 0% 5% 10% 15%
ERD Lead Flow
Figure 4 – ERD Mixing Versus Lead Flow
International Desalination Association World Congress: Dubai 2009 - DB09-083
IV. PROCESS MODEL
The Alicante plant SWRO process was modeled using the characteristic curves of the high-pressure
pump, the circulation pump, the membranes and the ERDs. Figure 5 shows the high-pressure pump
curve, with head plotted as a function of flow. Adjustment of the high-pressure pump feed pressure,
high-pressure throttling or permeate throttling can shift the pump curve up or down the chart. The duty
point of the pump, however, always stays on the curve. For example, higher membrane feed pressure
results in a lower flow rate as the duty point moves to the right and down the chart.
1000
900
Membranes
800 High Pressure Pump 45% Recovery
700 41%
37%
600
Head (m)
500
400
300
200
100
0
0 50 100 150 200 250 300 350 400 450 500 550
Flow (m3/hr)
Figure 5 – High-Pressure Pump and Membrane Characteristic Curves
Membrane responses from the projection software for three different recovery rates were superimposed
on the high-pressure pump curve in Figure 5. It is valid to consider these curves on the same chart
because the high-pressure pump flow rate and the permeate rate are always nearly equal as described
above. The membrane curves indicate that increased feed pressure results in higher permeate flow rates.
These curves also indicate that higher pressures are required at higher recovery rates. Seawater
temperature, seawater salinity and membrane fouling shift the membrane curves up or down the chart.
The process operates at the pressure and flow rate where the membrane curve and the pump curve
intersect.
Two characteristic curves for the circulation pump are given in Figure 6, corresponding to two different
pump and motor rotation speeds. Although the duty point of the circulation pump always stays on a
flow-head curve, adjustment of the VSD allows the operator to shift pump duty point from one curve to
another. Therefore, the circulation pump can essentially be operated at any combination of flows and
pressures within the operating envelope provided by the pump, motor and VSD. The circulation pump
drives flow through the membrane concentrate channels and the ERDs such that the operation of these
elements is coupled.
International Desalination Association World Congress: Dubai 2009 - DB09-083
80
70
60
50
Head (m)
40
30 1475 rpm
20
1280 rpm
10
0
0 100 200 300 400 500 600 700 800 900
Flow (m3/hr)
Figure 6 – Circulation Pump Characteristic Curves
The performance of the ERDs is also described by characteristic curves. These are shown in Figure 7 for
a model PX-220 device. The flow rates through the ERD determine the pressure drops along the high-
pressure and low-pressure flow paths. Conversely, with multiple ERDs operating in a device array, the
flow rate through a particular device is determined by the pressure difference between the inlet and
outlet manifolds at the manifold positions of the ERD (3). The ERD high-pressure flow is driven by the
circulation pump and the low-pressure flow by the low-pressure supply pump.
DIFFERENTIAL PRESSURE
m3/hr
30 35 40 45 50
1.2
15 1.0
Pressure
0.8
10
HP DP 0.6
0.4
5
LP DP 0.2
psi bar
0 0.0
130 gpm 150 170 190 210 230
Flow Rate
Figure 7 – PX-220 ERD Characteristic Curves
Resolution of the SWRO operating point for a given set of flow, pressure, salinity and temperature
conditions involves the simultaneous consideration of the characteristic curves of the pumps,
membranes and ERD. It is, therefore, an iterative computational process. After the system flows,
pressures and corresponding pump hydraulic output requirements are determined, the energy
consumption of the pump motors are computed using the pump and motor efficiencies (4, 5).
4.1 Lead Flow Modeling Results
Modeling results for a range of system recovery and membrane recovery combinations under lead flow
conditions are shown in Figure 8. Specific energy in Figure 8 is the sum of the supply pump, the
circulation pump and the high-pressure pump energy consumption divided by the SWRO permeate flow
International Desalination Association World Congress: Dubai 2009 - DB09-083
rate. Lead flow and overall recovery were adjusted by changing the low-pressure flow rate through the
ERDs while holding constant the high-pressure flow rate of the circulation pump.
2.90
Specific Energy (kWh/m )
3
2.88 43%
membrane
39%
2.86 recovery
2.84
41%
2.82
2.80
36% 38% 40% 42% 44%
Overall Recovery
Figure 8 – Lead Flow Modeling Results
Considering any of the membrane recovery curves shown in Figure 8, as overall recovery is increased
with reduced seawater fed to the ERDs, energy consumption by the high-pressure pump increases. This
energy consumption increase corresponds with the increase in salinity in the membrane feed caused by
extra mixing in the ERD in accordance with Figure 4. As overall recovery is reduced with extra seawater
fed to the ERDs, excess energy is consumed by the system supply pump. The optimum overall recovery
rate corresponding with the lowest SWRO specific energy consumption, therefore, is achieved by
optimizing the low-pressure flow rate supplied to the ERDs by the supply pump.
At 41% membrane recovery, minimum energy consumption is predicted to occur at an overall recovery
rate of nearly one percent less than the membrane recovery rate, corresponding to positive lead flow. At
43% membrane recovery, the optimal low-energy point is at an overall recovery rate that is about 1.5%
lower than the membrane recovery, corresponding to more positive lead flow. At 39% membrane
recovery, the optimal overall recovery is above 39%, corresponding with slight negative lead flow.
These data suggest that the system has a “sweet spot” close to 41.5% membrane recovery and 41%
overall recovery. These recovery rates are close to balanced flow with 2% extra low-pressure seawater
supply or 2% lower concentrate flow. If the membrane recovery is increased above 41.5% by reducing
circulation pump speed, specific energy consumption can be reduced by reducing overall recovery by
applying more low-pressure flow to the ERDs. Similarly, if the system is operating at a lower-than-
optimal membrane recovery, overall recovery can be increased slightly by reducing low-pressure flow to
the ERDs to reduce energy consumption.
It is important to note that this analysis does not take into account the full cost of pretreatment, the
capacity of the pretreatment system or the energy or cost required for post-treatment. Recovery
optimization requires consideration of equipment, contractual and cost constraints. For example, there is
insufficient pretreatment capacity to allow the entire plant to operate at 41% recovery. However, the
analysis does take into account changes in mixing through the ERDs and the associated impact upon
International Desalination Association World Congress: Dubai 2009 - DB09-083
membrane feed pressure. Volumetric mixing ranges from 2 to 13% in the data presented in Figure 8 in
accordance with the lead flow dependency given in Figure 4.
4.2 Lag Flow Modeling Results
A similar iterative procedure was used to resolve the characteristic equations of the components to
generate specific energy curves for a range of overall and membrane recovery rates for lag flow
conditions. Lag flow and membrane recovery were adjusted by just changing circulation pump speed.
For example, increased circulation pump speed resulted in positive lag flow and reduced membrane
recovery. The results are shown in Figure 9.
2.90
Specific Energy (kWh/m )
43%
3
2.88 overall
39% recovery
2.86
2.84
41%
2.82
2.80
36% 38% 40% 42% 44%
Membrane Recovery
Figure 9 – Lag Flow Modeling Results
The overall trend of the data for the lag flow modeling was similar to the lead flow data. Considering
any of the overall recovery curves, positive lag flow and reduced membrane recovery increased energy
consumption by raising the salinity of the membrane feed and the corresponding duty of the high-
pressure pump. Negative lag flow and increased membrane recovery increased energy consumption by
raising the osmotic pressure within the membrane array. The optimum membrane recovery rate
corresponding with the lowest overall specific energy consumption, therefore, is achieved by optimizing
the high-pressure flow rate supplied to the ERDs by the circulation pump. As in the lead flow analysis,
the “sweet spot” is at about 41.5% membrane recovery and 41% overall recovery.
Note that the overall recovery curves in the lag flow modeling results are more curved than the lead flow
curves when the system is shifted away from its minimum energy points. Circulation pump speed
changes alter both the mixing within the ERD and the salinity within the membrane elements. The
compound affect of membrane recovery rate changes results in the steep curves in Figure 9. These
results show that the process is more sensitive to the high-pressure flow rate through the ERDs than it is
to the low-pressure flow rate through the ERDs.
It should also be noted that the overall range in specific energy indicated in Figures 8 and 9 is about
1.4% despite a variation in ERD flow of nearly 30%. This suggests that the system’s energy
performance is robust and somewhat insensitive to flow imbalance through the ERD. The optimization
International Desalination Association World Congress: Dubai 2009 - DB09-083
exercise described above is “fine tuning.” However, its importance is magnified when multiple years of
operation and multiple SWRO trains are considered. In the Alicante plant, if the price of power is $0.09
per kilowatt hour, a 1.4% reduction in energy consumption would save nearly $90,000 per year in power
costs.
Again, it is important to note that this analysis does not take into account equipment, contractual and
cost constraints which may shift the optimal recovery rate of the process. However, the trend of the
results can be useful for process optimization at any recovery rate. For example, if the process is
operated at an overall recovery rate of 45%, the analysis above suggests that reducing the membrane
recovery rate to below 45% will improve specific energy consumption. This can be achieved by
increasing the speed of the circulation pump with no other changes to the process. Conversely, at overall
recovery rates below 41%, increased membrane recovery rates and lower circulation pump speeds are
energetically favored.
V. PROCESS PERFORMANCE
Between when the analysis above was conducted and process performance data was collected, several
changes were made to the process. First, a partial open intake was added to the beach wells resulting in a
decrease in the average salinity of the feedwater to the process from about 42,300 to about 39,300 ppm
and reducing membrane feed pressure by about 1 bar. Second, the membranes of some SWRO trains
were chemically treated to address the fouling problem resulting in a membrane feed pressure reduction
of approximately 2 bar. Both of these changes reduced the specific energy consumption of the process
compared to the energy consumption measured at the start of the analysis and predicted by the model.
Process data was collected in the Alicante plant from the SCADA screen in the plant control room at 41
and 43.5% overall recovery. The circulation pump speed or the low-pressure feed rate to the PX devices
was varied and pump power consumption was measured. ERD efficiency ranged from 96.6 to 97.2%
and volumetric mixing ranged from 3.6 to 10.6%. Specific energy data as a function of recoveries is
shown in Figure 10. The curve identified as “43.5% overall recovery” was collected on a train that had
received chemical treatment for fouling.
41%
Specific Energy (kWh/m )
Overall
3
2.56
Recovery
2.52
2.48
2.44 43.5%
2.40
2.36
36% 38% 40% 42% 44%
Membrane Recovery
Figure 10 – Process Data
International Desalination Association World Congress: Dubai 2009 - DB09-083
These data clearly demonstrate the energy reduction benefits of the feedwater and membrane changes.
The data also suggest that the lowest specific energy was measured when the membrane recovery and
overall recovery were set equal, at which point the ERDs were operating at balanced flow. However, the
change in specific energy consumption and the ERD efficiency measured over the range of the
membrane recovery rates examined was minimal despite the rather large range of mixing observed from
the ERDs. This observation supports the conclusion that SWRO processes operating with isobaric ERDs
are relatively insensitive to flow and flow ratio variations through the ERDs.
VI. SUMMARY AND CONCLUSIONS
The isobaric energy recovery devices in use in the Alicante plant gave the operators the flexibility to
reduce recovery to achieve the plant’s permeate production target despite persistent fouling that
increased the membrane feed pressure. In addition, the PX devices used for energy recovery in the plant
allow the operators to adjust the ratio of high- and low-pressure flows through the devices and thereby
independently adjust the membrane recovery rate and overall recovery rate of the process. An analysis
of the system over a range of recovery rates reveals the existence of a specific combination of overall
and membrane recovery rates that results in minimum energy consumption. The analysis indicates how
the membrane recovery rate could be adjusted to minimize energy consumption at whatever overall
recovery rate that plant is operating. The results of the analysis were corroborated with process data.
VII. REFERENCES
1. MacHarg, J.P. and G.G. Pique, How to Design and Operate SWRO Systems Built Around a New
Pressure Exchanger Device, International Desalination Association World Congress Manama,
Bahrain 2002.
2. Stover, R.L., Development of a Fourth Generation Energy Recovery Device – A CTO’s Notebook,
Desalination, 165, pp. 313-321, August 2004.
3. Stover, R.L., J.G. Martin and M. Nelson, The 200,000 m3/day Hamma Seawater Desalination Plant,
Proceedings of the International Desalination Association World Congress, Maspalomas, Gran
Canaria, Spain, October 2007.
4. Moch, I. and C. Harris, What Seawater Energy Recovery System Should I Use? – A Modern
Comparative Study, Proceedings of the International Desalination Association World Congress,
Manama, Bahrain, March 2002.
5. Stover, R.L., Energy Recovery Device Performance Analysis, Proceedings of the Water Middle East
Conference, Bahrain, November 2005.
International Desalination Association World Congress: Dubai 2009 - DB09-083