Experimental Characterization of Membrane Fouling under Intermittent Operation and Its Application to the
Optimization of Solar Photovoltaic Powered Reverse Osmosis Drinking Water Treatment Systems
by
Marina Freire-Gormaly
A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy
Department of Mechanical and Industrial Engineering University of Toronto
© Copyright by Marina Freire-Gormaly 2018
ii
Experimental Characterization of Membrane Fouling under
Intermittent Operation and Its Application to the Optimization of
Solar Photovoltaic Powered Reverse Osmosis Drinking Water
Treatment Systems
Marina Freire-Gormaly
Doctor of Philosophy
Department of Mechanical and Industrial Engineering
University of Toronto
2018
Abstract
This thesis presents a novel experimental characterization of reverse osmosis membrane
fouling from the intermittent operation of solar powered water treatment systems. This thesis also
depicts the development of an analytical membrane fouling model and a design framework to
configure location-customized solar photovoltaic reverse osmosis systems.
The World Health Organization estimates that 760 million people worldwide lack access
to clean drinking water. The regions with the highest water scarcity are usually off-grid, remote
and have high solar insolation. Therefore, the use of solar powered reverse osmosis water treatment
systems is a viable solution. However, to minimize the costs, these systems are configured with
minimal battery storage and operated intermittently with extended shutdown periods. Literature
lacks an experimental characterization of the effect of this intermittent operation on membrane
fouling and an associated design optimization framework.
iii
This research work on reverse osmosis water treatment systems is divided into two main
parts: (1) the experimental characterization of membrane fouling under intermittent operation, and
(2) the development of an analytical membrane fouling model and a design optimization
framework for these systems.
A new fully-instrumented experimental lab-scale system was designed, built,
commissioned and operated with triplicate measurements of membrane permeability and
membrane salt rejection for the experimental characterization. A new pilot-scale experimental
system was also designed, built and operated. The membrane fouling was characterized
experimentally for intermittent and continuous operation. The effect of anti-scalant and rinsing
was also investigated. Two types of experimental water was tested: an experimental MilliQ-based
matrix and an experimental groundwater-based matrix. The groundwater was from Nobleton,
Ontario. In addition, membrane autopsy was performed using scanning electron microscopy.
An analytical membrane fouling model was developed based on the experimental results.
Furthermore, a novel design framework was developed using this new analytical membrane
fouling model. This design optimization framework can be used for the configuration of
community-specific solar photovoltaic reverse osmosis systems that are reliable throughout the
system life at a minimal cost. The design optimization framework can be adapted for other modular
systems such as renewable power systems for off-grid communities, remote First Nations, Métis,
and Inuit communities, or remote mining sites.
iv
Dedication
I dedicate this work to my friend, Susan Lee, whose enthusiasm for my research, curiosity
and encouragement was invaluable. Although she didn’t see the work to completion, dying of brain
cancer in late 2016, her spirit of generosity and social consciousness continue to inspire me.
I also dedicate this work to my great-aunt Andrea. Although she only saw me start my
doctoral research, and thought it was ‘real swish,’ her kind and friendly nature taught me the
importance of learning from everyone I meet.
v
Acknowledgments
I have been very fortunate to work with Prof. Amy Bilton for this doctoral program. Her
mentorship, leadership and work-ethic has been instrumental. I greatly appreciate her guidance,
assistance troubleshooting the experimental equipment and systematic approach to everything. Her
dedication to global engineering research, teaching and her students is inspirational. Working with
her, helping set-up the lab and seeing it grow over the past years has been an incredible experience.
I also sincerely appreciate the guidance and expertise provided by my committee members,
Dean Cristina Amon and Prof. Robert C. Andrews. Their review, in-depth critique of my research
and discussions over the past years have been incredibly helpful. I also would like to thank Prof.
Robert C. Andrews for lending his RO experimental equipment. This helped me immensely to set-
up the experimental systems in this thesis. Prof. A. H. M. Anwar Sadmani and Dr. Gwen Woods-
Chabane for their helpful guidance explaining the RO experimental equipment and answering any
questions I had during my doctoral studies.
I also greatly appreciate the expertise provided by my external committee members, Prof.
John H. Lienhard and Prof. Murray Metcalfe.
Jeff Vandenberg from the Toronto Regional Conservation Authority graciously taught me
how to use his water collection equipment and allowed me to collect water at the Nobleton, Ontario
deep water well. I also appreciate Prof. Christopher M. Yip’s generosity sharing his fridge for the
groundwater storage. I am grateful to George Kretschmann from Geology, who provided
invaluable technical support on the scanning electron microscopy and electron dispersive
spectroscopy.
The financial support from the Natural Sciences and Engineering Research Council
(NSERC) of Canada for the NSERC Discovery Grant and experimental equipment infrastructure
financially supported through the Canada Foundation for Innovation and the Ontario Research
Fund were invaluable. I am also very thankful for the NSERC CGS-D scholarship, the Department
of Mechanical and Industrial Engineering at the University of Toronto for the fellowship and
Conference Travel Grant, and to the Province of Ontario for the Queen Elizabeth II/Dupont
Graduate Scholarship in Science and Technology, the Hatch Graduate Scholarship for Sustainable
vi
Energy, the Paul Cadario Doctoral Fellowship in Global Engineering, the Mary Gertrude L'Anson
Scholarship and the Ontario Society of Professional Engineers (OSPE) Personal Scholarship.
I had the privilege of working with extremely helpful and dedicated undergraduate
researchers, Youngmok (Justin) Ko, Rashmi Satharakulasinghe, Francis Cruz, Daniel Powell,
Caiden Chi, and an M.Eng. researcher, Ravier Weekes. Their enthusiasm and support made
working in the lab a joy.
The Water and Energy Research Lab was a dynamic, supportive and innovative team. I am
really thankful to Shakya Sur and Ahmed Mahmoud for helping out with driving a van or vehicle
for water collection trips. As well, my trustworthy water moving crew labmates, who never
hesitated to come over and help move containers off the van and into the fridge. The lab group
was a great place of comradery and I appreciate the conversations, discussions, helpful critiques
of my presentations and the endless support of all of my labmates.
I am thankful for the support my mom, dad, sister, Marianne, and brother-in-law, Ray
provided throughout my studies. I also appreciate my friends and extended family for supporting
me and listening about my adventures in the lab. I also appreciate Dr. Faizul Mohee’s
encouragement all along the way. Whether it was listening to my practice presentations, providing
a sounding board for ideas, reading my drafts or encouraging me to take a break once in a while, I
greatly appreciate all of their support.
vii
Table of Contents
Dedication ...................................................................................................................................... iv
Acknowledgments............................................................................................................................v
Table of Contents .......................................................................................................................... vii
List of Tables ............................................................................................................................... xiii
List of Figures .............................................................................................................................. xiv
Chapter 1 ..........................................................................................................................................1
Introduction .................................................................................................................................1
1.1 Motivation ............................................................................................................................1
1.2 Problem Statement ...............................................................................................................4
1.3 Research Objectives and Goals ............................................................................................4
1.4 Research Scope ....................................................................................................................5
1.5 Thesis Contributions ............................................................................................................6
1.6 Research Program ................................................................................................................8
1.7 Thesis Organization .............................................................................................................9
Chapter 2 ........................................................................................................................................11
Background and Literature Review ..........................................................................................11
2.1 Renewable Powered Reverse Osmosis Desalination .........................................................11
2.1.1 Reverse Osmosis Water Treatment ........................................................................12
2.1.2 Overview of Renewable Powered RO Systems .....................................................13
2.2 Reverse Osmosis Membrane Fouling ................................................................................15
2.2.1 Biofouling ..............................................................................................................15
2.2.2 Scaling....................................................................................................................16
2.2.3 Pre-treatment to Minimize RO Membrane Fouling ...............................................16
2.2.4 Experimental Studies on RO Membrane Fouling ..................................................17
2.2.5 RO Membrane Fouling Models and Mechanisms .................................................18
viii
2.3 Design Approaches for Solar Powered RO Systems .........................................................19
2.4 Summary and Research Needs...........................................................................................20
Chapter 3 ........................................................................................................................................22
Experimental Systems and Methods .........................................................................................22
3.1 Introduction ........................................................................................................................22
3.2 Lab-scale Systems ..............................................................................................................22
3.2.1 Initial Lab-scale System Setup...............................................................................22
3.2.2 Improved Lab-scale System Setup.........................................................................25
3.2.3 Lab-scale Instrumentation ......................................................................................28
3.2.4 Experimental Water Tank ......................................................................................31
3.3 Pilot-scale System ..............................................................................................................32
3.3.1 Pilot-scale System Setup ........................................................................................32
3.3.2 Pilot-scale Instrumentation ....................................................................................35
3.4 Experimental Methods .......................................................................................................37
3.4.1 Experiment Water Preparation ...............................................................................37
3.4.2 Operating Conditions .............................................................................................40
3.4.3 Cartridge Filter Replacement .................................................................................43
3.4.4 Lab-scale System Procedures ................................................................................43
3.4.4.1 Membrane Coupon Preparation ...............................................................43
3.4.4.2 Membrane Autopsy .................................................................................44
3.4.5 Pilot-scale System Procedures ...............................................................................45
3.4.5.1 Membrane Preparation ............................................................................45
3.4.5.2 Membrane Autopsy .................................................................................45
3.4.6 Membrane Characterization ...................................................................................46
3.4.6.1 Pure Water Permeability ..........................................................................46
3.4.6.2 Scanning Electron Microscopy ................................................................47
ix
3.4.6.3 ATP Surface Deposit Analysis ................................................................48
3.5 Conclusion .........................................................................................................................48
Chapter 4 ........................................................................................................................................49
Membrane Fouling under Intermittent Operation with the Initial Lab-scale System and
using the Experimental MilliQ-based Matrix ...........................................................................49
4.1 Introduction ........................................................................................................................49
4.2 Experimental Program .......................................................................................................49
4.3 System Operating Conditions ............................................................................................50
4.4 Effect of Intermittent Operation vs. Continuous Operation...............................................50
4.5 Discussion ..........................................................................................................................55
4.6 Conclusions ........................................................................................................................57
Chapter 5 ........................................................................................................................................58
Membrane Fouling Characterization at the Lab-scale using the Experimental MilliQ-based
Matrix ........................................................................................................................................58
5.1 Introduction ........................................................................................................................58
5.2 System Operating Conditions ............................................................................................58
5.3 Experimental Program .......................................................................................................60
5.4 Effect of Anti-scalant Pre-treatment ..................................................................................61
5.5 Effect of Intermittent Operation vs. Continuous Operation...............................................62
5.6 Effect of Anti-scalant F135 for Intermittent Operation with no Rinse ..............................63
5.7 Effect of Rinsing with Anti-scalant F135 for Intermittent Operation................................64
5.8 Combined Comparison of the Effect of Operational Conditions on Membrane
Permeability .......................................................................................................................65
5.9 Combined Comparison of the Effect of Operational Conditions on Salt Rejection ..........67
5.10 Combined Comparison of the Effect of Operational Conditions on the Membrane
Autopsies............................................................................................................................68
5.11 Conclusions ........................................................................................................................71
Chapter 6 ........................................................................................................................................73
x
Membrane Fouling Characterization at the Lab-scale and Pilot-scale using the
Experimental Groundwater-based Matrix .................................................................................73
6.1 Introduction ........................................................................................................................73
6.2 Experimental Hypotheses ..................................................................................................73
6.2.1 Effect of Intermittent vs. Continuous Operation on Membrane Fouling with
Anti-scalant Usage at the Lab-scale .......................................................................73
6.2.2 Effect of Permeate Rinsing on Membrane Fouling with Anti-scalant Usage at
the Lab-scale ..........................................................................................................74
6.2.3 Effect of Experimental System on Membrane Fouling .........................................75
6.3 Experimental Program to Test Hypotheses........................................................................75
6.4 Experimental Results .........................................................................................................76
6.4.1 Lab-scale Membrane Permeability Decline ...........................................................76
6.4.2 Lab-scale Salt Rejection ........................................................................................80
6.4.3 Lab-scale Membrane Autopsy ...............................................................................81
6.4.4 Discussion for the Lab-scale Experimental Results ...............................................84
6.4.4.1 Effect of Intermittent vs. Continuous Operation on Membrane
Fouling with Anti-scalant Usage .............................................................84
6.4.4.2 Effect of Effect of Permeate Rinsing on Membrane Fouling with
Anti-scalant Usage ...................................................................................85
6.4.5 Pilot-scale Membrane Permeability Decline .........................................................86
6.4.6 Pilot-scale Membrane Autopsy ..............................................................................89
6.4.7 Discussion for the Pilot-scale.................................................................................91
6.4.7.1 Effect of Experimental System on Membrane Fouling ...........................91
6.5 Discussion of Potential Fouling Mechanisms ....................................................................92
6.6 Conclusions ........................................................................................................................93
Chapter 7 ........................................................................................................................................95
Design Optimization Framework for Solar Powered Reverse Osmosis Systems
Considering Membrane Fouling from Intermittent Operation ..................................................95
7.1 Introduction ........................................................................................................................95
xi
7.2 Design Optimization Framework Approach ......................................................................96
7.3 Design Optimization Framework .......................................................................................99
7.3.1 Optimization Setup ................................................................................................99
7.3.2 Simulation Model.................................................................................................101
7.3.2.1 Power System Model .............................................................................104
7.3.2.2 Water Treatment System Model ............................................................106
7.3.2.3 Analytical Membrane Fouling Model ...................................................107
7.3.3 Cost Model ...........................................................................................................109
7.3.3.1 Power System Cost Model ....................................................................112
7.3.3.2 Water Treatment System Cost Model ...................................................114
7.4 Optimization Results and Discussion ..............................................................................116
7.4.1 Effect of System Size on Optimal System Cost...................................................116
7.4.2 Effect of System Size on System Configuration..................................................119
7.4.3 Effect of System Reliability on Optimal System Configuration .........................120
7.4.4 Effect of Membrane Permeability Decline on System Reliability .......................121
7.4.5 Effect of Geographic Location on Optimal System Configuration .....................124
7.5 Conclusions ......................................................................................................................127
Chapter 8 ......................................................................................................................................128
Summary and Conclusions ......................................................................................................128
8.1 Summary ..........................................................................................................................128
8.1.1 Initial Experimental Characterization of Intermittent Operation .........................129
8.1.2 Improved Experimental Characterization of Intermittent Operation at the Lab-
scale......................................................................................................................129
8.1.3 Experimental Characterization of Intermittent Operation using Groundwater ....130
8.1.4 Design Optimization of Solar Powered Water Treatment Systems Considering
Membrane Fouling ...............................................................................................131
8.2 Conclusions ......................................................................................................................131
xii
8.3 Recommendations for Future Work.................................................................................134
8.4 List of Published Papers from this Thesis........................................................................136
8.5 List of Conference Proceedings from this Thesis ............................................................136
References ....................................................................................................................................138
xiii
List of Tables
Table 1-1: Water salinity ranges in terms of the total dissolved solids, adapted from [15]. .......... 2
Table 3-1: Water chemistry of La Mancalona, Mexico groundwater, Nobleton, Ontario
groundwater, the experimental MilliQ-based matrix and the experimental groundwater-based
matrix. ........................................................................................................................................... 38
Table 3-2: Filmtec BW30 membrane characteristics. ................................................................... 47
Table 4-1: List of experiments and operating conditions. ............................................................ 50
Table 5-1: List of experiments and operating conditions. ............................................................ 60
Table 6-1: List of experiments and operating conditions. ............................................................ 76
Table 7-1: List of design variables.............................................................................................. 100
Table 7-2: List of genetic algorithm parameters. ........................................................................ 101
Table 7-3: Membrane fouling parameters for cases investigated. .............................................. 108
Table 7-4: Individual component costs for the solar powered water treatment system. ............. 110
Table 7-5: Water cost (USD/m3) for various system sizes (1, 5, 10 m3/day) and LOWP (1%, 5%,
10%). Shows decreasing water costs for reduced system reliability (increasing LOWP). ......... 119
Table 7-6: Effect of changing the design goal LOWP for a 10 m3/day system for La Mancalona,
Mexico. ....................................................................................................................................... 121
Table 7-7: Effect of considering membrane fouling on the optimal system configuration for a
design goal loss of water probability of 1%. ............................................................................... 122
Table 7-8: Effect of geographic location on the optimal system configuration for a design goal
loss of water probability (LOWP) of 1%. For all the shown design configurations the optimal
operating condition was Case 3 (with anti-scalant and with rinsing) and the optimization found
the optimal system cost at the maximum membrane life of five years. ...................................... 126
xiv
List of Figures
Figure 1-1: The world map of a) water scarcity [1], and b) average solar irradiance [24]. Regions
with high water scarcity also tend to have high solar irradiance. ................................................... 3
Figure 1-2: Flowchart of the research program conducted in this thesis with the associated
chapter numbers. ............................................................................................................................. 9
Figure 2-1: Reverse osmosis membrane resistance as a function of time. ................................... 13
Figure 2-2: Simple PVRO system schematic for remote communities. ....................................... 14
Figure 3-1: Initial experimental lab-scale system schematic including instrumentation. ............. 24
Figure 3-2: Physical initial experimental lab-scale setup corresponding to the lab-scale schematic
in Figure 3-1. ................................................................................................................................. 25
Figure 3-3: Improved experimental lab-scale system schematic with instrumentation. ............... 26
Figure 3-4: Physical experimental lab-scale reverse osmosis system. ......................................... 27
Figure 3-5: Electrical system diagram with sensors, actuators, DAQ and user interface. ............ 29
Figure 3-6: User-interface for data monitoring in Labview.......................................................... 30
Figure 3-7: User control interface in Labview. ............................................................................. 30
Figure 3-8: a) Isometric view of the experimental water tank; and b) cut-away view of the inner
components of the experimental water tank. ................................................................................ 32
Figure 3-9: Pilot-scale system schematic. ..................................................................................... 34
Figure 3-10: Physical setup of the pilot-scale system. .................................................................. 35
Figure 3-11: Labview continuous monitoring interface for the pilot-scale system. ..................... 36
Figure 3-12: Labview control interface for the pilot-scale system. .............................................. 37
xv
Figure 3-13: Nobleton deep well (Nobleton, Ontario) with the a) well's green access box and
b) portable variable frequency drive for the submersible pump located inside the well. ............. 40
Figure 3-14: Groundwater collection at the Nobleton deep well, Ontario. .................................. 40
Figure 3-15: Operating pressure for the a) intermittent experiments and b) the continuous
experiments for a representative three days of operation. ............................................................ 42
Figure 3-16: Location of SEM and ATP samples for the lab-scale samples. ............................... 44
Figure 3-17: Unraveled TW30-2514 membrane for removal of the membrane autopsy samples.
....................................................................................................................................................... 45
Figure 3-18: Pilot-scale membrane autopsy sample locations shown on the unraveled membrane.
....................................................................................................................................................... 46
Figure 3-19: Laser-cut holder for the aluminum disks containing the SEM samples (gold sputter
coated reverse osmosis membrane samples)................................................................................. 47
Figure 4-1: Membrane permeability as a function of operating time for the intermittent run. The
hours when the system was shutdown are not included. .............................................................. 51
Figure 4-2: Transmembrane pressure for the intermittent and continuous experimental run. ...... 52
Figure 4-3: Membrane permeability vs. operating time for the continuous run (averaged over 30
min intervals). ............................................................................................................................... 53
Figure 4-4: Comparison of initial continuous and intermittent experiments. ............................... 54
Figure 5-1: Experimental system operational conditions: a) permeability, b) pressure, c) recovery
ratio, d) salt rejection. ................................................................................................................... 59
Figure 5-2: Normalized membrane permeability for continuous operation with and without the
use of F135 anti-scalant. ............................................................................................................... 61
Figure 5-3: Normalized membrane permeability of continuous operation and intermittent
operation with F135 anti-scalant. .................................................................................................. 63
xvi
Figure 5-4: Normalized permeability decline for intermittent operated experiment both with and
without anti-scalant and without rinsing. ...................................................................................... 64
Figure 5-5: Normalized membrane permeability of decline for intermittent operated experiment
with anti-scalant when operated with rinsing and without rinsing. .............................................. 65
Figure 5-6: Comparative bar graph of the experimental conditions investigated, the permeability
shown is the permeability at the fourth hour of the day. .............................................................. 66
Figure 5-7: Comparative bar graph of the average salt rejection for the experimental conditions
investigated. .................................................................................................................................. 67
Figure 5-8: Comparative SEM images of the various operating conditions at x 1000
magnification: a) clean un-used membrane, b) continuous with anti-scalant, c) continuous
without anti-scalant, d) intermittent with anti-scalant and with rinse, e) intermittent with anti-
scalant and and no rinse, f) intermittent without anti-scalant. ...................................................... 68
Figure 5-9: Comparative SEM cross-sections of the various operating conditions at x 500
magnification: a) clean un-used membrane, b) continuous with anti-scalant, c) continuous
without anti-scalant, d) intermittent with anti-scalant and with rinse, e) intermittent with anti-
scalant and and no rinse, f) intermittent without anti-scalant and without rinse. ......................... 69
Figure 5-10: Concentration of ATP on the membrane surface for various operating conditions. 70
Figure 6-1: Normalized membrane permeability for F260 anti-scalant when operated
continuously and intermittently. ................................................................................................... 77
Figure 6-2: Normalized membrane permeability for intermittent operation and F260 anti-scalant
with and without rinse. The rinse significantly improved the membrane permeability. .............. 78
Figure 6-3: Normalized permeability for the various operating conditions investigated at the
fourth hour of each day for day-to-day comparison. .................................................................... 79
Figure 6-4: Salt rejection for the groundwater experiments with F260. ....................................... 80
Figure 6-5: Membrane autopsy of the lab-scale experiments at x 1000 magnification and the
length bar represents 20 µm (a-c). Also, the membrane autopsies are shown at x 3000
xvii
magnification and the length bar represents 5 µm (d-f). For the continuous with F260 anti-scalant
(a, d), intermittent with F260 anti-scalant and no rinse (b, e) and intermittent with F260 anti-
scalant and with rinse (c, f) experiments. ..................................................................................... 82
Figure 6-6: Concentration of ATP on the membrane surface for lab-scale operating conditions
with anti-scalant F260. .................................................................................................................. 83
Figure 6-7: a) Normalized membrane permeability with the pilot-scale system operated with the
F260 anti-scalant and no rinsing in intermittent operation. The circled permeability rose above
the declining trend due to several start-up attempts after the submersible pump failed. .............. 87
Figure 6-8: Normalized membrane permeability for intermittent operation and the F260 anti-
scalant without rinsing for both the pilot-scale system and the lab-scale system. ........................ 88
Figure 6-9: Pilot-scale SEM autopsy a) compared to lab-scale SEM autopsy b) for intermittent
operation with F260 and no rinse at x 1000 magnification. Pilot-scale SEM autopsy c) compared
to lab-scale SEM autopsy d) for intermittent operation with F260 and no rinse at x 3000
magnification. ............................................................................................................................... 89
Figure 6-10: Pilot-scale membrane ATP compared to the lab-scale for intermittent operation with
F260 and no rinse. ......................................................................................................................... 90
Figure 6-11: Schematic of the difference between the lab-scale system and the pilot-scale system
during extended shutdown periods. .............................................................................................. 92
Figure 6-12: Proposed fouling mechanism process for intermittent operation of reverse osmosis
membranes .................................................................................................................................... 93
Figure 7-1: Framework for performing the cost optimization of the PVRO system. ................... 96
Figure 7-2: Solar powered reverse osmosis water treatment system architecture. ....................... 97
Figure 7-3: PVRO operating costs and pre-treatment................................................................... 98
Figure 7-4: System simulation flowchart outlining the hourly and daily steps in the simulation to
determine the loss of water probability. ...................................................................................... 102
xviii
Figure 7-5: Power management strategy for the solar powered water treatment model. simulation
for a day of sample operation in Mexico for a 1 m3 system with 4 solar panels and 2.6 kWh of
energy storage. ............................................................................................................................ 106
Figure 7-6: Membrane fouling model based on experimental data [110,131]. ......................... 108
Figure 7-7: Normalized membrane permeability decline (KFF) vs. operating time in days for Case
1-4 used in the water treatment system model. ........................................................................... 109
Figure 7-8: Tank capacity and costs from a supplier of plastic drinking water tanks [144]. ...... 111
Figure 7-9: a) Annualized system cost vs. design goal LOWP for variable system size (m3/day)
and b) the actual LOWP compared to the design goal LOWP for a 10 year simulation period for
La Mancalona, Mexico. .............................................................................................................. 117
Figure 7-10: Water cost ($/m3) vs. the system size (m3/day) for La Mancalona, Mexico at various
design goal loss of water probabilities (1%, 5%, 10%). ............................................................. 118
Figure 7-11: System configuration varies with size of the system for a 10 year simulation period
and at 5% LOWP with experimental fouling and anti-scalant usage with daily rinsing. ........... 120
Figure 7-12: Actual loss of water probability (LOWP) when simulated vs. the Designed goal
LOWP for several system sizes ( a) 1 m3/day, b) 5 m3/day, c) 10 m3/day). The optimal system
design was determined considering no fouling then simulated for a 10 year period with
experimental fouling (with anti-scalant and rinsing) and Abbas et. al. [110] fouling. ............... 123
Figure 7-13: Several communities were selected for evaluating the optimization algorithm in four
countries (Mexico, Ghana, India and Bangladesh). .................................................................... 124
Figure 7-14: Annualized system cost (USD) vs. variable LOWP for three communities in India,
Ghana, and Bangladesh) for variable system sizes (m3/day) a) 1 m3/day, b) 5 m3/day, c)
10 m3/day and a 10 year simulation period. ................................................................................ 125
1
Chapter 1
Introduction
1.1 Motivation
Globally, over 760 million people lack access to clean drinking water (Figure 1-1a) and its
associated health benefits [1]. They rely instead on seasonally available water sources (e.g. ponds,
rivers, rain cisterns, wells) that are often contaminated with excess dissolved minerals, chemicals,
bacteria and other pollutants [1]. Furthermore, these communities have limited access to electricity
[2]. Connecting these communities to large water treatment plants and the central electricity grid
is not a viable option due to their remote locations. Diesel generators are a common power source
for remote communities [3]. However, the high cost of diesel and environmental pollution has
limited their use for water treatment [4]. A viable alternative is the use of solar power since
communities facing high water scarcity (Figure 1-1a) also tend to be in regions of high solar
insolation (Figure 1-1b). To date, researchers have developed non-diesel water treatment
technologies appropriate for off-grid locations, such as solar ultra-violet (UV) water purification
systems [5], solar thermal desalination systems [6–8], solar photovoltaic reverse osmosis (PVRO)
systems [9–11] and wind powered reverse osmosis systems [12]. For these solutions to be
economic, the technologies must be fine-tuned to the location-specific energy source, the water
demand and the water characteristics of the community [13]. As well, due to their remote location,
the water treatment systems require minimal operator intervention, and should be reliable over the
lifetime of the systems at a minimum cost.
Solar photovoltaic powered reverse osmosis (PVRO) systems are promising for small-scale
(also called community-scale) desalination applications due to their low-cost, small footprint,
water purification efficacy and simple design from modular components [14]. Desalination is the
2
process of removing salts from brackish or seawater sources to produce drinking water. For the in-
land water-stressed communities considered in this thesis, the groundwater is often brackish with
a high level of dissolved minerals from the local geology. The level of dissolved minerals is
quantified using the total dissolved solids (TDS) which is a bulk measurement of the water salinity.
The water salinity ranges of typical water sources are provided in Table 1-1.
Table 1-1: Water salinity ranges in terms of the total dissolved solids, adapted from [15].
Water type Salinity Range
(mg/L of TDS)
Drinking water <500
Freshwater 50 - 500
Brackish 1,000 - 10,000
Seawater 30,000 - 45,000
Community-scale PVRO systems have been deployed previously in remote communities
[11, 15, 16]. PVRO systems require minimal maintenance because their repair is challenging in
remote communities [17–20]. Furthermore, energy storage is a large component of the system cost
due to the need for frequent battery replacements. To reduce the energy storage costs, PVRO
systems are often operated intermittently to reduce the need for batteries [8]. This intermittent
operation is a unique aspect of these PVRO systems since industrial-scale reverse osmosis plants
are designed to operate continuously. The effectiveness of reverse osmosis membranes is reduced
by scaling, biofouling and chemical degradation [22]. These effects can be minimized by selecting
effective pre-treatment [7] and chemical cleaning methods [23]. For PVRO systems, chemical
cleaning methods are rarely performed [8]. In addition, for PVRO systems, the effect of
intermittent operation, anti-scalant and membrane rinsing on membrane fouling needs to be
quantified [8].
3
Figure 1-1: The world map of a) water scarcity [1], and b) average solar irradiance [24]. Regions
with high water scarcity also tend to have high solar irradiance.
Another challenge for remote communities is the system affordability. To reduce the cost
of PVRO systems, they can be designed from commercially available modular components. This
reduces the cost by facilitating construction, parts replacement and system troubleshooting.
Complex design decisions are required to determine the appropriate system architecture.
Furthermore, to predict the membrane permeability of the intermittently operated PVRO systems,
a detailed experimental characterization of the reverse osmosis membrane fouling under
intermittent operation is required. Since every community has their own unique needs, each
location requires a customized design for the best performance. The performance is greatly
affected by the variability in the water characteristics, the water demand, the membrane fouling,
and the solar energy availability. Typically, skilled engineers are required to analyze the location-
specific conditions for a community to make the appropriate design decisions. For small remote
communities this kind of detailed design is not economically possible. An automated design
optimization framework can facilitate these complex design decisions and make this technology
accessible to these resource-constrained communities.
Low High No Data
Average Solar Irradiance (W/m2)
| | |
175 200 225 [2]
a) b)
Average Solar Irradiance (W/m2)
175 200 225
4
1.2 Problem Statement
The literature lacks a fundamental understanding of how the intermittent operation from
extended shutdown periods affects the permeability of the reverse osmosis (RO) membrane. This
intermittent operation is common in solar powered reverse osmosis water treatment systems for
resource-constrained communities. Both the fouling behavior of reverse osmosis membranes
under extended shutdown periods and the operating conditions to mitigate this fouling need to be
better understood. In addition, the current design methods for PVRO systems lack an analytical
model describing the membrane fouling behavior caused by intermittent operation. Design
optimization frameworks for PVRO systems that consider the membrane fouling from intermittent
operation are required to increase adoption of this technology in remote communities.
1.3 Research Objectives and Goals
The goal of this research is to develop a fundamental understanding of reverse osmosis
membrane fouling for brackish water under intermittent operation with extended shutdown
periods. This intermittent operation is common for small-scale (1 - 10 m3/day) solar powered
reverse osmosis systems. The subsequent goal is to use the experimental findings to develop an
analytical model of the membrane permeability decline for use in the development of a design
optimization framework for these water treatment systems. To accomplish these goals, the research
objectives are as follows:
1. Development of an experimental system and protocol for the characterization of
intermittent operation for small-scale solar powered water treatment systems.
5
2. Experimental characterization of reverse osmosis membrane fouling by scaling for
intermittent operation compared to continuous operation when using anti-scalant pre-
treatment and shutdown procedures appropriate for remote locations.
3. Experimental characterization of reverse osmosis membrane fouling for intermittent
operation at the lab-scale and pilot-scale.
4. Development of a design optimization framework for small-scale PVRO brackish water
systems that considers membrane fouling from intermittent operation using an analytical
membrane fouling model.
1.4 Research Scope
The research conducted in this thesis focused on developing an improved understanding of
membrane fouling caused by intermittent operation of PVRO systems using experimental
characterization. As well, the research focused on the development of a design framework for these
systems considering the membrane fouling from intermittent operation. The characterization of
membrane fouling caused by intermittent operation focused on the effect of extended shutdown
periods, commonly experienced in PVRO systems with limited battery storage. The experimental
investigations concentrated on the characterization of the membrane fouling for two types of
experimental water. The first experimental water used MilliQ® (a high purity lab-grade water) as
the solvent and was termed ‘experimental MilliQ-based matrix.’ The second experimental water
used a local groundwater from Nobleton, Ontario as the solvent and was termed ‘experimental
groundwater-based matrix.’ The experimental waters were mixed with lab-grade chemicals to
mimic groundwater high in dissolved minerals, common of inland brackish desalination. Using
the experimental results, the research then focused on the development of a design optimization
framework for small-scale (1 - 10 m3/day) brackish water PVRO systems.
6
1.5 Thesis Contributions
The main contributions of this PhD research work were the experimental characterization of
membrane fouling from intermittent operation and the development of a novel design optimization
framework for PVRO systems that considers membrane fouling from intermittent operation. This
thesis resulted in the following contributions:
The experimental quantification of reverse osmosis membrane fouling under intermittent
operation and continuous operation was conducted, and compared to each other. It was the first
of its kind in literature. A fully instrumented and automated experimental system was
developed for quantifying the RO membrane fouling in triplicate. The experimental system
was designed, built, commissioned and operated to characterize the membrane fouling
resulting from extended shutdown periods overnight.
o Published Journal Paper: Freire-Gormaly, M., Bilton, A., (2017) An Experimental System
for Characterization of Membrane Fouling of Solar Photovoltaic Reverse Osmosis Systems
under Intermittent Operation, Desalination and Water Treatment, volume 73, pp.54-63.
The characterization of membrane fouling during extended shutdown periods for reverse
osmosis membranes in a controlled experimental system with anti-scalant pre-treatment and
rinsing with permeate water was performed. This was conducted using the experimental
MilliQ-based matrix. The membrane permeability and membrane autopsy using scanning
electron microscopy (SEM) showed that the use of anti-scalant and rinsing maintained high
membrane permeability.
7
o Published Journal Paper: Freire-Gormaly, M., Bilton, A., (2018) Experimental
Quantification of the Effect of Intermittent Operation on Membrane Performance of Solar
Powered Reverse Osmosis Desalination Systems, Desalination, volume 435, pp.188-197.
The characterization of reverse osmosis membrane fouling under intermittent operation was
carried out in controlled experiments using the lab-scale and pilot-scale experimental systems
with the experimental groundwater-based matrix. The experiments were operated with anti-
scalant pre-treatment and rinsing with permeate water prior to shut down. The lab-scale
membrane fouling results showed that the membrane permeability was maximized when the
system was operated intermittently with a daily permeate water rinse prior to shut down. The
lab-scale results also represented the pilot-scale experimental results.
o Journal Paper: Freire-Gormaly, M., Bilton, A., (In-Preparation) Experimental Lab-scale
and Pilot-scale Characterization of the Effect of Intermittent Operation on Membrane
Fouling for Solar Powered Reverse Osmosis Desalination Systems, Desalination.
A novel design optimization framework for PVRO systems was developed, which considered
the membrane fouling under intermittent operation. The framework configured the PVRO
system components (e.g. number of solar panels, number of RO membranes, size of the water
tank) and operating conditions (i.e. with or without rinse, with or without anti-scalant) for a
user-defined geographic location and water demand. It was found that lower reliability systems
reduced the annualized system costs. It was demonstrated that considering membrane fouling
was critical to design reliable and cost-optimal PVRO systems. The cost-optimal system
configurations were compared for several geographic locations (Mexico, Ghana, India and
Bangladesh). It was also shown that despite variations in solar insolation and water salinity
between these geographic locations, a similar system configuration met the water demands.
8
o Journal Paper: Freire-Gormaly, M., Bilton, A., (Under Review) Design of Solar
Powered Reverse Osmosis Desalination Systems Considering Membrane Fouling
caused by Intermittent Operation, Desalination, DES_2018_301
1.6 Research Program
The research in this thesis followed the methodology shown in Figure 1-2. The research
program comprised of experimental investigations and a design optimization of solar powered
reverse osmosis systems. Based on the results of the experimental program, an analytical model of
membrane fouling was developed, and a design optimization framework was implemented for
solar powered reverse osmosis water treatment systems. The experimental program allowed for
quantification of the effect intermittent operation had on membrane fouling. The experimental
results also allowed for identifying potential fouling mechanisms which are unique to intermittent
operation compared to continuous operation. As well, in resource-constrained communities,
minimal pre-treatment, operator intervention and brine volume are required since brine is typically
release to the environment with minimal post-treatment. Therefore, developing a design
framework to automatically configure the cost optimal system with the required reliability and
which considers the membrane fouling caused by intermittent operation was required.
9
Figure 1-2: Flowchart of the research program conducted in this thesis with the associated
chapter numbers.
1.7 Thesis Organization
This thesis is organized into eight chapters. The first chapter provides the motivation, the
problem statement, the objectives, and the contributions. The second chapter covers the
background and an extensive literature review of PVRO systems, membrane fouling and system
design optimization. The third chapter outlines the experimental methods and the experimental
equipment setup and instrumentation. The fourth chapter provides the initial experimental results
when the system was operated intermittently and continuously with extended shutdown periods
for the experimental MilliQ-based matrix. The fifth chapter presents the detailed experimental
results with an improved experimental system to characterize the membrane permeability for
intermittent operation with anti-scalant pre-treatment and rinsing with permeate water for the
experimental MilliQ-based matrix. The sixth chapter describes the experimental results with the
Develop Design Framework for PVRO
Design Optimization Framework
Experimental Characterization of Reverse Osmosis Fouling
Simulation
RO Fouling
Pre-treatment and Membrane Rinsing
Design Optimization Module
RO SystemOperation
PVRO System Costs
Improved Lab-scale
MilliQ-based matrix
Initial Lab-scale
MilliQ-based matrix
Ch. 7
Ch. 4 Ch. 5
Design, Build and Commission New Experimental Systems Ch. 3
Groundwater-based matrix
Pilot-scale Ch. 6
Groundwater-based matrix
10
improved experimental system and pilot-scale system using a second experimental water matrix
to characterize the membrane permeability for intermittent operation with anti-scalant pre-
treatment and rinsing with permeate water for the experimental groundwater-based matrix. The
seventh chapter presents the design optimization framework for cost-optimal PVRO systems that
are designed to meet a community’s water needs based on their geographic location and water
characteristics. The eighth chapter provides the conclusions of this thesis and the directions for
future work.
11
Chapter 2
Background and Literature Review
2.1 Renewable Powered Reverse Osmosis Desalination
Renewable powered reverse osmosis desalination has been a topic of intense research
due to the growing need for stand-alone water treatment systems for remote locations [25–34].
Several existing renewable powered reverse osmosis systems have been installed and studied
previously [20,35–38]. Studies on existing renewable powered reverse osmosis systems tend to
focus on either the pilot-scale implementation [39–42] or the design and cost optimization [43–
46]. All renewable powered desalination systems have inherent intermittent operation due to the
intermittent nature of the power source. For example, wind power is only available during wind
speeds within the cut-off range of the wind turbine. As well, solar power is only available during
daylight hours and varies with cloud cover. Some of these systems are designed for high solar
insolation locations and therefore use solar photovoltaics coupled with batteries as their main
source of power [33,43,47,48]. To reduce the costs of these systems, very little battery storage
is incorporated in the design leading to intermittently operated systems with extended shutdown
periods [33,43,47,48]. Many researchers claim intermittent operation leads to increased
membrane fouling rates [33,39,42,49] and high membrane fouling rates have been observed in
operating plants with several years of data [50]. However, no one to date validated this
experimentally and evaluated the effectiveness of simple treatments to alleviate membrane
fouling.
12
2.1.1 Reverse Osmosis Water Treatment
The reverse osmosis (RO) process produces fresh water by applying a pressure higher
than the osmotic pressure across a semi-permeable membrane [5,6,8,51,52]. Water passes
through the membrane producing permeate and leaves behind excess dissolved solids
constituting a concentrated brine (retentate). RO is one of the main commercial desalination
technologies due to its high energy efficiency, no thermal requirements, and modularity. Spiral
wound membranes are the most common RO membrane type in use. The permeate flow rate
across the membrane is proportional to the difference between the applied pressure and the
osmotic pressure, given by Equation 2-1:
(2-1)
where 𝑄𝑃 is the permeate flow rate (m3/s), 𝐾𝑊 =1
𝑅𝜇 is the membrane permeability for water
(ms-1bar-1), R is the membrane resistance as a function of fouling (m-3s2bar), µ is the dynamic
viscosity of water (m2/s), Amem is the membrane surface area (m2), KT is the water permeability
temperature correction factor, ∆�̅� is the average pressure applied across the membrane (bar), and
∆�̅� is the average osmotic pressure applied across the membrane (bar). The membrane flux
(Lm-2h-1) is given by 𝐽 =𝑄𝑃
𝐴𝑚𝑒𝑚 and is commonly reported to describe the amount of permeate
produced per unit of membrane area. The membranes can become fouled by particulates, scaling
or biological growth leading to a decline in the membrane permeability and an increase in
membrane resistance. A typical response of the membrane resistance is shown in Figure 2-1.
Some of the increase in membrane resistance can be reversed by cleaning, but some fouling is
irreversible. This leads to an overall decrease in membrane water permeability and increases the
energy requirements. Typically, small-scale RO plants do not have cleaning cycles due to the
𝑄𝑃 = 𝐾𝑊𝐴𝑚𝑒𝑚𝐾𝑇(∆�̅� − ∆�̅�)
13
high costs of cleaning chemicals and lack of facilities for chemical disposal in their remote
locations.
Figure 2-1: Reverse osmosis membrane resistance as a function of time.
2.1.2 Overview of Renewable Powered RO Systems
RO membrane desalination can be coupled with power systems to produce standalone,
modular water treatment systems. To reduce costs, numerous studies have looked at the overall
design and optimization of RO membrane systems [9,10,53–55] and investigated small scale
solar photovoltaic reverse osmosis (PVRO) systems that produced less than 10 m3/day of clean
drinking water [11,16,17,19,33,48,56]. Figure 2-2 shows a schematic of a basic PVRO system.
The feed water pump transports brackish water from an intake to a simple pre-treatment system,
typically a cartridge filter and anti-scalants [11]. A high-pressure pump provides the pressure
required to operate the RO membrane module. The PVRO system is powered by solar
photovoltaics and the power is distributed through the system using control electronics. The
permeate water is stored in a fresh water tank. In remote communities, the reject brine is typically
released slowly to the environment [16,33].
Mem
bra
ne
Res
ista
nce
(m
-1)
Operating Time (Days)
Rinitial
Rreversible
Rirreversible
Cleaning Cleaning
14
Figure 2-2: Simple PVRO system schematic for remote communities.
Membrane fouling is one of the major challenges to RO membrane systems and as a
result has been the subject of intense research over the past decades [40,57–65]. These studies
focused on lab-scale experiments to identify the onset of crystal nucleation and precipitation for
scaling [57–60] and the effects of biofilm on RO performance [61,64,65]. Membrane fouling is
a concern for small-scale systems in remote communities due to limited financial flexibility for
membrane replacement costs, chemical cleaning costs, limited operator knowledge and, lack of
access to resources to improve system performance by chemical or mechanical cleaning once
the system is operational.
Considering membrane fouling in the design of PVRO systems has been identified as a
major area for further research since existing design methods do not account for RO membrane
fouling, nor the effects of pretreatment [11]. An added challenge for small-scale PVRO systems
in remote communities is intermittent operation to reduce the need for batteries, which are an
expensive component of the system [8]. It is generally regarded that this intermittent operation
leads to premature fouling of the reverse osmosis membranes, however, this behavior has not
been characterized experimentally [8].
Brackish Water Intake
Control Electronics
Reject Brine
Fresh Water Tank
Reverse Osmosis ModuleFeed Water
Pump
High Pressure PumpPre-Treatment
Solar PV Array
15
2.2 Reverse Osmosis Membrane Fouling
Reverse osmosis membrane foulants can be categorized into several groups (biological,
mineral, particle, colloidal, organic and oxidant). For particle fouling, typically filtration with
5 µm sediment filters is sufficient to remove suspended solids. For colloidal fouling and organic
fouling coagulation and filtration are typically used as treatment methods. Oxidant fouling
occurs in the presence of free chlorine and oxidants, which can be present if chlorination or
ozonation are used to treat biological content. The reverse osmosis membrane is very sensitive
to oxidant foulants, and hence these pre-treatment methods are not recommended. Mineral
scaling and biofouling are the dominant sources of reverse osmosis membrane fouling in
operating plants [8]. Membrane fouling causes increased pressure requirements [8], energy
requirements [7], costs due to early replacement of membranes [15,66] and chemical costs for
cleaning [67]. High flux rates correspond to higher permeate volumes, however, exceeding the
critical flux rate leads to increased membrane fouling due to the boundary effects from
concentration polarization and hydrodynamic effects at the membrane interface [68]. High flux
rates result in high levels of scaling by locally exceeding the solubility limit of dissolved minerals
and by compacting dispersed foulants onto the membrane surface.
2.2.1 Biofouling
Biofouling is caused by the growth of biofilm on the membrane surface [65,67,69,70].
Biofouling has been associated with increasing membrane resistance, even with 99.99%
biological component removal from the feed water [71]. Studies on biofilm growth mechanisms
have been conducted in the literature [52,64,72–74]. These researchers performed modeling,
simulations and lab-scale experiments to quantify biofouling and the influence on system
parameters [52,64,72–74]. Radu et al. [64,72] presented a detailed biomass growth model in two-
dimensions and three-dimensions in small lab-scale channels. This work was extended to spiral-
16
wound elements by Vrouwenvelder et. al. [74]. None of the previous research has studied how
intermittent operation from extended shutdown periods affect the growth of biofilms in RO
systems.
2.2.2 Scaling
Scaling is caused by the precipitation of dissolved minerals onto the RO membrane and
system components [75]. Scaling in brackish water is a major concern due to the high levels of
dissolved calcium carbonate, silicates and other dissolved minerals often present in groundwater
due to the local geology. Several researchers have studied the scaling of RO membranes by
calcium carbonate [76–79], silica and iron oxides [80–82] for operational conditions analogous
to large-scale desalination plants. By treating fouling on early on-set, the amount of irreversible
fouling (fouling that cannot be cleaned) can be minimized [83]. None of the previous scaling
studies investigated how intermittent operation affects the crystal growth on RO membranes.
2.2.3 Pre-treatment to Minimize RO Membrane Fouling
Fouling of RO membranes can be reduced by adjusting the operating characteristics
(flux, recovery rate, feed channel pressure drop) or by using appropriate pre-treatment methods
selected based on the characteristics of the feed-water (e.g. anti-scalants, sand filtration,
coagulation and flocculation), and overall system design [84]. For small-scale RO systems in
remote communities, minimal operator intervention is required. Anti-scalants are a suitable
chemical pre-treatment for small-scale brackish water systems because they can be injected in-
line autonomously without operator intervention.
Anti-scalants shift the solubility curves of the sparingly soluble minerals to delay
precipitation [85]. The dose of anti-scalant and the chemical composition of anti-scalants for
optimal performance remains an area of active research [23,83,86–94]. Anti-scalants are
17
commonly used to minimize mineral scaling as they delay the onset of targeted mineral scales
[92,95,96]. Previous studies have outlined the mechanisms of scale formation [49,57,63,76,97]
and the effect of pH on scale formation [98] for continuously operated systems. These studies
were focused on the cleaning schedule for RO membrane modules in large-scale RO desalination
systems operated continuously. For small-scale brackish water systems, cleaning of the RO
modules is rarely incorporated in the design [99]. Recent studies have shown determining an
optimal cleaning schedule before implementation for small-scale renewable powered RO
systems is unreliable [18]. These previous studies did not consider the intermittency inherent
from small-scale renewable powered RO systems. Hence, characterization of scaling of RO
membranes in small-scale renewable powered RO systems operated intermittently is required.
2.2.4 Experimental Studies on RO Membrane Fouling
The existing experimental studies on membrane fouling have explored both biofouling
and scaling for industrial-scale desalination systems. RO membrane experiments on full-scale
membrane cartridges were conducted to observe the role monochloramine has on fouling rate
and fouling potential [100]. They demonstrated that monochloramine can reduce fouling rates,
but can also have a negative impact on the life of the membrane due to degradation of the
membrane polymer. Periodic cleaning of the RO membranes with air and water (air sparging)
has been shown to be effective in removal of biological fouling and organic nutrients [101].
Real-time monitoring systems and alarm systems for early identification of fouling [102,103]
have also been developed. However, these are complicated systems that were developed for use
in industrial-scale plants. The literature lacks experiments which characterize the effectiveness
of pre-treatment options on reducing membrane fouling rates of brackish RO systems operated
intermittently with extended shutdown periods.
18
2.2.5 RO Membrane Fouling Models and Mechanisms
Existing scaling models in literature [49,79,98,104–108] focus on single compound
scales, and are typically at the lab-scale. Some studies were conducted on modeling the
membrane fouling of full-scale system operation of spiral-wound modules over a period of
4 months [106]. Hoek et al. [106] developed a new semi-empirical model and found it was able
to predict the experimentally observed membrane fouling. Hoek et al. [106] also proposed a new
mechanism of cake-enhanced concentration polarization. However, this model was for
secondary wastewater effluent which had different fouling compounds than the brackish
groundwater considered in this thesis. Oh et al. developed a mathematical model of scale
formation which focused on two main mechanisms of scale formation, bulk crystallization and
surface crystallization [108]. Bulk crystallization is when crystals form in the feed solution and
deposit on the surface of the membrane resulting in reduced permeability. Surface crystallization
is when crystal growth occurs directly on the surface of the membrane. Concentration
polarization plays a key role in scale formation, increased concentration polarization leads to
surface crystallization as the dominant crystal growth mechanism. These previous studies
focused on modeling the various factors that affect scale formation (e.g. pH, cross-flow velocity,
pre-treatment) in continuous operation. None of the previous studies analyzed the effect of
intermittent operation on membrane scaling.
One fouling model studied the effect of both biofouling and scaling simultaneously [109]
and showed the presence of a biofilm can increase the concentration polarization and thereby
induce scaling. A study by Thompson et. al. [109] used a single model foulant (gypsum) and
focused on continuous operation. The literature lacks fouling models under intermittent
operating conditions for real groundwater which capture the effects of both scaling and
biofouling.
19
Although renewable powered RO systems have been in operation, very few long-term,
physics-based mathematical models of reverse osmosis membrane fouling have been proposed
in literature [110,111]. Recent studies have built upon this work by developing fouling models
using neural networks and machine-learning techniques [111–113]. These studies required
significant training data and were developed for industrial-scale continuously operated RO
systems. Despite these recent advances, mathematical fouling models are required for long-term
simulations of reverse osmosis systems operated intermittently.
2.3 Design Approaches for Solar Powered RO Systems
Researchers have formulated cost and energy optimization design approaches for small-
scale renewable powered reverse osmosis systems [25,50] and small-scale PVRO systems
[11,14,33,114,115]. Bilton et al. developed a modular design architecture for the cost-optimal
PVRO systems [11]. The computer-based approach configured systems from an inventory of
components by applying design principles to limit the design space and employing a genetic
algorithm to find the minimum cost solution. These previous design algorithms focused
predominantly on the costs of the system and detailed physics-based models of the components
[33,115]. The RO models in these existing design algorithms focus on the selection of the RO
modules and do not consider a detailed fouling model. Instead, some researchers have arbitrarily
assumed fixed fouling rates [25,115]. These researchers also considered the down time of
extended shutdown periods to size water storage and to over-size the required RO membrane
array [25,115]. None of the previous design optimization algorithms considered the impact RO
membrane fouling from intermittent operation has on the system life or cost.
Although membrane permeability decline has been predicted using neural networks for
large-scale continuously operated RO water treatment plants [111–113], only limited research
20
has been performed for predicting the membrane performance decline of small-scale PVRO
systems [18]. Previous optimization studies for PVRO systems do not consider how operating
conditions, such as the use of anti-scalant or permeate rinsing affect membrane permeability and
membrane replacement rates [14,114,116]. As a result, in the real operating system, as the
membrane permeability declines, water production would decrease and would not be able to
meet demand. These previous optimization studies therefore over-estimate the system reliability
and the community’s water demand would not be adequately met by these systems. An integrated
design framework for PVRO systems with extended shutdown periods that considers the impacts
of membrane fouling from intermittent operation, pre-treatment, and cost and energy
requirements is required.
2.4 Summary and Research Needs
This chapter provided a review of the existing literature on reverse osmosis membrane
fouling by biofouling and scaling; and methods of minimizing membrane fouling using pre-
treatment technology. A review of the experimental studies on RO membrane fouling and fouling
mechanisms was also provided. No previous studies have considered the impact intermittent
operation would have on membrane fouling. Also, none of the previous work investigated the
effect of simple interventions appropriate for remote areas, such as rinsing and anti-scalant, on
membrane fouling in intermittently operated RO systems. The existing literature provided
insight to the challenges of designing solar photovoltaic reverse osmosis systems, but none
completely solve the challenges of designing a system that will meet the water demands of a
community considering membrane fouling from intermittent operation.
The research undertaken in this thesis seeks to resolve some of these remaining
challenges to improve the provision of clean drinking water to off-grid remote communities.
This research seeks to develop an improved understanding of how reverse osmosis membranes
21
become fouled by the daily extended shutdown period common of intermittently operated
renewable powered and solar powered desalination systems. Then to use this new information
to develop a new design algorithm to design solar powered desalination systems that considers
the membrane permeability decline from membrane fouling caused by intermittent operation.
22
Chapter 3
Experimental Systems and Methods
Some of the content of this chapter has been previously published (full citation: Freire-Gormaly,
M., Bilton, A., (2017) An Experimental System for Characterization of Membrane Fouling of
Solar Photovoltaic Reverse Osmosis Systems under Intermittent Operation, Desalination and
Water Treatment, vol.73, pp.54-63 and in Freire-Gormaly, M., Bilton, A., (2017) Experimental
Quantification of the Effect of Intermittent Operation on Membrane Performance of Solar
Powered Reverse Osmosis Desalination Systems, Special Issue: Desalination, vol.435, pp.188-
197. It has been reproduced here. Permission to use this content was secured from the editor.
3.1 Introduction
This chapter outlines the experimental systems and methods used to quantify the reverse
osmosis membrane fouling under intermittent operation. There were three experimental systems
that were used. An initial lab-scale system, an improved lab-scale system and a pilot-scale
system. The initial lab-scale system is described. The detailed description of the improved lab-
scale system setup is provided. The instrumentation and data acquisition are also described. The
auxiliary system components are also described, for example the experimental water storage
tank. The pilot-scale system and its instrumentation are described. Finally, the experimental
methods are outlined.
3.2 Lab-scale Systems
3.2.1 Initial Lab-scale System Setup
As part of this research, a custom experimental lab-scale system was designed, built and
commissioned to allow for smaller water volumes than typically needed for real systems for the
experiments. The initial lab-scale system schematic (Figure 3-1) shows the various system
automation and monitoring equipment. The lab-scale system (Figure 3-2) allowed for testing
different two different operating conditions (continuous and intermittent operation). The system
23
was configured with the ability to perform automated rinsing and anti-scalant dosing. However,
the initial experiments did not use these features.
The system was designed to allow for continuous monitoring of the system operating
conditions including the pressure, conductivity and flow rates. This data was required to
determine the membrane permeability. The initial lab-scale system used an equalization tank
with a gravity fed float valve to maintain the recovery ratio at a consistent 75%. As well, the
system configuration was on a small portable stand to allow for transfer of the equipment through
the lab door.
The initial experimental setup allowed for triplicate measurements of the permeate flow
rates using three identical cross-flow cells (SEPA cells). The cross-flow cells permitted smaller
water volumes to be used per experiment than a spiral-wound membrane module since a much
smaller membrane area was used (0.0266 m2) for the cross-flow cells compared to the spiral
wound elements (7-34 m2). The online monitoring of the system operating conditions allowed
for detailed analysis of the system behaviour.
The initial experimental system was equipped with four analog flowmeters (FLR1000,
Omega Inc., Laval, Quebec) to measure the permeate flowrate for each SEPA CF module and to
measure the brine flowrate going to drain. The experimental system was also equipped with two
digital flowmeters (FPR301, Omega Inc., Laval, Quebec) to measure the feed flowrate and
recirculation flowrate. Four inline conductivity sensors (A1002, EC/pH Sensors, Boston MA)
measured the electrical conductivity, three for the permeate water (0-500 µS/cm) from each
SEPA cross-flow cell and one for the feed flow (0-5000 µS/cm).
24
Figure 3-1: Initial experimental lab-scale system schematic including instrumentation.
The initial experimental system used a variable frequency driven (VFD) high pressure
positive displacement pump (Hydracell Pump Model M03SASGSSSPA, Wanner Engineering
Inc., Minneapolis, MN) to maintain the desired feed flowrate. A pressure release valve (C46
Valve, Wanner Engineering Inc., Minneapolis, MN) and a pressure regulator were also used to
ensure the system pressure did not exceed safe operating pressures. To remove particulate matter,
a 5-inch tall, 3½-inch diameter filter housing with a spun polypropylene 5 µm cartridge filter
(Pentek, Upper Saddle River, NJ) was installed in front of the high pressure pump. The initial
experimental system did not have computer-based control of the variable-frequency drive. The
25
experimental system was manually started and shut down using the manual interface of the
variable frequency drive.
Figure 3-2: Physical initial experimental lab-scale setup corresponding to the lab-scale
schematic in Figure 3-1.
3.2.2 Improved Lab-scale System Setup
After the initial experiments were completed, the initial experimental lab-scale system
was improved to a new configuration which permitted more equal distribution of the feed flow
to the three cross-flow cells (Figure 3-3). The improved lab-scale system allowed for testing
different operating conditions (e.g. anti-scalant, intermittent operation, continuous operation,
rinsing before shutdown) and online continuous monitoring of the system parameters (e.g.
Feed WaterTank
Agitator
DAQ
Custom Electronics
LabVIEW User Interface
High Pressure
Pump
Equalization Tank
RO Module 3RO Module 2
RO Module 1
Chiller
26
pressure, conductivity, permeate flow rates). The experimental lab-scale system (Figure 3-4) also
had autonomous control systems to maintain a consistent recovery ratio using an automated
needle valve, and to maintain a consistent level in the equalization tank using a float-switch
coupled to a solenoid valve.
Figure 3-3: Improved experimental lab-scale system schematic with instrumentation.
To ensure the inlet pressure of the high-pressure pump did not drop below the rated
suction pressure, a submersible pump (Rule iL200 Marine 200 GPH Inline Submersible Pump
(12-Volt, Intermittent Duty), Xylem Inc., Beverly, MA) was used on the inlet of the feed water
inlet before the cartridge filter. This was an improvement compared to the initial lab-scale system
which was reliant on the suction of the pump. The equalization tank was used to vent air to
27
atmosphere and to enable recirculation of the brine and experiment water to reach high recovery
ratios typical of brackish water desalination systems. The float switch (.6A NO/NC POLY,
RSF88Y100R, Cynergy3 Components LLC, Garden Grove, CA) coupled with the actuation of
a solenoid valve (SV3201, Omega Inc., Laval, Quebec) was used to refill the equalization tank
with experiment water at the same rate of net water withdrawal (the three SEPA cross-flow
permeate flowrates plus the brine flowrate).
Figure 3-4: Physical experimental lab-scale reverse osmosis system.
RO Module 3
RO Module 2
RO Module 1
pH control
Conductivity Sensors
Flowmeters
VFD LabVIEW Interface
Custom Electronics
DAQ
Agitator
High Pressure Pump & Chiller
(Outside of View)
Automated Needle Valve
28
The experimental system also consisted of equipment for automated rinsing, anti-scalant
dosing, and start-up and shutdown procedures. A solenoid valve (SV3202, Omega Inc., Laval,
Quebec) and an automatic three-way control valve (Electric Actuated PVC 3-Way Ball Valves
- Multi-Voltage 5615, Valworx ®, Cornelius, NC) was installed for permeate rinses. An anti-
scalant dosing peristaltic pump (100.PH.030/4 Peristaltic Pump, Williamson Inc., Brighton,
Sussex) was installed for accurate dosage of anti-scalant pre-treatment.
3.2.3 Lab-scale Instrumentation
The lab-scale system was instrumented with continuous data collection and monitoring.
The lab-scale system was equipped with three analog flowmeters (Alicat Scientific L-10ccm-
D/5V) which allowed for high accuracy measurements at low flowrates in the range of
0-10 mL/min to measure the permeate flowrate for each SEPA cross-flow module. A single
analog flowmeter (Alicat Scientific L-100ccm-D/5V) was used to measure the brine flowrate in
the range of 25-100 mL/min. The lab-scale system was also equipped with two digital
flowmeters (FPR301, Omega Inc., Laval, Quebec) to measure the feed flowrate and recirculation
flowrate. Four inline conductivity sensors (A1002, EC/pH Sensors, Boston MA) measured the
electrical conductivity, three for the permeate water (0-500 µS/cm) from each SEPA cross-flow
cell and one for the feed flow (0-5000 µS/cm).
The high pressure before the SEPA cross-flow cells was measured using a high-pressure
transducer (P51-1000-S-A-I36-4.5V, SSI Technologies, Janesville, WI). The temperature of the
feed water was monitored using a thermistor (TH-44000-NPT Thermistor, Omega Inc., Laval,
Quebec). The feed water temperature was maintained in the equalization tank using an
immersion chiller (13271-500, VWR, Mississauga, Ontario) and a temperature controller
29
(61161-300, VWR, Mississauga, Ontario) set at 151oC, analogous to the source water
temperature of the target community.
The improved experimental lab-scale setup included continuous pH control through
automated acid (HCl) or base (KOH) addition and continuous pH monitoring. A data acquisition
system (DAQ NI USB-6343, National Instruments) and custom electronics were used to collect
the data. A diagram of the electrical connections is shown in Figure 3-5.
Figure 3-5: Electrical system diagram with sensors, actuators, DAQ and user interface.
Custom circuits were designed and soldered to interface the sensors with the data
acquisition system. Driver circuits were also designed for the dosing pump, feed pump, solenoid
valve, float switch and three-way control valve. Custom power circuits were also configured to
provide appropriate and well-conditioned power to all sensors and actuators.
30
Figure 3-6: User-interface for data monitoring in Labview.
Figure 3-7: User control interface in Labview.
31
The system control and monitoring were performed using a custom-made Labview
program. Figure 3-6 shows the user interface which provides online monitoring of the
conductivity, temperature, pressure, flow rates, and recovery ratio. User control was also
performed through a secondary interface (Figure 3-7). For example, the anti-scalant dosing pump
could be actuated, the three-way control valve could be activated, and the system could be turned
on or off using the variable frequency drive. In addition, automated scripts were written to
provide timing for the intermittent experiments presented below.
3.2.4 Experimental Water Tank
The experimental water tank (Figure 3-8) was equipped with UV disinfection,
temperature control, and continuous agitation to ensure a consistent source water for the
experiments. The UV disinfection system (Tank MasterTM TM22 UV Tank Storage Sanitizer,
Atlantic Ultraviolet Corp., Hauppauge, NY) provided continuous UV exposure to the water in
the tank. The temperature control was provided using a second immersion chiller (13271-500,
VWR, Mississauga, Ontario) and a temperature controller unit (61161-300, VWR, Mississauga,
Ontario) ensured the experiment water was maintained at a constant temperature (10oC)
throughout the experiment duration. Continuous agitation was provided using a mechanical
agitator (Arrow Model 1200, Arrow Engineering Ltd., Hillside, New Jersey). The continuous
agitation ensured the chiller did not become encrusted in ice and that there was an even
temperature in the tank and sufficient mixing to ensure the concentration was uniform throughout
the tank. The level of water in the tank was maintained above the minimum reach of the chiller
coil and agitator (approximately 80 L). All ports were covered to prevent dust or particles from
the ambient air entering the tank during operation of the experiments.
32
Figure 3-8: a) Isometric view of the experimental water tank; and b) cut-away view of the inner
components of the experimental water tank.
3.3 Pilot-scale System
3.3.1 Pilot-scale System Setup
In real systems implemented in the field, full-scale spiral wound reverse osmosis
elements are used. However, to limit the water requirements, the lab-scale system was
implemented with smaller membrane coupons. Therefore, to ensure the lab-scale experimental
system adequately represented the membrane fouling behavior of a real-system, the pilot-scale
system was built using a spiral-wound element and operated intermittently. The pilot-scale
system was designed to a size that would minimize the daily feed water requirements to a range
that would be practical to fulfill through local groundwater collection.
To keep the initial PVRO plant in La Mancalona, Mexico a cost-effective system, the
pilot-plant was not fully instrumented, therefore real operating data was unavailable. Although
an alternate option would be to perform experiments on a pilot-scale system in the target
Chiller Coil
Agitator
Tank LevelIndicator
a) b)
33
community, it was impractical to go to the field to perform controlled experiments. As well, the
PVRO plant in La Mancalona, Mexico is an operating drinking water plant for the community.
It would be an unnecessary burden on the community to disrupt the community’s only source of
clean drinking water to perform experiments. It was decided that this alternate option would not
be responsible engagement with the community.
The pilot-scale system was designed in a similar configuration to the real operating pilot-
plant and it mirrored the lab-scale system configuration (Figure 3-9). The main difference
between the pilot-scale and the lab-scale system was that the pilot-scale system was not run in
triplicate. If the pilot-scale system was run in triplicate the water volumes would be prohibitively
large. Instead, the pilot-scale system was run with a single spiral-wound membrane element
(Figure 3-10). Again, similar to the lab-scale system an equalization tank was used in the pilot-
scale system to achieve high recovery ratios, similar to those achieved in the field with a larger
spiral-wound element.
The pilot-scale system was pressurized with a high pressure pump (Hydracell Pump
Model M03SASGSSSPA, Wanner Engineering Inc., Minneapolis, MN) driven with a variable
frequency drive to maintain the desired feed flowrate. A pressure release valve (C46 Valve,
Wanner Engineering Inc., Minneapolis, MN) was used keep the system pressure below the safe
operating pressures. To remove particulate matter, a 10-inch tall x 3.5-inch diameter filter
housing with spun polypropylene 5 µm cartridge filter (Pentek, Upper Saddle River, NJ) was
installed in front of the high-pressure pump. A submersible pump (Rule iL200 Marine 200 GPH
Inline Submersible Pump, 12-Volt, Intermittent Duty, Xylem Inc., Beverly, MA) pumped the
feed water through the filter cartridge.
34
Figure 3-9: Pilot-scale system schematic.
The TW2514 RO spiral wound element (Dow Filmtec, The Dow Chemical Company)
was housed in an RO fiberglass pressure vessel (ROPV, Harbin ROPV Industrial Co., Ltd,
Nangang District, China). The temperature in the equalization tank was maintained using an
immersion chiller (13271-500, VWR, Mississauga, Ontario) and a temperature controller
(61161-300, VWR, Mississauga, Ontario) at (151oC) analogous to the source water temperature
of a brackish water RO system. A float switch (.6A NO/NC POLY, RSF88Y100R, Cynergy3
Components LLC, Garden Grove, CA) actuated a solenoid valve (SV3201, Omega Inc., Laval,
Quebec) to refill the equalization tank with water from the experiment water feed tank. The
equalization tank included continuous pH control using automated acid (HCl) and base (KOH)
addition using two peristaltic dosing pumps (100.PH.030/4 Peristaltic Pump, Williamson Inc.,
35
Brighton, Sussex). Anti-scalant dosing was provided using a peristaltic dosing pump
(100.PH.030/4 Peristaltic Pump, Williamson Inc., Brighton, Sussex).
Figure 3-10: Physical setup of the pilot-scale system.
3.3.2 Pilot-scale Instrumentation
The pilot-scale system also incorporated continuous data monitoring and autonomous
control of system operating parameters. The pilot-scale system used automated control of the
needle valve for maintaining the recovery ratio. Three analog flowmeters measured the feed flow
rate (500-2000 mL/min, FLR1010, Omega Inc., Laval, Quebec), permeate flowrate (200-1000
mL/min, FLR1009, Omega Inc., Laval, Quebec) and the brine flowrate (200-1000 mL/min,
High Pressure
Pump
Cartridge Filter
Equalization Tank
Flowmeters
Conductivity Sensor
RO Pressure Vessel
36
FLR1010, Omega Inc., Laval, Quebec). Two inline conductivity sensors (A1002, EC/pH
Sensors, Boston MA) measured the electrical conductivity, one for the permeate water (0-
500 µS/cm) and one for the feed water (0-10,000 µS/cm). The pressure was measured using a
high-pressure transducer (P51-1000-S-A-I36-4.5V, SSI Technologies, Janesville, WI). The
temperature of the feed water was monitored using a thermistor (TH-44000-NPT Thermistor,
Omega Inc., Laval, Quebec).
Figure 3-11: Labview continuous monitoring interface for the pilot-scale system.
New components were used for the pilot-scale system to ensure both the pilot-scale
system and the lab-scale system could be used alternatively. The only two components that were
shared between the lab-scale system and pilot-scale system were the NI-USB-6343 data
acquisition system and the experimental water tank storage container. The user-interface was
37
implemented in Labview to provide continuous monitoring of the system (Figure 3-11). The
Labview interface provided continuous feedback regarding the system operating conditions
(pressure, temperature, flow rates, recovery ratio). As well, the user-control in Labview (Figure
3-12) allowed for actuation of the various auxiliary systems (anti-scalant dosing pump and
submersible pump).
Figure 3-12: Labview control interface for the pilot-scale system.
3.4 Experimental Methods
The overall experimental method for both the lab-scale system and the pilot-scale system
were analogous. In the case of the lab-scale system there were some additional procedures that
needed to be performed for preparing the membrane coupons. The overlapping methods are
presented followed by the methods that were distinct for the lab-scale and the pilot-scale systems.
3.4.1 Experiment Water Preparation
The water used in the experiments was mixed to mimic brackish groundwater common
of communities requiring solar powered desalination. La Mancalona, Mexico is one of many
38
communities with similar water characteristics, socio-economic conditions, and high solar
insolation. The water from La Mancalona, Mexico was analyzed from a 2 L sample that was
shipped from the community (Table 3-1).
Table 3-1: Water chemistry of La Mancalona, Mexico groundwater, Nobleton, Ontario
groundwater, the experimental MilliQ-based matrix and the experimental groundwater-based
matrix.
Ions in the Water
La
Mancalona,
Mexico
(mg/L)
Nobleton,
Ontario
Deep Well
(mg/L)
Experimental
MilliQ-based
matrix
(mg/L)
Experimental
groundwater-
based matrix
(mg/L)
Sodium (Na+) 164 11 164 164
Magnesium (Mg2+
) 99 26 99 99
Calcium (Ca2+
) 502 64 502 502
Chloride (Cl-) 166 2.3 166 166
Sulfate (SO42-
) 1773 - 1713 1436
Potassium 6.4 - - -
Iron 5.9 - - -
Nitrate as Nitrogen 3.9 - - -
Silicon 23 22 - 22
Strontium 12 0.75 - 0.75
General Characteristics
pH 7.83 7.63 7 7
Alkalinity (As CaCO3) 142.5 1.2 - -
Conductivity 3770 550 4140 3740
Total Dissolved Solids 2262 340 2648 2393
Hardness 1661 270 1660 1660
Dissolved Organic Content 1.3 2.9 - 2.9
Total Ammonia-N 0.37 3.7 - 3.7
Orthophosphate - 0.31 - 0.31
Each lab-scale experiment required approximately 150 L of experiment water and the
pilot-scale experiment required approximately 600 L of experiment water. Due to the remote
location of the partner community, it was unfeasible to ship experimental water volumes to the
experimental lab-scale system. Instead, experimental water matrices were prepared using either
lab-grade MilliQ® water or the local Nobleton, Ontario groundwater as the solvent. For the
39
experimental MilliQ-based matrix, the MilliQ® lab-grade water was prepared using an Elix®
(EMD Millipore Ltd, Etobicoke, Ontario) water purification system and was mixed with
scientific grade inorganic salts. The resultant experimental MilliQ-based matrix was used to
mimic a brackish groundwater. Inorganic salts were added in the various proportions and
dissolved in the lab-grade water. The inorganic salts used were magnesium sulfate, sodium
chloride, and sodium sulfate from Fisher Scientific Company (Fisher Chemical, Fair Lawn, New
Jersey) and calcium sulfate from Anachemia, a VWR Company (VWR, Mississauga, Ontario).
The water analysis of the experimental MilliQ-based matrix mixed with the inorganic salts is
presented in Table 3-1.
The groundwater analysis from the community in La Mancalona, Mexico had a dissolved
organic content of 1.3 mg/L. Since the water volumes required for the experimental program
undertaken were too large to ship from the remote community in Mexico, a local groundwater
well with a similar dissolved organic content was identified from the Ontario government’s
provincial groundwater monitoring program through a partnership with the Toronto Regional
Conservation Authority. The Nobleton, Ontario deep well, (Figure 3-13) was selected since its
historical dissolved organic content was (1.1-3.3) mg/L and was in the closest range of dissolved
organic content of the wells in Ontario’s provincial groundwater monitoring program.
The groundwater was collected throughout the experimental program using several 20 L
water containers (Figure 3-14) to ensure fresh water was used in the experiments. The collected
water was stored in 20 L containers in a walk-in refrigerator at 4oC for a maximum of three
weeks. The water analysis of the Nobleton, Ontario groundwater is presented in Table 3-1 and
shows the concentrations of dissolved minerals and dissolved organic content. The water
chemistry of the experimental groundwater-based matrix that used the Nobleton, Ontario
groundwater as the solvent for inorganic salts is also presented in Table 3-1.
40
Figure 3-13: Nobleton deep well (Nobleton, Ontario) with the a) well's green access box and
b) portable variable frequency drive for the submersible pump located inside the well.
Figure 3-14: Groundwater collection at the Nobleton deep well, Ontario.
3.4.2 Operating Conditions
To evaluate the effects of intermittent operation on membrane fouling, separate
experiments were conducted. For each experiment, fresh membranes were prepared and used
41
following the method described in the respective sections for the lab-scale and pilot-scale. The
intermittent experiments were operated for 8 hours per day with 16 hours of shutdown. The start-
up procedure was to turn on the VFD and increase the VFD frequency to 15 Hz, then the needle
valve on the recirculation line was adjusted until the system operating pressure reached 20.7 bar
and the back pressure of the RO modules was slightly adjusted using the individual needle valves
on each cross-flow cell until the permeate flow rates were similar within ±5%. The recovery
ratio was also adjusted using the needle valve for the waste stream (Figure 3-1, Figure 3-3 and
Figure 3-7). After 8 hours of operation the system was shut down by reducing the frequency on
the VFD to zero. After 16 hours, the system was re-started by slowly increasing the VFD
frequency to 15 Hz. The needle valve on the recirculation line was not re-adjusted nor were the
needle valves on the individual RO modules.
A new set of membranes were prepared in the same process and used in the continuous
experiments. The continuously operated experiments were initiated using the same startup
procedure as the intermittent experiments and then were operated continuously. The runtime was
equal to or longer than the intermittent experiments to enable a fair comparison between the
operating modes. The system operating pressure vs time for three consecutive days of
intermittent operation and of continuous operation are shown in Figure 3-15.
The experimental systems were run with a recirculation loop to enable recovery ratios
representative of the full-scale system. The experimental recovery ratio was selected as 75%
since in full-scale PVRO systems, the brine waste is typically percolated to the environment [33]
and minimal volumes were preferred. In addition, when operating at higher recovery ratios, the
permeate water produced is much greater than at lower recovery ratios leading to lower specific
energy consumption. The experiments were then run in the various operating conditions.
42
Figure 3-15: Operating pressure for the a) intermittent experiments and b) the continuous
experiments for a representative three days of operation.
The three main operating conditions investigated were the use of anti-scalant (BWA
Additives Flocon 135 and Flocon 260) versus no anti-scalant, intermittent (8 hours on, 16 hours
off) vs. continuous (24 hours on) operation and rinsing of the system with lab-grade clean water
vs. no rinsing prior to shutdown. The rinsing volumes (8 L at the lab-scale, 21 L at the pilot-
0
5
10
15
20
0 8 16 24 32 40 48 56 64 72
Ap
plie
d P
ress
ure
(bar
)
Houra)
0
5
10
15
20
0 8 16 24 32 40 48 56 64 72
Ap
plie
d P
ress
ure
(bar
)
Hourb)
43
scale) were selected to be three times the system volume including the tubing, reverse osmosis
module(s), submersible pump and all the system components in-line to provide sufficient volume
of clean water to rinse the system components. In a real system, the rinse would be performed
with permeate water, however, for the experiments, lab-grade water was selected to reduce
variability from the collected experiment permeate water quality.
3.4.3 Cartridge Filter Replacement
During the experiments, the cartridge filters required daily monitoring and were replaced
daily throughout the experiments for both the lab-scale and pilot-scale system. This ensured that
the filters were not clogged during operation of the system. The high pressure pump would be at
risk of damage if the cartridge filter became too blocked with particulates. In an operating system
in the field, a larger cartridge filter would be used and a maintenance schedule would be followed
to prevent clogging.
3.4.4 Lab-scale System Procedures
The experimental method for the lab-scale system required some distinct methods from
the pilot-scale system. First, the membrane coupon preparation was required since membranes
for the lab-scale system were purchased in a flat-sheet roll. Second, after the experiment was
completed, the membrane autopsy was performed in a slightly distinct method.
3.4.4.1 Membrane Coupon Preparation
For each experiment, membrane coupons were cut using a die (Sterlitech Sepa CF Steel
Rule Die, 19 cm x 14 cm) from a membrane flat-sheet roll (40” x 60” Dow Filmtec BW30). The
membrane coupons were soaked in lab-grade water for 24 hours to wet the membranes and to
ensure uniform polymer swelling. The membranes were then placed in the cross-flow cells,
compressed in the pressure vessel to 68.9 bar and the system was operated at an applied pressure
44
of 20.7 bar (pre-compacted) for 24 hours using lab-grade water (MilliQ® water) in a closed loop
maintained at 10oC. After pre-compaction, the experiments were commenced by spiking the
experimental MilliQ-based matrix (5 L in the equalization tank) to the required conductivity to
represent the operating recovery ratio. The experiments were then run in the designated operating
condition.
3.4.4.2 Membrane Autopsy
Membrane autopsy was performed to characterize the dominant fouling mechanisms
using two techniques, adenosine triphosphate (ATP) and scanning electron microscopy (SEM),
described in more detail in Section 3.4.6. Each ATP sample and SEM sample were taken from
the same locations on the membrane coupon, as shown in Figure 3-16. This was done to ensure
the results would be comparable between experiments and various operating conditions.
Figure 3-16: Location of SEM and ATP samples for the lab-scale samples.
45
3.4.5 Pilot-scale System Procedures
The pilot-scale system methodology differed from the lab-scale system since the
membrane was a spiral-wound element. The spiral-wound element fit directly into the housing
and did not require cutting. For the pilot-scale experiment, a new Filmtec TW30-2514 was used.
The Filmtec TW30 used the same BW30 membrane but it was wrapped in tape instead of
fiberglass, which allowed for simple disassembly for the membrane autopsy.
3.4.5.1 Membrane Preparation
The TW30-2514 was prepared by performing pre-compaction. Prior to the start of pre-
compaction, the TW30-2514 was rinsed with 40 L of lab-grade water to remove the wetting
chemicals. Pre-compaction was performed by operating the reverse osmosis system at an applied
pressure of 20.7 bar (pre-compacted) for 24 hours using lab-grade water in a closed loop
maintained at 10oC.
3.4.5.2 Membrane Autopsy
After completion of the pilot-scale experiment, the TW30-2514 was immediately
removed from the high-pressure housing for analysis of the membrane surface using ATP and
SEM. The TW30-2514 spiral-wound membrane element was opened using a sharp knife to
unravel and reveal the membrane surface (Figure 3-17). The location of the ATP sample and
SEM sample on the unraveled membrane are shown in Figure 3-18.
Figure 3-17: Unraveled TW30-2514 membrane for removal of the membrane autopsy samples.
PermeateCollection Tube
Brine Outlet
Feed Inlet
46
Figure 3-18: Pilot-scale membrane autopsy sample locations shown on the unraveled
membrane.
3.4.6 Membrane Characterization
The reverse osmosis membranes investigated in this thesis were analyzed using three
main techniques (pure water permeability, scanning electron microscopy (SEM) and adenosine
triphosphate (ATP) surface deposit analysis). The pure water permeability was measured at the
beginning of the experiment. The membrane autopsy was performed using the other two
techniques, SEM and ATP. These three techniques are described in the subsequent sections.
3.4.6.1 Pure Water Permeability
The pure water permeability provides the maximum flux that the virgin membrane can
achieve. The pure water permeability for the virgin reverse osmosis membrane Filmtec BW30
was determined experimentally. Membranes were pre-compacted for 24 hours using MilliQ®
water at 20.7 bar to allow the permeate flow rate to reach a constant flow rate (±5%). The pure
water permeability was measured by performing pre-compaction followed by adjusting the
applied pressure from 0-20.7 bar in 3.4 bar increments and recording the corresponding permeate
flow rate. The pure water permeability was determined to be (0.57±0.05) L m-2 day-1 kPa-1 also
equivalently as (2.4±0.2) L m-2 h-1 bar-1. In comparison, Antony et. al. found the pure water
PermeateCollection Tube
Brine Outlet
Active Membrane Area
SEM Sample Location
ATP Sample Location
Cut Lines
14”
1cm
9”
2”Feed Inlet
47
permeability of Filmtec BW30 membranes to be (2.5±0.3) L m-2 h-1 bar -1 [117]. Table 3-2 lists
the membrane characteristics of the Filmtec BW30 membranes investigated in this thesis.
Table 3-2: Filmtec BW30 membrane characteristics.
Membrane Pure Water
Permeability
(L m-2 h-1 bar -1)
Contact angle (o) Zeta
potential
(mV)
Average
Roughness
(nm)
Filmtec BW30 2.5±0.3 [117] 61.42±0.56 [118] -8.0±1.2 [119] 45.9±0.2 [118]
3.4.6.2 Scanning Electron Microscopy
Membrane autopsy using SEM was also performed. SEM samples were analyzed to
observe the scales and deposits accumulated on the membrane surface. SEM was performed
using a JEOL JSM6610-Lv Scanning Electron Microscope (JEOL Ltd., Tokyo, Japan). First, the
samples were dried in ambient air for at least 24 hours and then gold sputter-coated to provide a
200 nm thick conductive layer for the SEM’s electron beam.
Figure 3-19: Laser-cut holder for the aluminum disks containing the SEM samples (gold
sputter coated reverse osmosis membrane samples).
48
The surface deposits were imaged to elucidate and evaluate the various fouling
mechanisms. Several aluminum sample holders were machined and a custom-built laser cut case
was made (Figure 3-19) to transport several sets of membrane samples simultaneously to the
SEM and to prevent ambient dust particles from settling on the dried membrane samples.
3.4.6.3 ATP Surface Deposit Analysis
After each experiment was completed, the membranes were immediately removed for
testing the adenosine triphosphate (ATP) on the membranes to evaluate the presence of
biological fouling. ATP is a measure of the living cells present on the membrane surface or in a
sample of water. The solid phase ATP content was measured using a luminometer (Kikkoman
Lumitester C-110, New Brunswick, Canada). Membrane coupons extracted from the membrane
were tested using the Deposit and Surface Analysis protocol [120].
3.5 Conclusion
In conclusion, the experimental systems and methods used in this thesis provided a means
of mimicking the real conditions of an operating plant in a controlled environment at the lab-
scale and at the pilot-scale. The experimental systems were equipped with continuous data
collection of the system parameters to ensure precise quantification of the membrane
permeability as a function of time. As well, the experimental systems were capable of
autonomous control for rinsing of the membranes, and maintenance of a consistent recovery ratio
during the experiments. The experimental methods outlined in this chapter highlight the key
aspects of the experimental program which was undertaken in this thesis.
49
Chapter 4
Membrane Fouling under Intermittent Operation with the Initial Lab-scale System and using the Experimental MilliQ-based Matrix
This work has been previously published (full citation: Freire-Gormaly, M., Bilton, A., (2017)
An Experimental System for Characterization of Membrane Fouling of Solar Photovoltaic
Reverse Osmosis Systems under Intermittent Operation, Desalination and Water Treatment,
vol.73, pp.54-63). It has been reproduced here. The experimental methods were described in
Chapter 3 to avoid repetition. Permission to use this content was secured from the editor.
4.1 Introduction
The initial experimental lab-scale system described in Chapter 3 was used to assess the
membrane permeability decline for the experimental MilliQ-based matrix. The experimental
MilliQ-based matrix was a solution of lab-grade water mixed with lab-grade chemicals to mimic
the concentration of groundwater representative of communities requiring solar powered reverse
osmosis water treatment systems. These initial experiments provided a quantification of the
membrane permeability decline for intermittently operated RO systems compared to
continuously operated RO systems.
4.2 Experimental Program
The experimental method described in Chapter 3 was followed for these experiments
with the experimental MilliQ-based matrix. The experimental MilliQ-based matrix consisted of
lab-grade (MilliQ®) water and lab-grade chemicals. The experiments were performed to test the
effect of the operating conditions (intermittent vs. continuous operation). Table 4-1 lists the
experimental program. The membrane permeability was determined using instrumentation
measurements and Equation 2-1.
50
Table 4-1: List of experiments and operating conditions.
Experimental System Base Water Operating
Condition
Anti-scalant
(AS) Use
Permeate
Rinse
Initial Lab-scale Lab-Grade Intermittent None None
Initial Lab-scale Lab-Grade Continuous None None
4.3 System Operating Conditions
The initial lab-scale system was operated using a recirculation loop and an automated
needle valve to maintain a high recovery ratio of 75%. The membranes were first run in a pre-
compaction cycle with lab grade water for twenty-four hours at 20.7 bar. The intermittent run
was operated for 8 hours per day with 16 hours of shutdown for three days. The start-up
procedure was to turn on the variable frequency drive and increase the operating frequency until
the system operating pressure reached 20.7 bar and slightly adjust the back pressure of the RO
modules until the permeate flow rates were within ±0.5mL/min and about 8 mL/min. The
recovery ratio was also adjusted using the needle valve for the waste stream. The intermittent
run was operated for three consecutive days. After 8 hours of operation, the system was shut
down by manually reducing the frequency on the variable frequency drive to zero. After 16
hours, the system was re-started by manually increasing the variable frequency drive to 15 Hz.
To compare the membrane fouling to a continuous run, the system was operated continuously
for 24 hours. This was also the runtime for the intermittent experiments to enable a fair
comparison between the two operating conditions. A new set of membranes were prepared and
used in the continuous run.
4.4 Effect of Intermittent Operation vs. Continuous Operation
The membrane permeability was calculated using the measured permeate flowrates, the
operating pressure, and feed water conductivity. The time-axis of the experimental results
51
represents the operating time of the experiment. The hours when the experimental system was
shut down were not included in the time-axis to clearly compare the two operating conditions.
Figure 4-1: Membrane permeability as a function of operating time for the intermittent run.
The hours when the system was shutdown are not included.
The permeability data set for the intermittent run, shown in Figure 4-1, was averaged for
each thirty minutes of operation. The pressure regulator was engaged during the intermittent
experimental run and continuous experimental run maintaining a consistent transmembrane
pressure (Figure 4-2). The recovery ratio was also maintained at an average recovery ratio of
75%. Overall, in Figure 4-1 there is a consistent trend of decreasing permeability. For the
intermittent run, there was a rapid decrease in membrane permeability that can be observed
immediately after the first time the system was shut down for 16 hours (shown in Figure 4-1 at
52
hour eight). The permeability decreased rapidly to below 4x10-7m-1s bar-1 by the end of 24 hours
of intermittent operation.
Figure 4-2: Transmembrane pressure for the intermittent and continuous experimental run.
There was a small initial increase in the membrane permeability after start-up and then a
decrease in permeability at the start of Day 3 for SEPA 2 and SEPA 3. Since the results in Figure
4-1 were averaged over thirty minutes, initial increases in membrane permeability after start-up
were not shown. The initial increase in permeability is anticipated to be due to the rapid removal
of salts that had accumulated on the surface of the membrane during the time when the system
was shut down. The slow decrease is likely due to fouling during the operating time from scaling.
The permeability in SEPA 1 dropped less rapidly than in SEPA 2 and SEPA 3 (Figure 4-1). This
indicates a lower rate of fouling on SEPA 1’s membrane. This could be due to unit to unit
variation or due to an unequal division of flow in the piping flow manifold. Further experiments
53
and refinements of the experimental system design were undertaken and are described in detail
in Chapter 5.
The initial continuous experimental run was operated for approximately 24 hours. Figure
4-3 shows the calculated membrane permeability for the continuous experimental run. The
membrane permeability decreased during continuous operation. The membrane permeability for
SEPA 2 and SEPA 3 show a similar trend of decline. SEPA 1 has a distinctive pattern of decline
with a sharp linear decrease starting at the sixth hour of the experiment, the permeability declined
at a steeper rate than SEPA 2 and 3 from this point onwards to a minimum at the 24th hour of the
experiment.
Figure 4-3: Membrane permeability vs. operating time for the continuous run (averaged over
30 min intervals).
The rapid sharp decline in permeability for SEPA 1 may have been caused from unequal
division of flow in the parallel manifold or from rapid fouling of the membrane in SEPA 1. In
54
subsequent experiments in Chapters 5 and 6, the improved lab-scale system was used which had
a larger footprint, improved parallel manifold of stainless steel and smaller needle valves for
improved control. The pressure throughout the experiment was very stable because the pressure
regulator was engaged from the beginning of the experiment, as shown in Figure 4-2. The needle
valves which pressurize the SEPA cells were set initially when all three SEPA cells had a similar
permeability and were not adjusted throughout the duration of the experiment.
A comparison of the initial intermittent and continuous run is shown in Figure 4-4.
Comparing the calculated permeability values, there is a difference between the two operating
conditions. This could be due to the severe pressure fluctuations observed for the intermittent
run compared to the continuous run (Figure 4-2).
Figure 4-4: Comparison of initial continuous and intermittent experiments.
55
Despite these variations, these initial results indicate that increased fouling due to scaling
for intermittent operation may not be as significant as anticipated. Previous experimental tests
showed even larger variations between the SEPA cells, and this current setup has resolved large
variations. However, small variations still exist between SEPA cells due to variability between
the membrane fouling rates. This variability originally motivated performing the experiments in
triplicate. Additional experiments with the improved experimental lab-scale setup (with
improvements to ensure a steady pressure) were performed and described in Chapter 5 to more
accurately quantify the difference in fouling rates between the intermittent and continuously
operated systems.
4.5 Discussion
This chapter presented an approach to evaluate the effects of intermittent operation with
extended shutdown periods on membrane fouling in PVRO systems, which had never been
previously quantified. To evaluate this effect, a custom experimental system was designed and
constructed. The system consisted of three stainless steel SEPA cross-flow reverse osmosis
membrane cells connected in parallel, and was equipped with computer-controlled valves and
pumps to autonomously test different operating conditions. The system was also instrumented
with pressure, flow, temperature, and conductivity sensors to characterize membrane
permeability decline.
These initial experiments had several challenges that were addressed before proceeding
with the experiments detailed in Chapter 5. The initial experiments had large pressure
fluctuations caused by the accumulation of air in the feed water line after the cartridge filter. To
resolve the accumulation of air in the feed water line, the improved experimental lab-scale
system used a submersible pump directly from the equalization tank to overcome the pressure
56
drop associated with the polyspun cartridge filter. As well, during the initial experiments, the
flowrate of the gravity fed float valve frequently fluctuated. To resolve this issue the gravity fed
float valve was replaced with a float switch which activated a solenoid valve directly from the
experiment water tank. As well, the Labview code which controlled the activation of the
automated needle valve was improved to take an average over 30 seconds of the instantaneous
recovery ratio and the valve was only adjusted if the recovery ratio fell outside of the upper and
lower limit (±5 % of the target recovery ratio). The Pelton wheel flowmeters were also replaced
with pressure-based flowmeters. Finally, the physical system setup was adapted to a larger
footprint and a new flow manifold was implemented with stainless steel tubing instead of flexible
stainless-steel tubing to ensure equal flow division of the feed flow to the three SEPA cross-flow
cells.
The initial experiments were conducted using the initial lab-scale system to evaluate the
effects of intermittent operation for brackish water without organic content. It was found in these
initial experiments that the intermittent operation did not have a significant impact on membrane
permeability. These experiments contradicted previous claims in literature that membrane
fouling is greatly accelerated by intermittent operation. However, it should be noted that these
experiments were only conducted for a short period of time and with pressure fluctuations. These
aspects were evaluated in more detail in the experimental results presented in Chapters 5 and 6.
Based on these initial experimental results, it is hypothesized that the intermittent
operation increased fouling by two main mechanisms: scaling and biofouling. Fouling by scaling
is hypothesized to be a dominant factor for the intermittent operation since the water in the SEPA
cross-flow cells remains stagnant over a 16-hour period. During this 16-hour period existing
nucleation sites would have substantial time for crystal growth. Although biological content was
kept to a minimum using the UV lamp in the experiment water tank, there was evidence when
57
the flowmeters were disassembled that there was minimal biological growth in the system from
slime build-up on the various internal components of the flowmeters. Further experiments
described in Chapter 5 were performed to evaluate the fouling mechanisms for intermittent
operation and the effects of intermittent operation over a longer operating period with the
improved experimental system.
4.6 Conclusions
These initial experiments were conducted using the initial lab-scale system to evaluate
the effect of intermittent operation compared to continuous operation for a lab-mixed brackish
water. The results showed that the intermittent operation did not cause a significant decline in
membrane permeability compared to the continuously operated experiment. Further experiments
described in Chapter 5 were performed using the improved experimental lab-scale system. The
experiments in Chapter 5 were performed to evaluate the effect of intermittent operation, anti-
scalant pre-treatment and rinsing of the membranes prior to shut down on the membrane
permeability over a longer operating period (three to six days).
58
Chapter 5
Membrane Fouling Characterization at the Lab-scale using the Experimental MilliQ-based Matrix
This work has been previously published (full citation: Freire-Gormaly, M., Bilton, A., (2017)
Experimental Quantification of the Effect of Intermittent Operation on Membrane Performance
of Solar Powered Reverse Osmosis Desalination Systems, Special Issue: Desalination, vol.435,
pp. 188-197.). The experimental methods were described in Chapter 3. The remaining parts of
the published work have been reproduced here. Permission to use this content will be secured
from the editor after final publication of the article.
5.1 Introduction
Experiments were performed using the improved lab-scale system to determine the effect
of various operating conditions (intermittent operation, anti-scalant pre-treatment and rinsing of
the membranes prior to shut down) on membrane permeability. The experiments in this chapter
were operated for a longer time period (three to six days) compared to the initial experiments
(three days) reported in Chapter 4. This chapter presents the results for the lab-scale experiments
when the experiments were performed with the experimental MilliQ-based matrix. These
experiments also investigated the effect of using anti-scalant pre-treatment. The graphs in do not
state the type of anti-scalant since all the graphs represent the results for anti-scalant F135.
5.2 System Operating Conditions
Consistent experimental conditions were applied in all of these experiments. The
pressure set at 20.7 bar and was maintained using a pressure regulator on the high-pressure pump.
The system was operated at a cross-flow velocity of 0.1 ms-1 to match the conditions in a pilot-
plant operating with a spiral-wound reverse osmosis element. The membrane flux ranged from
36.0 Lm-2h-1 at the beginning of the experiment to 13.5 Lm-2h-1 at the end of the experiments.
This is below the manufacturer’s recommended flux rate for a TW30-2514 spiral-wound RO
module (55 Lm-2h-1). The recovery ratio was set to 75% and was maintained using an automated
59
needle valve on the brine line. The high recovery ratio was selected to minimize the brine which
in a real operating pilot-plant would typically be percolated to the environment with minimal
treatment [33]. The equalization tank was refilled using a solenoid valve triggered by a float
switch to maintain a consistent conductivity and level in the tank. The system was instrumented
to allow for continuous data collection of the system conditions and for autonomous control of
the system’s recovery ratio.
Figure 5-1: Experimental system operational conditions: a) permeability, b) pressure, c)
recovery ratio, d) salt rejection.
System data from a sample day during a lab-scale experiment for intermittent operation
with anti-scalant and no rinse is shown in (Figure 5-1). There is good agreement between the
three SEPA cells, providing results in triplicate for each experiment (Figure 5-1a). Conditions
were consistent throughout the experiment, however, there were a few deviations. Pressure was
60
consistently maintained at 20.7 bar (Figure 5-1b). Despite the automated control on the system
to maintain the system conditions, there was some variability in the recovery ratio (Figure 5-1c)
due to the automated needle valve adjusting to the required recovery ratio.
5.3 Experimental Program
The experimental method described in Chapter 3 was followed for these experiments
with the experimental MilliQ-based matrix which consisted of lab-grade (MilliQ®) water as the
solvent for lab-grade chemicals. The experiments were performed to test the effect of the
operating conditions, pre-treatment, and a permeate rinse at shutdown. Table 5-1 lists the
experimental program for the results presented in this chapter. The pre-treatment investigated in
these experiments was a commercially available anti-scalant (Flocon 135, BWA Additives)
designed to minimize scale formation.
Table 5-1: List of experiments and operating conditions.
Experimental
System
Operating
Condition
Anti-scalant
(AS) Use
Permeate
Rinse
Improved Lab-scale Intermittent None None
Improved Lab-scale Continuous None None
Improved Lab-scale Continuous With F135 None
Improved Lab-scale Intermittent With F135 None
Improved Lab-scale Continuous With F135 None
Improved Lab-scale Intermittent With F135 With Rinse
The normalization of the membrane permeability was performed by dividing the
instantaneous membrane permeability by the average value over the first five minutes of the
experiment. This limited the effect of membrane to membrane variability. The time axis of the
experimental results represents the operating time of the experiment. The hours when the
experimental system was shutdown were not included in the time-axis to clearly compare the
various operating conditions.
61
5.4 Effect of Anti-scalant Pre-treatment
Pre-treatment methods in industrial-scale reverse osmosis systems can include several
steps. However, for resource-constrained communities, minimal pre-treatment is usually done to
minimize system complexity and cost. The results show the average membrane permeability of
the three SEPA cross-flow cells. Figure 5-2 compares the normalized permeability for
continuous experiments ‘with’ and ‘without’ anti-scalant F135 addition using the experimental
MilliQ-based matrix. The normalization exceeded one for ‘continuous with anti-scalant and no
rinse’ between hour zero and hour four likely due to the membranes compacting slightly during
operation. It should be noted that no rinsing was implemented in either case.
Figure 5-2: Normalized membrane permeability for continuous operation with and without the
use of F135 anti-scalant.
62
The experiment ‘with no anti-scalant’ decreased very rapidly when compared to ‘with
anti-scalant’ at the high recovery ratio of 75%. The effect of anti-scalant showed that anti-scalant
is required for operating these reverse osmosis systems at a high recovery ratio of 75%. As a
result, most of the other experiments were operated with anti-scalant since without it, the
experiment duration was too short to investigate simple remedial actions that could potentially
restore membrane permeability, such as rinsing with permeate water.
5.5 Effect of Intermittent Operation vs. Continuous Operation
The effect of intermittent operation versus continuous operation when no rinsing was
performed was investigated to determine the effect on the membrane permeability. The results
in Figure 5-3 show the average membrane permeability of the three SEPA cross-flow cells.
Figure 5-3 compares continuous to intermittent operation, in both experiments they were
operated with F135 anti-scalant and without rinsing.
The experimental results show that intermittent compared to continuous operation did
not cause a significant decrease in membrane permeability over this experiment duration at the
high recovery ratio (Figure 5-3). Although the permeability slightly increased at the start of the
8th hour of operation for the intermittently operated experiment, overall the permeability decline
of the intermittently operated experiment was comparable to the continuously operated
experiment. This is hypothesized to be a result of insignificant crystal growth on the membrane
surface during the extended shutdown period since anti-scalants were used. The increase in
membrane permeability at the start of the new day can be seen clearly in the intermittent
operation since these results are presented with the continuous data. The results in Chapter 4
were the average over thirty minutes of data and the system had several fluctuations in pressure
and flow rate that were resolved with the improved lab-scale system.
63
Figure 5-3: Normalized membrane permeability of continuous operation and intermittent
operation with F135 anti-scalant.
5.6 Effect of Anti-scalant F135 for Intermittent Operation with no Rinse
Intermittent operation with anti-scalant F135 maintained a consistently higher membrane
permeability than experiments operated without anti-scalant (Figure 5-4). The anti-scalant
reduced the onset of scale formation, thereby improving the membrane performance. There was
a visible improvement in membrane permeability at the start of the new day (at hour 8) for both
with and without anti-scalant usage. This is hypothesized to be a result of the local concentration
in the cross-flow reverse osmosis cell. At shutdown, the permeate water remaining in the tubing,
was able to flow back into the cross-flow cell through osmosis. At the membrane surface, there
64
is likely a localized decrease in salt concentration which permits some scales to dissolve and
become detached. As well, the start-up allows for a fast flow rate and air to scour the membrane
surface which can further contribute to the increased membrane permeability at the start of the
new day (at hour 8 and 16) (Figure 5-4).
Figure 5-4: Normalized permeability decline for intermittent operated experiment both with
and without anti-scalant and without rinsing.
5.7 Effect of Rinsing with Anti-scalant F135 for Intermittent Operation
Figure 5-5 compares the effect of permeate rinsing for intermittent operation with anti-
scalant addition. Rinsing had a significant improvement on the membrane permeability as a
function of time. Without rinsing the membrane, performance reduced significantly by the
65
second day of intermittent operation, but with rinsing, high permeability was still observed after
seven days of operation. This is hypothesized to be a result of the washing away of residual salts
and anti-scalant which otherwise remained on the surface of the membrane. For the case with
rinsing, the membrane sits in clean water overnight, potentially dissolving some mineral scaling.
Figure 5-5: Normalized membrane permeability of decline for intermittent operated experiment
with anti-scalant when operated with rinsing and without rinsing.
5.8 Combined Comparison of the Effect of Operational Conditions on Membrane Permeability
Figure 5-6 shows the average normalized permeability (for all three SEPA cells) at the
fourth hour for all the experimental conditions for each individual day. Each individual
experiment for the tested operating condition was ended when the permeate flow rate dropped
66
below 30% of the initial flow rate. The exception was for the intermittent operation with anti-
scalant and rinsing experiment which was ended once it was determined that it maintained high
membrane permeability well beyond the operating time of the other experiments. The trend of
consistent permeability for intermittent operation with anti-scalant and with rinse can be clearly
seen in the comparative bar graph. Figure 5-6 shows that for intermittent operation with anti-
scalant use and 8 L of permeate rinsing, the permeability declined only slightly to (87±9) %
while all the other operating conditions declined to zero except continuous operation with anti-
scalant and no rinse which reached (30±4) % by Day 5. The results also show that without anti-
scalant usage and without 8 L of permeate rinsing when operated continuously, the membrane
permeability decreased significantly by day 3 to (13±9) %. Intermittent operation with anti-
scalant use and rinsing provided consistent membrane permeability for six days of operation,
while all the other operating conditions, intermittent without rinsing and continuous operation
with or without anti-scalant decayed much more rapidly.
Figure 5-6: Comparative bar graph of the experimental conditions investigated, the
permeability shown is the permeability at the fourth hour of the day.
67
5.9 Combined Comparison of the Effect of Operational Conditions on Salt Rejection
The average salt rejection was examined for the operating conditions (Figure 5-7). For
these experimental conditions, a decreasing trend of salt rejection was observed except for
intermittent with anti-scalant and rinsing. The experimental conditions without anti-scalant (both
intermittent and continuous operation) showed the largest decline in salt rejection. The
significant decline in salt rejection when no anti-scalant was used is hypothesized to be a result
of the large amount of gypsum crystals which grew on the surface of the membranes, greatly
increasing the localized salt concentration on the feed side of the membrane. When the
experiment was operated intermittently with anti-scalant and with rinsing, the salt rejection was
maintained with minimal decline.
Figure 5-7: Comparative bar graph of the average salt rejection for the experimental conditions
investigated.
68
5.10 Combined Comparison of the Effect of Operational Conditions on the Membrane Autopsies
After each experiment, a membrane coupon was imaged using SEM (Figure 5-8 and
Figure 5-9). A summary of the surface features observed on the membrane coupons (Figure 5-8
and Figure 5-9) shows a visible reduction in surface scales when both anti-scalants were used
and rinsing for intermittent operation (Figure 5-8d). This coincides with the high permeability
still present for this case after the experiments.
Figure 5-8: Comparative SEM images of the various operating conditions at x 1000
magnification: a) clean un-used membrane, b) continuous with anti-scalant, c) continuous
without anti-scalant, d) intermittent with anti-scalant and with rinse, e) intermittent with anti-
scalant and and no rinse, f) intermittent without anti-scalant.
69
Mineral scales are minimal and were not the main cause of the decrease in membrane
permeability. However, in all other cases, where the permeability decreased, large crystals were
formed on the membrane surface, greatly reducing the membrane permeability. The membrane
coupon cross-sections (Figure 5-9) also show a visible reduction in the thickness of the surface
scales when anti-scalant F135 and rinsing was used for intermittent operation and a large increase
in scale thickness when compared to the clean membrane.
Figure 5-9: Comparative SEM cross-sections of the various operating conditions at x 500
magnification: a) clean un-used membrane, b) continuous with anti-scalant, c) continuous
without anti-scalant, d) intermittent with anti-scalant and with rinse, e) intermittent with anti-
scalant and and no rinse, f) intermittent without anti-scalant and without rinse.
70
The average ATP on the membrane surface for each of the experimental factors is shown
in Figure 5-10. The concentration of ATP (cATP), measured using the luminometer was
normalized to the concentration of ATP in the respective experiment feed water. The
concentration of ATP on the membrane surface after the experiments showed significantly lower
biological activity for intermittent operation than for continuous operation. During these
experiments, the experimental MilliQ-based matrix was used consisting of lab-grade chemicals
dissolved in lab-grade water. There was minimal organic content. Biological films nor cake
layers were observed in the SEM membrane autopsy (Figure 5-8). As a result, it is expected that
there was minimal biological fouling in these experiments. Further experiments presented in
Chapter 6 explored the effect a real groundwater with organic content had on the membrane
fouling results.
Figure 5-10: Concentration of ATP on the membrane surface for various operating conditions.
71
5.11 Conclusions
This chapter presented an experimental study on the effects of intermittent operation,
anti-scalant F135 addition, and permeate rinsing commonly seen in renewable energy powered
reverse osmosis systems. The comparison of intermittent operation and continuous operation,
as well as anti-scalant usage and membrane rinsing with permeate water showed that minimizing
membrane permeability decline is most dependent on performing a rinse of the membranes with
permeate water.
The experimental results showed that intermittent operation alone did not have a
significant negative impact on the membrane performance in the short-term (several days of
operation in the cross-flow unit). Anti-scalants were observed to improve the membrane
performance when used in isolation for intermittent operation. Rinsing the membranes with 8 L
of permeate water prior to shut down when anti-scalant was used had a significant improvement
on the membrane performance during intermittent operation. Membrane autopsy using scanning
electron microscopy showed the fewest scale deposits for intermittent operation with anti-scalant
and 8 L of permeate rinsing. For this case, on the sixth day of operation, the average normalized
permeability declined only slightly to (87±9) % for intermittent operation with anti-scalant and
with rinse; while all the other operating conditions declined to nearly zero except continuous
operation with anti-scalant (30±4) %.
This topic requires more investigation with a wider array of water sources. Further
experiments presented in Chapter 6 used an experiment water containing a real groundwater with
organic content. The experiments in this chapter were conducted with a lab-mixed brackish water
without organic content, over a short period of time, and in cross-flow cells with much smaller
membrane areas than in real systems. However, this work shows that systematic experimental
72
studies on conditions encountered for renewable powered desalination system are required for
cost-effective system operation. These experimental results represent an initial set of studies to
quantify the effect of intermittent operation and various simple interventions which may be taken
to improve the membrane performance for small-scale solar powered RO systems.
This chapter presented the results of experimental studies to be able to take into
consideration membrane fouling for intermittent operation for the system design of future
renewable powered reverse osmosis systems. This will enable the development of better design
algorithms that consider the membrane fouling in the design of solar powered water treatment
systems. The improved systems design considering membrane fouling was explored in Chapter 7
with the development of an optimization framework for PVRO systems. These results can also
be useful for other renewable powered desalination systems which exhibit intermittency due to
the inherent intermittent nature of the renewable power source (e.g. wind, tidal, wave).
73
Chapter 6
Membrane Fouling Characterization at the Lab-scale and Pilot-scale using the Experimental Groundwater-based Matrix
This work is in preparation for publication (full citation: Freire-Gormaly, M., Bilton, A., (2018)
Experimental Lab-scale and Pilot-scale Characterization of the Effect of Intermittent Operation
on Membrane Fouling for Solar Powered Reverse Osmosis Desalination Systems, Desalination).
It has been reproduced here. The experimental methods were described in Chapter 3 to avoid
repetition. Permission to use this content will be secured from the editor.
6.1 Introduction
Since systems in the field operate with brackish groundwater, the experiments in this
chapter were performed with a more realistic experiment water to elucidate the underlying
fouling mechanisms under intermittent operation. In addition, the effects of typical membrane
fouling mitigation used for PVRO systems, permeate rinsing and anti-scalants, were
experimentally evaluated using the experimental groundwater-based matrix. Furthermore, since
the previous experiments were performed at the lab-scale on smaller membrane coupons to
minimize water requirements, the agreement with larger spiral wound membranes was
experimentally investigated at the pilot-scale. Verification that the pilot-scale experiment match
the lab-scale system will allow for extrapolation of the lab-scale results to guide the design of
full-scale systems.
6.2 Experimental Hypotheses
6.2.1 Effect of Intermittent vs. Continuous Operation on Membrane Fouling with Anti-scalant Usage at the Lab-scale
The effect of intermittent operation compared to continuous operation is hypothesized to
increase the fouling rate, leading to lower membrane permeability compared to a continuously
operated RO system. The main mechanism is hypothesized to be due to the membrane being in
74
contact with stagnant water during the shutdown time, exacerbating biofouling, since biological
growth would have time to colonize. In addition, it is hypothesized that the anti-scalant may
increase the growth of biological content by providing food for micro-organisms [121]. The
extended duration of the shutdown period for 16 hours could also have an influence on
biological growth, since there would be large fluctuations in nutrient supply, this could cause
stress in the biofilm and result in detachment and starvation in parts of the biofilm [122]. It is
also hypothesized that the initial time post-shutdown may cause osmotic suck-back causing
water to flow from the permeate channel to the feed channel of the membrane. This could also
serve as a localized concentration reduction at the membrane interface immediately after shut
down. It is expected that this localized concentration reduction could reduce the likelihood of
significant crystal growth during shutdown.
6.2.2 Effect of Permeate Rinsing on Membrane Fouling with Anti-scalant Usage at the Lab-scale
The effect of permeate rinsing prior to shut down is hypothesized to reduce biofouling
caused by the presence of stagnant anti-scalant, which can act as a source of nutrients for
biological growth [121] since the stagnant water would be predominantly the rinse water. It is
expected that the permeate rinse will not be able to remove crystals firmly attached to the
membrane nor biofilm which may have securely adhered to the surface of the membrane.
Similarly, it is hypothesized that the permeate rinse would not be able to detach surface crystals,
or firmly attached biological growth. The permeate rinse is hypothesized to reduce the presence
of bulk crystals by providing a very low concentration stagnant water which could slightly
dissolve scales during the shutdown period.
75
6.2.3 Effect of Experimental System on Membrane Fouling
The effect of the lab-scale versus pilot-scale system on the normalized membrane
permeability decline is hypothesized to have a minimal effect. It is hypothesized that the two
experimental systems will behave analogously. This is anticipated because both experimental
systems were designed to operate at the same cross-flow velocity and operating conditions.
6.3 Experimental Program to Test Hypotheses
The experiments conducted to test the hypotheses are outlined in Table 6-1. These
experiments will be used to evaluate the effect of intermittent operation compared to continuous
operation at the lab-scale for the experimental groundwater-based matrix. As well, these
experiments will allow for analyzing the effect of a permeate rinse post-shutdown to restore the
membrane permeability for the intermittently operated lab-scale system. Finally, these
experiments will permit a comparison of the lab-scale experimental results to the pilot-scale
system results. This comparison is required for the design optimization to extrapolate the lab-
scale findings to full-scale RO spiral wound element modules.
The experimental method detailed in Chapter 3 was followed for these experiments. The
experimental MilliQ-based matrix used in these experiments was a real groundwater from the
Nobleton, Ontario deep well as the solvent for lab-grade chemicals. This was done to achieve a
water chemistry with dissolved organic content and dissolved minerals to represent brackish
groundwater commonly present in remote communities that require solar powered reverse
osmosis systems. The details about the water collection and groundwater analysis were presented
in Chapter 3, Section 3.8.1. The pre-treatment anti-scalant investigated for these experiments
was selected to be Flocon 260 (F260) from BWA Additives because it was designed to
counteract high foulant conditions and high scaling conditions. In contrast, the Flocon 135, used
76
in the experiments presented in Chapter 5, was designed solely for high scaling conditions. Since
the experiments in this chapter were performed with the experimental groundwater-based matrix
containing dissolved organic content, the F260 anti-scalant was better suited to counteract both
the mineral and organic foulants.
The membrane permeability was measured continuously during operation for all the
experiments conducted in this chapter. The salt rejection was also measured using the feed
conductivity and the permeate conductivity. The amount of biologic content was also measured
using an aggregate measure of the adenosine triphosphate (ATP) using the deposit surface
analysis of the reverse osmosis membrane after the experiment. Membrane autopsy using
scanning electron microscopy (SEM) was also conducted to visually compare the foulant
coverage and morphology on the membranes.
In all analyses, the membrane permeability was normalized using the same method as in
Chapter 5, by dividing by the average value of the membrane permeability over the first five
minutes of the experiment. This limited the effect of membrane to membrane variability.
Table 6-1: List of experiments and operating conditions.
Experimental
System
Operating
Condition
Anti-scalant
(AS) Use
Permeate
Rinse
Lab-scale Continuous With F260 None
Lab-scale Intermittent With F260 None
Lab-scale Intermittent With F260 With Rinse
Pilot-scale Intermittent With F260 None
6.4 Experimental Results
6.4.1 Lab-scale Membrane Permeability Decline
The effect of intermittent operation compared to continuous operation for the F260 anti-
scalant without rinsing (Figure 6-1) shows that there was some initial improvement in membrane
77
permeability at the beginning of each day (Day 3 - Day 7). These results (Figure 6-1) show that
without rinsing, anti-scalant pre-treatment alone cannot restore membrane permeability.
Continuous operation results in a similar decline in membrane permeability as intermittent
operation for the F260 anti-scalant, though there is a minimal increase in membrane permeability
at the beginning of each day. The membrane flux at the lab-scale ranged from 36.0 Lm-2h-1 at
the beginning of the experiment to 13.5 Lm-2h-1 at the end of the experiments. This range was
below the manufacturer’s recommended flux rate for a TW30-2514 spiral-wound RO module
(55 Lm-2h-1).
Figure 6-1: Normalized membrane permeability for F260 anti-scalant when operated
continuously and intermittently.
78
The experimental results show that when the F260 anti-scalant was used and a remedial
rinse with permeate water was performed at the end of the day, the membrane permeability was
effectively restored (Figure 6-2). The error region represents the 95% confidence interval of the
mean of three cross-flow cells. The normalized membrane permeability for intermittent with
rinse recovered 20% percent from the end of the previous day to the start of the next day for
Day 1 - Day 5. Similarly, the normalized membrane permeability for intermittent without rinse
recovered on average 20% from the end of the previous day to the start of the next day for
Day 2 - Day 7. In contrast from Day 1 to Day 2 for intermittent without rinse, the normalized
membrane permeability ended at Day 1 at 78% and started on Day 2 at 80%. This shows the first
day with no rinse had a large effect on the membrane permeability decline.
Figure 6-2: Normalized membrane permeability for intermittent operation and F260 anti-
scalant with and without rinse. The rinse significantly improved the membrane permeability.
79
The comparative graph for the operating conditions (intermittent, continuous) and pre-
treatment technologies (Figure 6-3) shows the effect intermittent operation had on the
normalized membrane permeability for each day of the experiments. The error represents the
95% confidence interval of the mean of three cross-flow cells. The daily value of the normalized
membrane permeability is from the fourth hour of operation and averaged between the three
cross-flow SEPA cells. The fourth hour of operation was selected to permit a comparison
between the daily decline of the normalized membrane permeability for the three experiments.
Figure 6-3: Normalized permeability for the various operating conditions investigated at the
fourth hour of each day for day-to-day comparison.
For the experiment with F260 anti-scalant and no rinsing, the membrane permeability
decreased at about the same rate as continuous operation when no rinsing was used for
intermittent operation. The only operating condition and pre-treatment that maintained
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1 2 3 4 5 6 7
No
rmal
ized
Per
mea
bili
ty
Days
Continuous F260 Intermittent F260 No Rinse Intermittent F260 Rinse
80
membrane permeability above 70% of the initial membrane permeability was the use of the F260
anti-scalant and with rinsing at the end of each day prior to shutdown, indicating this would be
appropriate pre-treatment for real applications.
6.4.2 Lab-scale Salt Rejection
The effect the operating conditions (intermittent, continuous), pre-treatment (F260 anti-
scalant) and rinsing (no rinse, rinse) had on the average daily salt rejection can be seen in Figure
6-4. The average salt rejection for the three SEPA cross-flow cells was averaged over each day.
The error bars represent the standard deviation of the salt rejection over each day. The
experiments that were operated intermittently with the F260 anti-scalant and permeate rinsing
provided the highest salt rejection.
Figure 6-4: Salt rejection for the groundwater experiments with F260.
99.0
99.1
99.2
99.3
99.4
99.5
99.6
99.7
99.8
2 3 4 5 6 7
Salt
Rej
ecti
on
(%
)
Days
Continuous F260 Intermittent F260 No Rinse Intermittent F260 Rinse
81
The experiment operated with the F260 anti-scalant and no rinsing performed worse than
with the F260 anti-scalant and rinsing. In addition, the experiment with the F260 anti-scalant and
no rinsing had the largest variability over the span of each day. This is likely due to the extended
shutdown period where the concentration next to the membrane surface declines slightly due to
forward osmosis from the permeate collection tubing into the cross-flow cell. The continuous
operation with the F260 anti-scalant had the lowest salt rejection of the three experiments. This
is likely because there were no extended shutdown periods where the localized concentration
next to the membrane surface decreased below the feed water concentration.
6.4.3 Lab-scale Membrane Autopsy
The lab-scale membrane autopsy using SEM (Figure 6-5) shows different surface
features for the different operating conditions. The membrane that was operated continuously
with F260 anti-scalant (Figure 6-5a) had uniform small crystals dispersed evenly on the surface
(on average about 1 µm or less in diameter) of calcium, sulfur and magnesium, confirmed by X-
ray electron dispersive spectroscopy (EDS). The membrane that was operated intermittently with
F260 anti-scalant and no rinse (Figure 6-5b) had small and medium-sized crystals, however, the
texture was quite different, since there was a layer on top and under the crystals of varying sizes.
This is hypothesized to be a biofilm that occurred because the anti-scalant remaining in the cross-
flow cell after shutdown provided a source of food initially for bacterial growth [121]. Due to
forward osmotic flow driven by the concentration gradient, there is likely sufficient oxygen for
bacteria to survive temporarily after the system is shutdown.
82
Figure 6-5: Membrane autopsy of the lab-scale experiments at x 1000 magnification and the
length bar represents 20 µm (a-c). Also, the membrane autopsies are shown at x 3000
magnification and the length bar represents 5 µm (d-f). For the continuous with F260 anti-
scalant (a, d), intermittent with F260 anti-scalant and no rinse (b, e) and intermittent with F260
anti-scalant and with rinse (c, f) experiments.
a)
20 µm
b)
20 µm
c)
20 µm 5 µm
5 µm
5 µm
d)
e)
f)
83
For the membrane operated intermittently with F260 anti-scalant and with rinse (Figure
6-5c) the membrane had small and medium-sized crystals. There were some regions of the
membrane surface that were not covered in dense crystals. This shows the rinse was effective in
washing away the residual anti-scalant and groundwater from the membrane surface. The anti-
scalant did not act as a food source, and therefore the biological content was less prevalent and
the dominating source of fouling was by scaling.
The average concentration of ATP on the membrane surface from the three SEPA cross-
flow cells for each of the experimental factors tested with the groundwater is shown in Figure
6-6. The error bars represent the variability between the three membranes from the three SEPA
cross-flow cells. The concentration of ATP (cATP) on the membrane surface after the
experiments showed significantly lower biological activity for continuous operation than
intermittent operation.
Figure 6-6: Concentration of ATP on the membrane surface for lab-scale operating conditions
with anti-scalant F260.
0
1
2
3
Continuous Int+No Rinse Int+Rinse
Mem
bra
ne
cATP
no
rmal
ized
to
fe
ed w
ater
cAT
P (
x10
00
)
84
The highest biological activity was observed on the membrane operated intermittently
but without a rinse. This is hypothesized to be a result of the shutdown period providing
sufficient time for the biological colony to use the anti-scalant and dissolved organic content in
the water as sources of food. The permeate water in the tubing can also flow into the cross-flow
cells through forward osmosis due to the concentration gradient which may provide sufficient
oxygen for the colonies to replicate during the shutdown period. While in the case of the rinse
and anti-scalant use, the membrane surface is swept clean with permeate water prior to the
extended shutdown period. This rinse washes away residual anti-scalant or nutrients. The
continuously operated system is hypothesized to have the least biologic activity because the
system is never stagnant and there is a consistent cross-flow sweeping the surface of the
membrane. The biologic activity grows but reaches a steady-state due to detachment caused by
the turbulent flow induced by the feed spacers. This is consistent with a previous study, which
showed the attachment and detachment of biological growth reaches a steady state [72] .
6.4.4 Discussion for the Lab-scale Experimental Results
6.4.4.1 Effect of Intermittent vs. Continuous Operation on Membrane Fouling with Anti-scalant Usage
Overall, the membrane permeability declined slightly faster when operated intermittently
with anti-scalant F260 compared to when the system was operated continuously (Figure 6-1).
This is expected to be a result of two competing mechanisms of fouling from scaling and
biofouling. For the intermittent operation, the dominating fouling mechanism was expected to
be from biofouling, as intermittent operation had the highest membrane ATP measurements
(Figure 6-6) and the SEM autopsy showed the thickest film deposits (Figure 6-5b,e). In contrast,
for continuous operation the dominating mechanism is expected to be from scaling, as the SEM
85
membrane autopsy showed the most scale coverage (Figure 6-5a, d). As well, the ATP
measurements (Figure 6-6) were the lowest for continuous operation.
For intermittent operation with anti-scalant and no rinse, there was an initial spike in the
membrane permeability at the start of Day 3 - Day 7. This is likely a result of the back-flow from
forward osmosis which occurs when the system is shut down for extended periods of time. The
permeate water in the tubing can flow back into the cross-flow cell by forward osmosis. Thus,
locally decreasing the concentration of the water at the membrane surface and slightly dissolving
some crystals which may have adhered to the surface and slightly loosening the foulants. At
start-up there is also an initial sparging and scouring of the membrane surface since there are air
bubbles and turbulent flow during the initial start-up of the system. This resulted in lower levels
of mineral scaling for this case as shown in (Figure 6-5b,e). For intermittent operation with anti-
scalant F260 and no rinse, the start-up did not appear to contribute to an improved permeability
on the second day (Figure 6-2). This is likely due to the biological and organic fouling which
was initiated in the first shutdown period, further decreasing the permeability. On the subsequent
days, it is anticipated that this fouling does not substantially increase. As a result, the osmotic
suck-back and start-up scouring effects dominate, increasing the start-up permeability.
6.4.4.2 Effect of Effect of Permeate Rinsing on Membrane Fouling with Anti-scalant Usage
For intermittent operation with anti-scalant F260 and with rinse, the membrane
permeability was significantly higher than without rinse (Figure 6-2). This is likely because the
anti-scalant effectively binds to the foulants and the rinse effectively washes away the majority
of the scales and deposits from each day, as seen in the membrane autopsy (Figure 6-5c, f). The
biological content was also lower with rinse than without (Figure 6-6). This is likely due to the
rinse which washes away some organic content and residual anti-scalant which may otherwise
86
act as a source of nutrients after shutdown. The salt rejection was also the highest for intermittent
operation with rinse (Figure 6-4). Which indicates, the membrane performance was better than
without the rinse. Even with rinsing, over several days, the membrane permeability began to
decline. This is likely due to some securely attached scales and deposits that were not removed
with the rinsing of the membrane.
6.4.5 Pilot-scale Membrane Permeability Decline
The pilot-scale experimental system was operated intermittently with the Nobleton
groundwater augmented with lab-grade chemicals using the F260 anti-scalant and no rinse
(Figure 6-7). The error bounds shown in the pilot-scale results are different from the lab-scale
error. In the lab-scale results, the error bounds are the standard deviation of the calculated
membrane permeability of the three cross-flow cells. While for the pilot-scale, the error was
calculated based on the accuracy and variation in the conductivity, pressure, and temperature
measurements. This difference was because the pilot-scale experiment was operated only once
since the water volumes required were prohibitively large. The membrane flux at the pilot-scale
ranged from 36.9 Lm-2h-1 at the beginning of the experiment to 13.8 Lm-2h-1 at the end of the
experiment which was below the manufacturer’s recommended flux (55 Lm-2h-1). The
membrane permeability was determined using the fully instrumented experimental system. It
was noted that the trend of declining permeability was slightly interrupted on the third day of
operation (circled in a dotted blue region in Figure 6-7a).
87
Figure 6-7: a) Normalized membrane permeability with the pilot-scale system operated with
the F260 anti-scalant and no rinsing in intermittent operation. The circled permeability rose
above the declining trend due to several start-up attempts after the submersible pump failed.
On the third day, the experiment was very difficult to start, and the submersible pump
failed. This caused air to enter the system and a loss of pressurization. As a result of the air and
turbulence next to the membrane, scales may have been scoured from the surface, thereby
increasing the overall permeability. The remaining days of operation were started without issue.
The normalized membrane permeability shows the general trend of a small increase in
membrane permeability initially at the start of each day. As well, the normalized membrane
permeability is comparable with experimental results from the lab-scale experimental system
(Figure 6-1). The comparison graph (Figure 6-8) shows the membrane permeability has a similar
88
trend of decreasing permeability and daily increase in membrane permeability after the extended
shutdown period for both the pilot-scale system and the lab-scale system.
Figure 6-8: Normalized membrane permeability for intermittent operation and the F260 anti-
scalant without rinsing for both the pilot-scale system and the lab-scale system.
The membrane permeability for the pilot-scale system however, shows a lower daily
increase at the start-up of the system. For the pilot-scale system, the normalized permeability
increased on average by (12±6) % from the shutdown time of the previous day. Compared to the
lab-scale system, where the normalized permeability increased on average by (23±12) % from
the shutdown time of the previous day.
89
6.4.6 Pilot-scale Membrane Autopsy
The surface coverage of the pilot-scale and lab-scale systems operated intermittently with
F260 and no rinse were compared using SEM as shown in Figure 6-9a-d. The pilot-scale SEM
surface appears to have several cracks.
Figure 6-9: Pilot-scale SEM autopsy a) compared to lab-scale SEM autopsy b) for intermittent
operation with F260 and no rinse at x 1000 magnification. Pilot-scale SEM autopsy c)
compared to lab-scale SEM autopsy d) for intermittent operation with F260 and no rinse at
x 3000 magnification.
This is expected to be a result of the unraveling process and drying of the spiral-wound
membrane. It is not expected that the cracking occurred during operation, since the salt-rejection
remained high, above 99.5%, throughout the experiment. The features of the membranes for both
the pilot-scale and lab-scale are similar since they both appear to have a layer on top and under
5 µm
c)a)
b)
20 µm 5 µm
d)
20 µm
90
the crystals (Figure 6-9c, d). The scales on the surface of the SEM in both the pilot-scale system
and lab-scale system were confirmed to have the same composition of calcium, sulfur, and
magnesium using EDS. The features of the membranes and the similar scales indicates similar
fouling mechanisms between the lab-scale system and the pilot-scale system.
The normalized concentration of ATP on the membrane surface for the pilot-scale with
intermittent operation and with F260 anti-scalant and no rinse is compared to the lab-scale in
Figure 6-10. The normalized concentration of ATP on the pilot-scale membrane was
significantly lower than the lab-scale membrane. This could be due to the size of the sample that
was used in the pilot-scale membrane autopsy (4.5 in2) compared to the lab-scale membrane
autopsy for ATP (0.5 in2).
Figure 6-10: Pilot-scale membrane ATP compared to the lab-scale for intermittent operation
with F260 and no rinse.
In addition, the membrane sample for the pilot was across the full width of the spiral
wound element, spanning from the inlet to the outlet. This cross-width sample contained regions
which were visibly fouled (dark yellow and orange) upon visual inspection and regions that
appeared non-fouled upon visual inspection. The outlet regions of the spiral wound membrane
0
1
2
3
Pilot Lab-scale
Mem
bra
ne
cATP
no
rmal
ized
to
fe
ed w
ater
cAT
P (
x10
00
)
91
sample were fouled the most, and therefore may have been a more representative region to select
the membrane sample for ATP analysis from the pilot-scale membrane.
6.4.7 Discussion for the Pilot-scale
6.4.7.1 Effect of Experimental System on Membrane Fouling
Despite similar fouling structures observed from the membrane autopsy (Figure 6-10),
the comparison graph (Figure 6-8) shows the membrane permeability for the pilot-scale system
had a smaller daily increase at the start-up of the system. This is likely due to differences in the
membrane structure and the ratio of permeate water flow reversal which would occur through
osmosis to membrane surface area during shutdown by osmotic suck-back. In the TW30-2514
spiral-wound element in the pilot-scale system, the membrane area is 6503.2 cm2 compared to
the cross-flow membrane area in the lab-scale system, which is 138.7 cm2. This corresponds to
approximately 46 times the area. By contrast, in the pilot-scale system the permeate water tubing
is only about three times the volume of the permeate water tubing of the lab-scale system. This
corresponds to a larger permeate water volume to membrane area ratio for the lab-scale system
(8x10-5 m) compared to the pilot-scale system (1x10-5 m), approximately an order of magnitude
greater.
The osmotic suck-back due to forward osmosis after shutdown during the extended
shutdown periods likely has a lower impact on decreasing the localized concentration on the feed
side for the pilot-scale RO membrane compared to the lab-scale membrane. As well, the
construction of the spiral-wound elements is designed such that only the final fractions of the
membrane sleeves are close to the permeate collection tube. Therefore, unlike the lab-scale
system, there is not a significant amount of permeate water that interacts with the membrane and
that can provide a decreased salt concentration during the extended shutdown periods. A
92
schematic presentation of the difference between the lab-scale and pilot-scale systems during the
extended shutdown periods is shown in Figure 6-11.
Figure 6-11: Schematic of the difference between the lab-scale system and the pilot-scale
system during extended shutdown periods.
6.5 Discussion of Potential Fouling Mechanisms
The results indicate the permeability decline under intermittent operation can be
maintained by using anti-scalant pre-treatment and a permeate rinse prior to shut down. The use
of a simple rinse with permeate water can improve and maintain the membrane permeability
above continuous operation, as discussed in Section 6.4.2. This is explained using the following
mechanistic diagram (Figure 6-12).
Intermittent operation has several steps between operating days. First, in step one, the
extended shutdown period allows sufficient time for forward osmotic flow to permeate into the
cross-flow cell changing the localized salt concentration and loosening scales and deposits from
the membrane surface (Figure 6-12a). Second, in step two, the extended shutdown period allows
for osmotic suck-back and a decreased concentration on the feed side of the membrane (Figure
Feed Spacer
Permeate Spacer
Membrane
Feed Channel
Direction of Osmotic Suck-Back
Legend:
Permeate Collection
Lab-scalePermeate Collection Tube
Pilot-scale
93
6-12b). The experimental system minimized bacterial content using UV-disinfection in the
experiment water tank, however, if there was bacterial growth inside the cross-flow reverse
osmosis SEPA cell, the bacteria would experience nutrient deficiency due to the stagnant water
overnight. Third, in step three, after the extended shutdown period, the system is re-started.
During start-up, very turbulent flow and air enters the system as the water is pressurized by the
high-pressure pump. It is anticipated that this provides sparging and scouring of the membrane
surface and lifts away some foulants and scales which become loosened from the surface during
the extended shutdown period (Figure 6-12c).
Figure 6-12: Proposed fouling mechanism process for intermittent operation of reverse osmosis
membranes
6.6 Conclusions
This chapter presented an experimental study on the effects of intermittent operation with
anti-scalant addition and permeate rinsing on membrane permeability decline when the improved
1 2 3Flow reversal by osmosis during extended shutdown periods
Decreased concentration at the membrane surface and flow reversal loosens foulants accumulated on membrane surface
Startup introduces air and turbulent flow at the membrane surface and helps remove loosened foulants accumulated on membrane surface
Permeate water
Membrane
Feed water
Legend:
Foulant
Scale
Permeate water flow direction
Feed water flow direction
Step: a) b) c)Step: Step:
94
experimental system was operated with a local groundwater augmented with lab-grade
chemicals. The comparison of intermittent operation and continuous operation with anti-scalant
usage and membrane rinsing with permeate water was consistent with the experimental MilliQ-
based matrix results from Chapter 5, the membrane permeability decline can be minimized by
rinsing with permeate water prior to the extended shut down period. The experimental results
showed that intermittent operation alone did not have a significant negative impact on the
membrane performance in the short-term (several days of operation in the cross-flow unit). This
trend remained consistent for the groundwater results compared to the earlier results presented
in Chapter 5 for the experimental MilliQ-based matrix. As well, in both Chapter 5 and Chapter 6,
the membrane permeability was maximized when the system was operated intermittently with a
daily permeate water rinse prior to the extended shutdown period. The pilot-scale system results
compared to the improved experimental system showed that the lab-scale system adequately
represented the normalized membrane permeability decline of a full-scale spiral wound reverse
osmosis membrane module. The discussion on fouling mechanisms provided potential
mechanisms that occur during the extended shut down periods.
95
Chapter 7
Design Optimization Framework for Solar Powered Reverse Osmosis Systems Considering Membrane Fouling from Intermittent Operation
This work is under peer-review for publication (full citation: Freire-Gormaly, M., Bilton, A.,
(Under Review) Design of Solar Powered Reverse Osmosis Desalination Systems Considering
Membrane Fouling caused by Intermittent Operation, Desalination, DES_2018_301). It has
been reproduced here. Permission to use this content will be secured from the editor.
7.1 Introduction
In the previous chapters, the experimental results showed that rinsing the membranes
with permeate water prior to the extended shutdown periods improved the membrane
permeability. Previous optimization studies in the literature for solar powered water treatment
systems do not consider how operating conditions, such as intermittent operation, the use of anti-
scalant or permeate rinsing affect membrane permeability and membrane replacement rates
[14,114,116]. As a result, in the real application of these solar powered water treatment systems,
as the membrane permeability declines, water production would decrease and would not be able
to meet demand. These previous optimization studies therefore over-estimate the system
reliability.
Membrane fouling needs to be considered in the design of PVRO systems to improve the
long-term reliability of these systems. If systems are designed without considerations of
membrane fouling, PVRO systems may be cost-optimally designed for ideal conditions (low
membrane resistance) and be under-sized once fouling occurs. When fouling occurs, membrane
permeability decreases as a function of time and a larger system would be required to meet the
demands of the community.
96
The design optimization framework in this chapter incorporates a membrane fouling
model of the membrane permeability decline for intermittently operated systems towards the
development of a design algorithm. The analytical membrane fouling model for intermittent
operation was developed from the experimental characterization outlined in Chapter 5. The
sensitivity of the system design to the daily water demand, system reliability and membrane
fouling model were also investigated. Several case studies were performed to test the application
of the design framework for several geographic locations to configure the solar powered water
treatment system from commercially available components.
7.2 Design Optimization Framework Approach
For a water treatment system to be practical for remote communities, the system must be
cost-effective and require minimal operator intervention. To determine the ideal design and
operating conditions, an optimization structure, shown in Figure 7-1, was developed. The
approach uses a genetic algorithm combined with a simulation model and a cost model
implemented in Matlab 2016a (MathWorks®, Natick, MA, USA).
Figure 7-1: Framework for performing the cost optimization of the PVRO system.
Genetic Algorithm Optimizer
Hourly Simulation
Model
Design Variables
Water Demand Constraint
System Cost
Minimum Water Production per
day
System Cost Model
System Model
97
The simulation model consists of an hourly simulation for a ten-year period to adequately
evaluate the performance of the solar powered reverse osmosis system configuration. A ten-year
simulation period was selected based on a sensitivity analysis to the simulation duration
described in Section 7.3.2. At each time step, the water produced was determined considering
membrane fouling as a function of time. This approach ensures the system size is sufficient to
provide the community’s daily water demands once membrane fouling occurs.
The technologies investigated were all market available components and included spiral-
wound membranes, pressure vessel housings, pumps, a polycarboxylic anti-scalant, and solar
photovoltaic panels. The polycarboxylic anti-scalant was selected since very small volumes are
required and they are environmentally benign [123]. The solar powered reverse osmosis system
architecture investigated in this chapter is shown in Figure 7-2.
Figure 7-2: Solar powered reverse osmosis water treatment system architecture.
The system architecture consists of a set of components for the power sub-system and
the water treatment sub-system. The water treatment components include a high pressure pump,
reverse osmosis pressure vessel, reverse osmosis membranes, pre-treatment, and water storage
Water Storage Tank
Solar Panels
Battery Storage
Control Electronics
Motor
Pump RO Pressure Vessel
Pre-Treatment
Well
RO Membrane
Brine DisposalCartridge Filter
& Anti-scalant Dosing
98
tank. A representative inventory of modular components was compiled with parameters
extracted from manufacturer data. The inventory of components is described in more detail in
Section 7.3.3.
In addition to the physical system configuration, the operating conditions influence the
rate of membrane fouling and the overall water production of the system. The operating
conditions included in the fouling model (Figure 7-3) were the use of an anti-scalant (with or
without anti-scalant) and the use of clean water for rinsing the membranes prior to extended
shutdown periods. Only intermittent operation was considered, since it had been previously
shown that continuous operation is uneconomic for resource-constrained communities [124].
Figure 7-3: PVRO operating costs and pre-treatment.
A multi-objective optimization was performed to investigate the cost-reliability trade-off
of the various operating conditions. The system was designed to minimize the annualized cost
such that the system could meet the community’s daily water demands with a given reliability
constraint. The reliability of the system was quantified using a loss of water probability (LOWP)
metric which is the number of hours that the community’s water demand was not met divided
by the total number of hours in the simulation. The trade-off between the costs of more expensive
Pre-Treatment and Membrane Cleaning
F135
No-F135
Rinse
No-Rinse
RO System Operating Cost
F135$1.60/lb
Rinse Volume Accounted
Cartridge Filters
99
operating conditions and the benefits for improved membrane permeability were evaluated. For
example, the use of anti-scalants versus no anti-scalants and the benefits of improved membrane
permeability were investigated in the simulation-based optimization framework.
7.3 Design Optimization Framework
The optimization framework described above (Figure 7-1) relied on a physical system
simulation and a cost model. The simulation model used an experimentally derived model of the
membrane fouling under the various operating conditions (Figure 7-3). The following sections
describe the optimization framework, the cost models and how the membrane fouling model was
incorporated into the optimization framework.
7.3.1 Optimization Setup
The optimization problem addressed is the minimization of the annualized system cost
of a solar powered reverse osmosis system design subject to a reliability constraint (loss of water
probability which quantifies the system’s ability to meet the community’s water demands) and
constraints on the optimization design variables. The solar powered reverse osmosis system is a
modular system composed of discrete components (pumps, reverse osmosis membranes,
pressure vessels, storage tanks, solar panels) which makes their optimization difficult. As well,
the cost function to determine the annualized system cost over a 25-year lifetime and the
reliability constraint are non-linear. This results in a mixed integer nonlinear program (MINLP).
Several approaches exist to optimize MINLPs, for example, piecewise linear modeling, spatial
branch-and-bound, and search heuristics [125]. The genetic algorithm [126] is an example of a
search heuristic method for MINLPs and it is a global optimization technique. The genetic
algorithm was used to determine the minimum annualized system cost subject to a reliability
constraint, loss of water probability (LOWP, defined in Equation 7-1). The system reliabilities
100
considered were between 1% LOWP - 10% LOWP. The optimization problem was solved using
eight design variables listed in Table 7-1. The number of batteries was not selected as a design
variable, because only minimal energy storage was required for the solar powered system to
operate during the day with extended shutdown periods at night.
Table 7-1: List of design variables.
Design Variable Number Design Variable Name
1 Use of F135 Anti-scalant
2 Rinsing
3 Length of Membrane Life
4 Number of RO Membranes
5 Diameter of RO Membranes
6 Length of RO Membranes
7 Permeate Water Tank Size
8 Number of Solar PV Panels
The genetic algorithm was selected because it required only an initial population of
potential solutions and a fitness evaluation of the population. The genetic algorithm then mutates
the population towards the Pareto front. Genetic crossovers and mutations occur over several
generations until the optimal design variables which satisfy the reliability constraint are
identified. The genetic algorithm was coupled with a penalty function to evaluate whether or not
the reliability constraint was met.
The genetic algorithm requires the selection of several parameters including the
population size, elite count, cross-over fraction, function tolerance, constraint tolerance, the
number of stalled generations for convergence, and the maximum generations. To determine
these parameters a tuning procedure was performed by changing each parameter individually by
10% to evaluate the effect on the optimal system design. Table 7-2 lists the tuned parameters
that were used for the remainder of analysis in this chapter.
101
Table 7-2: List of genetic algorithm parameters.
Parameter Value
Population size 150
Elite count 1
Crossover fraction 0.3
Function tolerance 0.01
Constraint tolerance 1E-10
Number of stalled generations for convergence 50
Maximum number of generations 150
7.3.2 Simulation Model
The simulation model predicts the water production of a full-scale system using the
power system model coupled to a water treatment system model. The simulation that was
performed for the solar powered reverse osmosis system is outlined in Figure 7-4. The water
treatment system model estimates the membrane fouling as a function of time and the water
produced based on the energy produced by the power system model. Water is stored and
withdrawn from a tank to accommodate for fluctuations in weather and water demand. The
simulation was run hourly for ten years with solar insolation data (global horizontal irradiance
which includes cloud cover) from the geographic location to characterize the system reliability
using the LOWP.
During each hour of the simulation (Figure 7-4), an energy balance was performed on
the power system and the water production was calculated. First, the solar power produced over
the given day was calculated. Second, the total number of hours the water treatment system can
run in the given day was calculated. Third, the water production in the given hour was calculated
considering membrane fouling. Fourth, the water level in the water storage tank at the given hour
was calculated considering the water production and the water demand in that hour.
102
Figure 7-4: System simulation flowchart outlining the hourly and daily steps in the simulation
to determine the loss of water probability.
103
There was a check to determine if the maximum tank level was exceeded. If the
maximum tank level was reached, a flag was reset to the maximum tank level, so that the tank
would not be overfilled. There was also a check to ensure the tank level did not drop below
empty. If the tank was empty, the water demand of the community in that hour was not met, and
the counter for the LOWP was incremented. If it was the time to rinse the system (one operating
condition required a daily rinse prior to the extended shutdown period) the rinse volume was
removed from the tank. Finally, the time was incremented to the subsequent hour. To end the
simulation, the day must meet the total number of days for the simulation (3650 days). The last
step in the simulation was to calculate the actual LOWP for the system design. This value was
used by the genetic algorithm’s penalty function to determine the lowest cost system that met
the desired loss of water probability constraint.
To limit the computational effort, the sensitivity of simulation duration was investigated
(5 to 25 years) for the system size of 5 m3/day and LOWP of 1%. Simulation durations longer
than ten years did not result in a new system configuration and the computational time was 2.5
times longer. Therefore, a simulation length of ten years was selected to ensure the membrane
life was tested beyond two maximum membrane replacements and sufficient inter-annual
variations were considered. The following sections provide the details of the individual models
and the associated assumptions.
In the simulation, the loss of water probability is used as the metric for the system
reliability. The loss of water probability was defined as follows:
𝐿𝑂𝑊𝑃 = 𝑁𝑤𝑎𝑡𝑒𝑟
𝑁𝑠𝑖𝑚 (7-1)
where Nwater is the total number of hours during the simulation that the water demand was not
met in hours, and Nsim is the total number of hours for the simulation. A custom MATLAB
104
program was used for performing the simulations. The details of the individual technology
models are presented in the following individual technology modeling sections.
7.3.2.1 Power System Model
The power for the water treatment system is provided by a solar array with minimal
battery storage. The batteries are sized to provide sufficient power to allow the system to operate
at a constant flow rate and pressure once the system is turned on. Although some previously
installed solar powered reverse osmosis systems operate with variable flow and variable pressure
[50,127], several existing plants operate at a constant flow rate and a set pressure point
[33,34,128]. The solar powered reverse osmosis system configuration investigated in this chapter
is simulated to operate at a constant feed flow rate and a set pressure point.
The solar photovoltaic system model uses solar radiation data from the geographic
location of interest to determine the power produced. For all scenarios, the hourly solar radiation
data was calculated from the global horizontal insolation data from the National Renewable
Energy Laboratory’s (NREL) National Solar Radiation Data Base (NSRDB) [24,129]. The
global horizontal insolation data includes the cloud attenuation [24,129]. Using the hourly solar
insolation data, the power produced by the solar photovoltaic system at a given hourly time step
is given by:
𝑃𝑠𝑜𝑙𝑎𝑟 = 𝑁𝑝𝑎𝑛𝑒𝑙𝜂𝑝𝑎𝑛𝑒𝑙𝜂𝑚𝑝𝑝𝑡𝐼𝑠𝑜𝑙𝑎𝑟𝐴 (7-2)
where Psolar is the power generated by the solar panels in kW, Npanel is the number of solar panels,
which is a design variable, ηpanel is the efficiency of the panel, ηmppt is the efficiency of the
maximum power tracking power electronics (98%), Isolar is the incident radiation from the sun
in kW/m2, and A is the area of the panel in m2. The panel efficiency, ηpanel, and the panel area in
m2, A, are a function of the panel that was modelled. A single solar panel type was used in the
105
design for ease of construction, assembly and maintenance. As a conservative estimate, the panel
was assumed to be mounted horizontally.
The amount of power required by the pump is determined by:
𝑃𝑝𝑢𝑚𝑝 = 𝐹𝑐𝑜𝑛𝑣𝑒𝑟𝑡𝑃𝑓×𝑄𝑓,𝑚𝑒𝑚𝑏
𝜂𝑝𝑢𝑚𝑝 (7-3)
where Ppump is the power required by the high pressure pump in kW, Fconvert is the unit conversion
factor, Pf is the reverse osmosis membrane feed pressure in bar, Qf,memb is the feed flow rate
based on the membrane configuration selected in the design in m3/hour, ηpump is the efficiency
of the high pressure pump.
The solar powered water treatment system is designed to operate at a single operating
point (pressure and flow rate) for several hours a day. Therefore, a small amount of battery
capacity was included in the power system model. It is assumed that the battery is charged in the
morning and then the system will turn on and run at a constant rate. The amount of battery storage
required (Estorage,reqd) was determined based on the ten-year simulation for each individual design
to determine the maximum required energy storage to account for the periods of time when there
was insufficient solar power available to operate the system at the set operating point. To provide
a conservative estimate of the battery storage requirements, no excess energy from the previous
day was used in the subsequent day. A sample day of operation of the solar powered system is
shown in Figure 7-5 with the hourly solar panel power production, battery state of charge and
system power requirements.
To take into account round trip losses from storing the energy in the battery [130], the
total power of the solar system with battery storage is given by:
𝑃𝑠𝑦𝑠𝑡𝑒𝑚 = 𝑃𝑠𝑜𝑙𝑎𝑟𝜂𝑟𝑡𝑏𝑎𝑡𝑡 (7-4)
106
where Psystem is the power produced at a given time step in kW, Psolar is the solar power and ηrtbatt
is the round-trip battery efficiency [130].
Figure 7-5: Power management strategy for the solar powered water treatment model.
simulation for a day of sample operation in Mexico for a 1 m3 system with 4 solar panels and
2.6 kWh of energy storage.
7.3.2.2 Water Treatment System Model
The water treatment system model determines the water production for a given energy input.
The water production (Qp) in m3/hr for each hour was determined based on the membrane
selected in the design vector, the membrane fouling and the operating pressure. The water
production is given by:
𝑄𝑝 = 𝐾𝐹𝐹𝐾𝑊𝐾𝑇𝐴𝑚𝑒𝑚(𝑃 − 𝜋) (7-5)
107
where KFF is the fouling factor, KW is the membrane permeability to water (m hr-1bar-1), KT is the
temperature correction factor, Amem is the membrane area (m2), P is the average operating
pressure in bar, and π is the average osmotic pressure of the water being treated in bar.
Experiments were performed to determine the membrane permeability as a function of
time for various operating conditions (anti-scalant use and rinsing) [131] described in Chapter 5
for an intermittently operated reverse osmosis membrane water treatment system. The system
was operated for 8 hours per day at a fixed operating pressure and was shut down for the
remaining 16 hours of the day. The improved lab-scale experimental system, described
previously [132] and in Chapter 3, was operated using the experimental MilliQ-based matrix to
test a water matrix containing high-levels of dissolved minerals common of brackish water in
remote regions facing high water scarcity. One such community, La Mancalona, is located in the
Yucatan Peninsula (GPS coordinates: Latitude: 18.81°N, Longitude: 89.29°W) was selected
because it is representative of other communities with high solar insolation served by brackish
groundwater. As well, La Mancalona currently has an operating solar powered water treatment
system [33] and is a partner community for the research in this thesis.
7.3.2.3 Analytical Membrane Fouling Model
An analytical membrane fouling model was developed for the different operating
methods for the reverse osmosis water treatment systems. The analytical model was based on
the experimental results [131] and built on the mathematical form for fouling [110]. The
analytical membrane fouling model of the normalized membrane permeability decline (fouling)
per day uses an exponential function, and is defined as follows:
𝐾𝐹𝐹 = 𝑎𝑒𝑏
(𝑑𝑎𝑦+𝑐) (7-6)
108
where day is the number of days since the start of use of the new membranes. The parameters,
a, b, c for the four cases investigated are included in Table 7-3. Additionally, for the daily
behavior of the membrane fouling, a linear fit was determined from the experimental data and
included in Table 7-3 as d. The daily decrease in the membrane permeability (KFF) and the hourly
decline of the membrane permeability each day are shown for two representative days of Case 3
(Figure 7-6). The daily exponential fouling model for Case 1-4 are shown in Figure 7-7.
Table 7-3: Membrane fouling parameters for cases investigated.
Case Description a b c d
1 No AS & No Rinse 0.0048 46 7.7 -0.042
2 AS & No Rinse 0.063 20 6.1 -0.042
3 AS & Rinse 0.54 3.8 6.0 -0.027
4 AS & No Rinse [110] 0.68 79 200 -0.042
Figure 7-6: Membrane fouling model based on experimental data [110,131].
0.70
0.75
0.80
0.85
0.90
0.95
1.00
0 8
No
rmal
ized
Mem
bra
ne
Perm
eab
ilty
Operating Time (t in hours)
Daily Decline
Hourly Decline
𝐾𝐹𝐹 = 𝑎𝑒𝑏
( 𝑎𝑦 𝑐)
𝐾𝐹𝐹, 𝑜𝑢𝑟𝑙𝑦, 𝑎𝑦1 = 𝐾𝐹𝐹,
𝐾𝐹𝐹, 𝑜𝑢𝑟𝑙𝑦, 𝑎𝑦 = 𝐾𝐹𝐹,
1 Operating Time (day in days)2
109
Figure 7-7: Normalized membrane permeability decline (KFF) vs. operating time in days for
Case 1-4 used in the water treatment system model.
The anti-scalant usage of the full-scale system was determined based on the amount of
feed water used per day and the required anti-scalant dose from the water characteristics. The
anti-scalant dose was calculated using the anti-scalant manufacturer’s dose calculation software
Flodose (BWA Water Additives, Version 4.0). This dose rate was used to determine the cost of
using anti-scalant over the system life.
7.3.3 Cost Model
The total cost of the solar powered water treatment system was determined based on the
cost of individual components selected in the design and the operating costs required to run the
system. The power system cost model is described first, followed by the water treatment cost
model. Table 7-4 provides the individual costs of the system components (e.g. pump, filters).
The costs of the water tanks (Figure 7-8) shows that larger tanks tend to cost more, yet there are
certain sizes which break this trend due to economies of scale. In this analysis it was assumed
that only one tank could be selected for simplicity. The cost models for each individual
subsystem are outlined below.
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0 30 60 90 120 150 180
No
rmal
ized
Mem
bra
ne
Per
mea
bilt
y
Operating Time (days)
Case 1
Case 2
Case 3
Case 4
110
Table 7-4: Individual component costs for the solar powered water treatment system.
Subsystem Description USD Product
Power
PV Panel 340.73 Sunmodule Plus SW280 [133]
PV BOS Structural 12%*Wdc Market Price [134]
PV BOS Electrical 27%*Wdc Market Price [134]
Lead-Acid Batteries 80/kWh Market Price [135]
Water
Treatment Membranes
2.5” x 14” 151.00 TW30-2514 [136]
2.5” x 21” 160.00 TW30-2521 [136]
2.5” x 40” 208.00 BW30-2540 [136]
4” x 14” 245.00 TW30-4014 [136]
4” x 21” 255.00 TW30-4021 [136]
4” x 40” 285.00 BW30-4040 [136]
Pressure Vessels
2.5x14 FRP 1000 psi 320.41 ROPV - MHFG-2514-14-1000 [137]
2.5x21 FRP 300 psi 234.69 ROPV - MHFG-2521-14-300 [137]
2.5x40 FRP 300 psi 246.40 ROPV - MHFG-2540-14-300 [137]
4x14 SS 300 psi 104.00 AMI PV4014SSAU-316 [136]
4x21 SS 300 psi 155.00 AMI PV4021SSAU-316 [136]
4x 40, 1 FRP 300 psi 180.00 Codeline 40E30N-1 [138]
4x40, 2 FRP 300 psi 257.78 Codeline 40E30N-2 [138]
4x 40, 3 FRP 300 psi 273.33 Codeline 40E30N-3W [138]
Pumps
Less than 1 Lpm 700.00 Danfoss APP 0.8 [11]
Less than 2 Lpm 4239.00 Danfoss APP 1.8 [11]
Less than 4.8 Lpm 4782.00 Danfoss APP 2.5 [11]
Motors
Less than 1 Lpm 845.00 Leeson 116698.00 [11]
Less than 2 Lpm 1319.00 Leeson G141121.00 [11]
Less than 4.8 Lpm 2141.00 Leeson 170615.60 [11]
Filter
Filter Housing 65.54 Pentek 20" Housing [139]
Filter Cartridge 20.00 Pentek 20" Filter [136]
Anti-scalant
Peristaltic Pump 42.51 Williamson [140]
Container 14.99 Aquapak [141]
Flocon 135 1.6/lb BWA Additives [142]
To compare the various system configurations on an equal basis, the annualized cost
method was used for a 25-year system life. The annualized system cost (Asystem) was determined
111
for each system configuration based on the cost of the power system (Cpower) and the cost of the
water system (CRO) which were annualized using the annualization factor (Fequiv, annual). The
Fequiv,annual is given by:
𝐹𝑒𝑞𝑢𝑖𝑣,𝑎𝑛𝑛𝑢𝑎𝑙 =𝑖(1 𝑖)𝑙𝑖𝑓𝑒
(1 𝑖)𝑙𝑖𝑓𝑒−1 (7-7)
where i is the discount rate of 12% [143], and life is the system life of 25 years.
Figure 7-8: Tank capacity and costs from a supplier of plastic drinking water tanks [144].
The annualized system cost is given by:
𝐴𝑠𝑦𝑠𝑡𝑒𝑚 = 𝐴𝑝𝑜𝑤𝑒𝑟 𝐴𝑅𝑂 (7-8)
where Apower is the power system costs (capital and replacement). The power system cost was
converted to annual costs using:
𝐴𝑝𝑜𝑤𝑒𝑟 = 𝐶𝑝𝑜𝑤𝑒𝑟(𝐹𝑒𝑞𝑢𝑖𝑣,𝑎𝑛𝑛𝑢𝑎𝑙) (7-9)
where Cpower is the cost of the power system and Fequiv,annual is the equivalent annual cost factor.
0
2000
4000
6000
8000
10000
12000
0 20 40 60 80
Tan
k C
ost
(U
SD)
Tank Capacity (m3)
112
The capital costs of the reverse osmosis system were converted to annualized costs (ARO)
as follows:
𝐴𝑅𝑂 = 𝐶𝑅𝑂(𝐹𝑒𝑞𝑢𝑖𝑣,𝑎𝑛𝑛𝑢𝑎𝑙) 𝐴𝑅𝑂,𝑅𝑒𝑝𝑙 (7-10)
Where CRO is the capital cost of the water system, Fequiv,annual is given by Equation (7-7) and
ARO,Repl is the annual replacement cost of the components for the reverse osmosis water treatment
system, given in Equation (7-22).
7.3.3.1 Power System Cost Model
The power system cost (Cpower) can be broken into contributions from the solar system
(Csolar), the battery system (Cbatt) and the balance of system costs (CBOS). The power system cost
(Cpower) was calculated as follows:
𝐶𝑝𝑜𝑤𝑒𝑟 = 𝐶𝑠𝑜𝑙𝑎𝑟 𝐶𝐵𝑂𝑆 𝐶𝑏𝑎𝑡𝑡 (7-11)
The cost of the solar photovoltaic system (Csolar) considers the use of a single brand of solar
panels at a set power rating per panel and it was calculated as follows:
𝐶𝑠𝑜𝑙𝑎𝑟 = 𝑁𝑝𝑎𝑛𝑒𝑙𝑠𝐶𝑃𝑎𝑛𝑒𝑙 (7-11)
where Npanels is the number of panels and CPanel is the cost of the solar panel [145].
The balance of system costs (CBOS) of the solar photovoltaic system were calculated as
follows:
𝐶𝐵𝑂𝑆 = 𝐶𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝐶𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑎𝑙 (7-12)
where Celectrical is the balance of system costs for the electrical components of the solar
photovoltaic system and Cstructural is the balance of system costs for the structural components.
113
The balance of system costs for the electrical components (Celectrical) were determined as
follows:
𝐶𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 = 0. 7(𝑊 𝑐,𝑠𝑜𝑙𝑎𝑟) (7-13)
where the Wdc,solar is the rated DC power rating of the solar photovoltaic system. The factor of
27% is based on U.S. photovoltaic solar costs from NREL [134] for the conductors, switches,
grounding equipment, fuses and breakers required for setting up a solar photovoltaic power
system.
The balance of system costs for the structural components (Cstructural) was determined as:
𝐶𝑠𝑡𝑟𝑢𝑐𝑡𝑢𝑟𝑎𝑙 = 0. (𝑊 𝑐,𝑠𝑜𝑙𝑎𝑟) (7-14)
where the Wdc,solar is the rated DC power rating of the solar photovoltaic system. The factor of
12% is based on U.S. photovoltaic solar costs from NREL [134] for purchasing the racking and
mounting hardware required to set up the solar panels in a structurally sound manner.
The capital cost of the battery storage (Cbatt) required for the system design was calculated
as follows:
𝐶𝑏𝑎𝑡𝑡 = (𝐶𝑐𝑎𝑝,𝑏𝑎𝑡𝑡)(𝐹𝑒𝑞𝑢𝑖𝑣,𝑎𝑛𝑛𝑢𝑎𝑙) 𝐶𝑟𝑒𝑝𝑙𝑎𝑐𝑒,𝑏𝑎𝑡𝑡 (7-15)
where Ccap,batt is the capital cost of the batteries, Fequiv,annual is the equivalent annual cost factor
given by Equation (7-7) and Creplace,batt is the replacement costs for the batteries. The capital
cost of the batteries (Ccap,batt) was given by:
𝐶𝑐𝑎𝑝,𝑏𝑎𝑡𝑡 = 𝐶𝑠𝑡𝑜𝑟𝑎𝑔𝑒(𝐸𝑠𝑡𝑜𝑟𝑎𝑔𝑒,𝑟𝑒𝑞 ) (7-16)
114
where Cstorage is the cost per kWh of lead-acid battery storage ($80 USD/kWh in 2015) [135] and
Estorage,reqd is the amount of battery storage required to ensure steady operation of the reverse
osmosis system during daylight hours as computed based on the simulation model.
The cost to replace the batteries (Creplace,batt) was given by:
𝐶𝑟𝑒𝑝𝑙𝑎𝑐𝑒,𝑏𝑎𝑡𝑡 = 𝐶𝑐𝑎𝑝,𝑏𝑎𝑡𝑡𝑅𝑏𝑎𝑡𝑡𝐹𝑒𝑞𝑢𝑖𝑣,𝑎𝑛𝑛𝑢𝑎𝑙 (7-17)
where Ccap,batt is the capital cost of the batteries, Rbatt is the replacement rate of the batteries,
Fequiv,annual is the equivalent annual cost factor. The only battery operating and maintenance cost
considered was the cost of battery replacement because it is the main operating and maintenance
cost for a lead-acid battery system [38]. The batteries are assumed to require replacing about
every 5 years for a system life of 25 years [38] based on 40,000 hrs between the mean time
between failure [146].
7.3.3.2 Water Treatment System Cost Model
The costs of the water system can be broken down into the capital and operating costs. The
capital costs of the reverse osmosis system were included for the optimization of the solar
powered water treatment system design. The capital costs of the reverse osmosis water treatment
system (CRO) were given by:
𝐶𝑅𝑂 = 𝐶𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 𝐶𝑡𝑎𝑛𝑘 𝐶𝑝𝑖𝑝𝑖𝑛𝑔 (7-18)
where Ccomponents is the cost of the components of the reverse osmosis system, Ctank is the cost of
the tank for water storage, and Cpiping is the cost of the piping required to install the reverse
osmosis system. The Ccomponents was given by:
𝐶𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠 = 𝐶𝑚𝑒𝑚𝑏 𝐶𝑃𝑟𝑒𝑠𝑉𝑒𝑠 𝐶𝑝𝑢𝑚𝑝 𝐶𝑚𝑜𝑡𝑜𝑟 𝐶𝑝𝑟𝑒−𝑡𝑟𝑒𝑎𝑡 (7-19)
115
where Cmemb is the capital cost of the membrane selected in the design, CPresVes is the capital cost
of the pressure vessel, Cpump is the cost of the pump, Cmotor is the cost of the motor, Cpre-treat is
the cost of the pre-treatment system. The cost of the pre-treatment system (Cpre-treat) was given
by:
𝐶𝑝𝑟𝑒−𝑡𝑟𝑒𝑎𝑡 = 𝐶𝑓𝑖𝑙𝑡𝑒𝑟 𝑜𝑢𝑠𝑖𝑛𝑔 𝐶𝑓𝑖𝑙𝑡𝑒𝑟 𝐶𝐴𝑆,𝑝𝑢𝑚𝑝 𝐶𝐴𝑆,𝑡𝑎𝑛𝑘 (7-20)
where Cfilterhousing is the capital cost of the filter cartridge housing, Cfilter is the capital cost of the
filter cartridge, CAS,pump is the capital cost of the anti-scalant dosing peristaltic pump and CAS,tank
is the capital cost of the a small tank for holding the anti-scalant. The cost of the piping (Cpiping)
was given by [11]:
𝐶𝑝𝑖𝑝𝑖𝑛𝑔 = 0. (𝐶𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠) (7-21)
The annual operating costs for the reverse osmosis system were given by:
𝐴𝑅𝑂,𝑅𝑒𝑝𝑙 = 𝐶𝑚𝑒𝑚𝑏(𝑅𝑚𝑒𝑚) 𝐶𝑓𝑖𝑙𝑡𝑒𝑟(𝑅𝑓𝑖𝑙𝑡) 𝐶𝑝𝑢𝑚𝑝(𝑅𝑝𝑢𝑚) 𝐶𝑚𝑜𝑡𝑜𝑟(𝑅𝑚𝑜𝑡) 𝐶𝑝−𝑐 𝑒𝑚 𝐶𝐴𝑆−𝑐 𝑒𝑚 (7-22)
where Rmem is the replacement rate of the reverse osmosis membrane selected in the design
optimization, Rfilt is the annual replacement rate of the cartridge filter pre-treatment, Rpum is the
replacement rate of the pump, Rmot is the replacement rate of the motor, Cp-chem is the cost of the
post-treatment chemicals for re-mineralizing the water to a drinkable standard, Rpc is the
replacement rate of the post-chemicals and CAS-chem is the annual cost of the anti-scalant
chemicals.
The cost of chemicals (Cp-chem) for re-mineralizing the water post-treatment was given
by:
𝐶𝑝−𝑐 𝑒𝑚 = 𝑚𝑟𝑒−𝑚𝑖𝑛 (𝐶𝑟𝑒−𝑚𝑖𝑛) (7-23)
116
where mre-min is the mass of re-mineralizing chemicals used per year in kg as calculated by the
simulation model and Cre-min is the cost of the re-mineralizing chemicals in USD/kg [147].
The cost of anti-scalant chemicals (CAS-chem) was given by:
𝐶𝐴𝑆−𝑐 𝑒𝑚 = 𝑚𝑎𝑛𝑡𝑖−𝑠𝑐𝑎𝑙(𝐶𝐴𝑆) (7-24)
where manti-scal is the mass of anti-scalant used per year in kg as calculated by the simulation
model and CAS is the cost of the anti-scalant in USD/kg. The total annualized cost of the reverse
osmosis water treatment system (ARO) was given by Equation (7-10). The water cost (Cwater) of
the reverse osmosis water treatment system in USD/m3 was given by:
𝐶𝑤𝑎𝑡𝑒𝑟 = 𝐴𝑅𝑂 ÷ 𝑉𝑦𝑒𝑎𝑟 (7-25)
where ARO is the total annualized reverse osmosis system cost from Equation (7-10) and Vyear is
the total volume of drinking water in m3 that was produced per year by the system.
7.4 Optimization Results and Discussion
The design optimization was performed for La Mancalona, Mexico for water production
between one to ten m3/day to show the application of the design optimization for a broad range
of water system sizes. In addition, sensitivity to the fouling models were analyzed for La
Mancalona, Mexico. Finally, the influence of system location on the design was analyzed using
three additional geographic locations.
7.4.1 Effect of System Size on Optimal System Cost
The sensitivity of the system configuration to the system size for three sizes (1 m3/day,
5 m3/day, 10 m3/day) was analyzed for La Mancalona, Mexico. For all system sizes, the variation
in system configuration was determined for a range of LOWP. It was found that the annualized
system cost decreased with increased loss of water probability for all system sizes (Figure 7-9a).
The larger system size (10 m3/day) showed the most reduction in annualized system cost for
117
higher LOWP. This was because for the smaller system size (1 m3/day), the system design was
slightly over-sized. For the 1 m3/day system, there is a limitation in place due to the module
availability and any smaller system would not meet the reliability constraints.
The change in system cost is a piece-wise linear function, due to the modular nature of
the design problem and the same system configuration meets a wide-range of reliability
constraints. The actual LOWP of the optimal configurations were less than the design goal
LOWP (Figure 7-9b). The system configuration is the same for the design goal LOWP 1%-3%
for all three system sizes (1 m3/day, 5 m3/day, 10 m3/day), and it was not until the system design
goal LOWP constraint was 4% that the system configuration changed, resulting in a decreased
annualized system cost.
Figure 7-9: a) Annualized system cost vs. design goal LOWP for variable system size (m3/day)
and b) the actual LOWP compared to the design goal LOWP for a 10 year simulation period
for La Mancalona, Mexico.
The water cost (Cwater in USD/m3) was also determined for La Mancalona, Mexico for
systems ranging from 1 m3/day to 10 m3/day (Figure 7-10). The water cost decreased slightly for
increased loss of water probabilities, however, the cost savings from lower reliability systems
are very small and overlap in Figure 7-10. The water cost vs. system size follows an exponential
1800
2300
2800
3300
3800
4300
0% 2% 4% 6% 8% 10%
An
nu
aliz
ed C
ost
(USD
)
Design Goal LOWP
1m³/day
5m³/day
10m³/day
a)
0%
2%
4%
6%
8%
10%
0% 2% 4% 6% 8% 10%
Act
ual
LO
WP
Design Goal LOWP
1m3/day
5m3/day
10m3/day
1800
2300
2800
3300
3800
4300
0% 2% 4% 6% 8% 10%
An
nu
ali
zed
Co
st (
USD
)
Design Goal LOWP
1m³/day
5m³/day
10m³/day
b)
118
decay, where the smaller system size (1 m3/day) is significantly more expensive than the larger
systems (5 m3/day and 10 m3/day) (Figure 7-10 and Table 7-5). This is due to economies of
scale, since the larger pressure vessels and larger RO modules are cheaper per volume treated
than the smaller RO components (Table 7-4). Therefore, even if a community requires only
1 m3/day during the design process, it would be better for the community to increase their system
size within their budgetary range to ensure they can benefit from these economies of scale. These
larger systems would also facilitate meeting future demand growth over the system lifetime of
25 years.
Figure 7-10: Water cost ($/m3) vs. the system size (m3/day) for La Mancalona, Mexico at
various design goal loss of water probabilities (1%, 5%, 10%).
0.5
1.5
2.5
3.5
4.5
5.5
0 2 4 6 8 10
Wat
er C
ost
(Cwater
in $
/m3 )
System Size (m3/day)
1% LOWP
5% LOWP
10% LOWP
119
Table 7-5: Water cost (USD/m3) for various system sizes (1, 5, 10 m3/day) and LOWP (1%,
5%, 10%). Shows decreasing water costs for reduced system reliability (increasing LOWP).
Design Goal LOWP
System Size
(m3/day) 1% 5% 10%
1 $5.53/m3 $5.42/m3 $5.42/m3
5 $1.56/m3 $1.54/m3 $1.45/m3
10 $1.05/m3 $1.00/m3 $0.93/m3
7.4.2 Effect of System Size on System Configuration
The cost-optimal system configurations for several system sizes (1 m3/day, 5 m3/day, and
10 m3/day) for the community in La Mancalona, Mexico at a 5% loss of water probability
constraint are shown in Figure 7-11. The 5% design goal LOWP was selected as a mid-point in
the analysis range. The larger system design for 10 m3/day required a larger tank size, more solar
panels, and larger energy storage than the smaller system designs (1 m3/day and 5 m3/day). In
all cases, the system configuration uses a tank size that is approximately two to three times the
daily water requirement. Storing the end-product, clean drinking water, is much cheaper than
energy storage, which requires expensive battery replacement costs. The most economical choice
for the membrane modules is the 2.5”x40” module for 1 m3/day and 5 m3/day. The membrane
module is 4”x40” for 10 m3/day. The number of solar panels increased drastically from one solar
panel for 1 m3/day to ten solar panels for 5 m3/day and increased slightly to fourteen solar panels
for 10 m3/day. The system design had increased power requirements to provide larger volumes
of water. For all system sizes, the cost-optimal operating condition was to use anti-scalant and
daily rinsing of the membranes. As well, for all system sizes, the membranes required
replacement at the maximum replacement time (five years).
120
Figure 7-11: System configuration varies with size of the system for a 10 year simulation
period and at 5% LOWP with experimental fouling and anti-scalant usage with daily rinsing.
7.4.3 Effect of System Reliability on Optimal System Configuration
Changing the reliability design goal (LOWP) for a 10 m3/day system for the community
of La Mancalona, Mexico changed the system configuration (Table 7-6). The system was
oversized for a design goal LOWP of 2% and 3% as well as for 5%-8%, 9%, 10%. This
emphasizes the effect discrete design choices have on the overall system configuration. If a
community can withstand a system LOWP of 4% (which translates to approximately 14 days per
year, typically not consecutively, without sufficient water) the annualized system cost can be
decreased by about 150 USD. As well, since the solar panels and the battery storage are the main
components that change to increase the system reliability, a community could purchase the water
components of the system and the solar panels and battery storage for the lower reliability
system, and then upgrade to a higher reliability system by purchasing the solar panels and extra
batteries at a later point in time.
1m3/day 5m3/day 10m3/day
# RO Membranes
Size of RO Membranes
# Solar PV Panels
Battery Storage
Tank Size
2.5”x40” 2.5”x40” 4”x40”
3.0 m3 9.8 m3 18.9 m3
9.5 kWh3.9 kWh1.3 kWh
121
Table 7-6: Effect of changing the design goal LOWP for a 10 m3/day system for La Mancalona,
Mexico.
Replace
Membranes
(Days)
# RO
Memb-
ranes
RO
Dimen-
sions
Tank
Size
(m3)
# Solar
PV
Panels
LOWP Design
LOWP
Goal
Annual
Cost
(USD)
Battery
Storage
(kWh)
1825 6 4" x 40” 18.9
17 0 1%-3% 3827 9.361
14 0.0391 4%-8% 3657 9.499
10 0.0823 9%-10% 3396 8.025
7.4.4 Effect of Membrane Permeability Decline on System Reliability
To evaluate the need to consider membrane fouling in the design of PVRO systems, a
system was optimized for La Mancalona, Mexico without considering fouling and then simulated
for ten years under different fouling cases. A pair-wise comparison of the optimal system
configuration when optimized considering fouling and without considering fouling is shown in
Table 7-7. The number of solar panels selected for the optimal configuration was much smaller
when no fouling was considered. As well, for larger system sizes (5 m3/day and 10 m3/day) the
annualized system cost was significantly higher when fouling was considered.
The annualized system cost for the 5 m3/day system optimized considering fouling was
19% more expensive than the system optimized without considering fouling. The annualized
system cost for the 10 m3/day system optimized considering fouling was 31% more expensive
than the system optimized without considering fouling. By contrast, the 1 m3/day system
optimized considering fouling was only 8% more expensive than the system optimized without
considering fouling.
122
Table 7-7: Effect of considering membrane fouling on the optimal system configuration for a
design goal loss of water probability of 1%.
System Size of
1 m3/day
System Size of
5 m3/day
System Size of
10 m3/day
Optimized
considering:
No
Fouling Fouling
No
Fouling Fouling
No
Fouling Fouling
# RO
Membranes 3 3 5 6 5 6
Size of RO
Membranes 2.5"x40" 2.5"x40" 2.5"x40" 2.5"x40" 2.5"x40" 4"x40"
# Solar PV
Panels 1 3 5 11 12 17
Tank Size
(m3) 3 3 9.8 9.8 18.9 18.9
Battery Size
(kWh) 1.3 2.8 3.4 2.9 3.3 9.4
Annualized
Cost (USD) 1862 2019 2402 2850 2918 3827
To investigate the sensitivity of the LOWP to the membrane fouling model, a set of
fouling cases were tested. The optimal system configuration determined without considering
fouling was then simulated with the experimental fouling model of Case 3 from Table 7-3
(labeled as ‘Exp. Fouling’ in Figure 7-12) and the fouling model from literature developed by
Abbas et. al. [110] which had similar water characteristics, Case 4 from Table 7-3 (labeled as
‘Abbas et. al. [20]’ in Figure 7-12). The difference in the system’s actual loss of water probability
can be seen in Figure 7-12 for 1 m3/day, 5 m3/day, and 10 m3/day systems.
The results show that for larger systems (5 m3/day and 10 m3/day) the effect of not
considering fouling is more important. A solar powered reverse osmosis system that is designed
considering no membrane fouling will severely underperform when membrane fouling occurs.
123
For the 5 m3/day system, the actual LOWP increases to 33% for the Abbas fouling model at a
design goal of 6%-8%, the actual LOWP increases to 20% at a design goal of 9%-10%. This was
because the system configuration at 6%-8% LOWP had more solar panels than RO membranes
compared to the system configuration for 8%-10% LOWP and as a result, the membrane fouling
had a smaller influence on performance.
Figure 7-12: Actual loss of water probability (LOWP) when simulated vs. the Designed goal
LOWP for several system sizes ( a) 1 m3/day, b) 5 m3/day, c) 10 m3/day). The optimal system
design was determined considering no fouling then simulated for a 10 year period with
experimental fouling (with anti-scalant and rinsing) and Abbas et. al. [110] fouling.
For a lower system size (1 m3/day), the difference in system reliability caused by fouling
is smaller. This is due to the oversizing of the system at the lower solar powered reverse osmosis
system size due to the modular components of the system design. These results demonstrate it is
crucial to consider fouling when performing the cost-optimized design of these systems.
Otherwise, the system would be unable to meet the demands of a community once membrane
fouling occurs and the community would be confronted with higher membrane replacement
operating costs than originally anticipated.
0%
10%
20%
30%
40%
0% 5% 10%
Act
ual
LO
WP
Design Goal LOWP
1m3/day
No Fouling
Abbas(AS,NoRinse)
Exp(AS,Rinse)
0%
10%
20%
30%
40%
0% 5% 10%
Design Goal LOWP
5m3/day
No Fouling
Abbas(AS,NoRinse)
Exp(AS,Rinse)
b)a)
0%
10%
20%
30%
40%
0% 5% 10%
Design Goal LOWP
10m3/day
No Fouling
Abbas et. al. [20]
Exp. Fouling
c)
124
7.4.5 Effect of Geographic Location on Optimal System Configuration
To evaluate the impact of variable solar radiation and required net driving pressure,
optimal system designs were configured for several distinct geographic locations (Figure 7-13).
The fouling models described in Table 7-3 were used in these analyses as all communities were
reported to have similar scaling compounds. Nelhal Village in the state of Karnataka, India was
selected because the village has a high total dissolved solids (TDS) in the range of 3000 μS/cm
[148]. As well, in the sub-district where Nelhal is located, 25% of the rural population only has
access to untreated water [149]. Tarkwa Bremang Village in the Prestea-Huni Valley District,
Ghana was selected because the Birimian formation in the region has a high TDS of greater than
2000 μS/cm [150–152]. As well, in the Prestea-Huni Valley District over 30% of rural
households use drinking water from groundwater [153]. The experimental fouling model, Case
3, was used in these geographic case studies.
Figure 7-13: Several communities were selected for evaluating the optimization algorithm in
four countries (Mexico, Ghana, India and Bangladesh).
In Bangladesh, coastal regions are experiencing increased groundwater salinity during
the dry season and this problem is only expected to increase due to climate change and sea-level
125
rise [154]. About 20 million people in the coastal communities of Bangladesh are affected by
saline water sources and these communities also rely heavily on rivers, groundwater (tube wells)
and rain-fed ponds [155]. This over-reliance on saline intruded water sources is degrading
maternal health. There is a 12% increase in hypertension during the dry-season for pregnant
mothers in these coastal communities [154]. The Tala community in the Khulna district of
Bangladesh’s southwest coastal region was selected because during the dry season, on average,
the groundwater reaches a high TDS of about 2600 μS/cm [154].
The variation in system cost versus LOWP for 1 m3/day, 5 m3/day, and 10 m3/day
systems are shown in Figure 7-14. The system configuration did not change significantly for the
three geographic locations for the 1 m3/day system size (Table 7-8). For the 1 m3/day system
size, between the three locations, only the battery storage required a slightly different size. The
battery storage for all three system configurations for 1 m3/day were within ±11% of the median
battery storage (2.7 kWh).
Figure 7-14: Annualized system cost (USD) vs. variable LOWP for three communities in
India, Ghana, and Bangladesh) for variable system sizes (m3/day) a) 1 m3/day, b) 5 m3/day, c)
10 m3/day and a 10 year simulation period.
It is interesting to note that in Figure 7-14 the annualized cost for India was the most
expensive for the 1 m3/day system, but the least expensive for the 5 m3/day and 10 m3/day. This
was because there was a trade-off between the energy availability versus the energy requirements
a) b) c)
1910
1920
1930
1940
1950
1960
1970
0% 5% 10%
An
nu
aliz
ed C
ost
(USD
)
Design Goal LOWP
2500
2550
2600
2650
2700
2750
0% 5% 10%
Design Goal LOWP
3100
3200
3300
3400
3500
3600
3700
0% 5% 10%
Design Goal LOWP
India
Ghana
Bangladesh
126
for the higher driving pressure. India had the highest solar insolation and also the highest TDS
of the three locations. For the 1 m3/day, the number of solar panels could not be reduced while
meeting the water demand, as it could be only done in discrete steps. However, for the larger
systems, the total number of solar panels was higher and there was more flexibility to reduce the
number of solar panels enabling reduction of the solar array size and the corresponding cost
savings.
Table 7-8: Effect of geographic location on the optimal system configuration for a design goal
loss of water probability (LOWP) of 1%. For all the shown design configurations the optimal
operating condition was Case 3 (with anti-scalant and with rinsing) and the optimization found
the optimal system cost at the maximum membrane life of five years.
1 m3/day 5 m3/day 10 m3/day
Location India Ghana Bangladesh India Ghana Bangladesh India Ghana Bangladesh
# RO
Membranes 3 6 6 5 6
Size of RO
Membranes 2.5" x 40" 2.5" x 40" 4” x 40”
# Solar PV
Panels 2 7 8 9 12 15 14
Tank Size
(m3) 3 9.8 18.9
Battery Size
(kWh) 2.9 2.7 2.4 2.5 2.9 2.6 9.6 8.9 8.7
These results show that for widespread adoption of this technology, for the smaller
system size, it would be possible to mass produce a system configuration for a wide range of
locations. For the 5 m3/day and 10 m3/day system sizes, the number of solar panels and the
battery storage varied for the three locations. As a result, for the larger system sizes, the water
system components could be ordered in bulk for widespread implementation of the technology
in various locations. This geographic comparative analysis provides useful comparisons of the
system configurations for consideration prior to widespread adoption of the technology.
127
7.5 Conclusions
In summary, the cost-optimal system configurations for solar powered reverse osmosis
water treatment systems for several geographic locations (Mexico, Ghana, India and
Bangladesh) were compared. It was shown that for lower reliability systems, the annualized
system cost was lower. As well, it was shown that despite variations in solar insolation and water
quality between these geographic locations, a similar system configuration meets the water
demands. Indicating that for widespread adoption, large-scale manufacturing of the systems may
be possible.
The optimization results show that considering membrane fouling during the design is
essential. Not considering membrane fouling during the system design would result in an under-
sized system that would not be able to produce the daily water requirements for the community.
Intermittent operation of solar powered water treatment systems is a useful method to decrease
the energy storage requirements and reduce the system costs to an achievable range
(one USD/m3). Decreased reliability systems can reduce the annualized system costs of solar
powered water treatment systems, however, the savings gained are not significant enough to
justify low quality system performance. The optimal system configuration used tank sizes
between two to three times the daily water requirements of the system to ensure water availability
even during low solar insolation periods. The time to replace the membranes did not seem to
have a significant effect. All optimization study results showed the membrane should be replaced
at the maximum allowable 5-year life. This is likely because the fouling model reached a steady
state membrane permeability after an initial rapid decrease. In practice, community members
should be trained to observe system performance parameters (e.g. pressure fluctuations, salt
rejection) and replace the RO membranes as required.
128
Chapter 8
Summary and Conclusions
8.1 Summary
This thesis presented the experimental characterization of reverse osmosis (RO)
membrane fouling caused by intermittent operation commonly experienced by solar powered
water treatment systems. These solar powered reverse osmosis (PVRO) water treatment systems
are frequently designed with minimal battery storage to decrease the system costs. Instead of
continuous operation, these systems are often operated at a constant pressure only when there is
sufficient solar insolation. This thesis involved the design, building, commissioning and
operation of an experimental system fully equipped for autonomous control and continuous data
collection to characterize the membrane permeability throughout various operating conditions.
Simple operational conditions appropriate for remote areas, including rinsing of the membranes
prior to shut down and pre-treatment with anti-scalant were investigated and shown to help
maintain membrane permeability. The membrane permeability decline seen in the intermittently
operated experiments was modeled mathematically for inclusion in a design optimization
framework.
A novel design optimization framework for solar powered reverse osmosis systems was
developed that considers the membrane permeability decline and selects the system components
(reverse osmosis modules, pumps, solar panels, tank capacity). It was shown that for small
system sizes 1 m3/day to 5 m3/day, systems can be configured to minimize component variation
for mass production. Several case studies were also presented, demonstrating that location to
location solar availability had minimal effect on the cost-optimal system configuration.
129
The details of the individual investigations presented in this thesis are summarized below.
In addition, this thesis resulted in several contributions to the field, detailed in Section 8.2.
Furthermore, recommended avenues for future research are presented in Section 8.3.
8.1.1 Initial Experimental Characterization of Intermittent Operation
The initial experiments on membrane fouling for intermittent operation were performed
on an experimental setup designed to collect data continuously and to have autonomous control
for rinsing with permeate water at the end of each day. The initial experimental results showed
that intermittent operation alone did not have a significant effect on the decline of the membrane
permeability when the experimental system was operated without anti-scalant. This was the first
report in literature of the effect of intermittent operation for extended shutdown periods.
8.1.2 Improved Experimental Characterization of Intermittent Operation at the Lab-scale
The experimental system was improved prior to further experimentation by increasing
the footprint and using a new flow manifold with stainless steel tubing to allow for equal division
of the flow rates. Other experimental system improvements included improved signal processing
to maintain the recovery ratio and replacement of the flow meters with improved pressure-based
flow measurements. The experimental characterization of membrane permeability studied the
effects of intermittent operation, anti-scalant F135 addition, and permeate rinsing. The
comparison of intermittent operation and continuous operation, as well as anti-scalant usage and
membrane rinsing with permeate water showed that a simple permeate rinse prior to shut down
can minimize membrane permeability decline.
The experimental results showed that intermittent operation alone did not have a
significant negative impact on the membrane performance in the short-term (several days of
130
operation in the cross-flow unit). Anti-scalants were observed to improve the membrane
performance when used in isolation for intermittent operation. Rinsing the membranes with 8 L
of permeate water prior to shut down when anti-scalant was used had a significant improvement
on the membrane performance during intermittent operation. On the sixth day of operation, the
average normalized permeability declined only slightly to (87±9) % for intermittent operation
with anti-scalant and with rinse; while all the other operating conditions declined to nearly zero
except continuous operation with anti-scalant (30±4) %.
8.1.3 Experimental Characterization of Intermittent Operation using Groundwater
Chapter 6 presented an experimental study on the effects of intermittent operation, anti-
scalant addition and permeate rinsing had on membrane permeability when the improved
experimental system was operated with the experimental groundwater-based matrix. The
comparison of intermittent operation and continuous operation, as well as anti-scalant usage and
membrane rinsing with permeate water showed the results were consistent with the experimental
MilliQ-based matrix results from Chapter 5, the membrane permeability decline was minimized
by rinsing with permeate water prior to the extended shut down period. The membrane
performance was not significantly impacted by intermittent operation over several days of
operation in the cross-flow unit.
The pilot-scale system results compared to the improved experimental system showed
that the lab-scale system adequately represented the normalized membrane permeability decline
of a full-scale spiral wound reverse osmosis membrane module. The fouling mechanisms
discussed provide a set of potential mechanisms that occur during the extended shutdown periods
towards an improved understanding of membrane fouling under intermittent operation.
131
8.1.4 Design Optimization of Solar Powered Water Treatment Systems Considering Membrane Fouling
A new design optimization framework was developed that considers the unique features
of the community (geographic location, water demand and water salinity) to design the cost-
optimal system considering membrane fouling. The cost-optimal design is configured from an
inventory of off-the-shelf components that are commercially available for the power and water
treatment system components. The cost-optimal system configurations for solar powered reverse
osmosis treatment systems for several geographic locations (Mexico, Ghana, India and
Bangladesh) were compared. It was shown that for lower reliability systems, the annualized
system cost was lower than higher reliability systems. It was demonstrated that considering
membrane fouling was crucial for designing reliable systems. As well, it was shown that despite
variations in solar insolation and water quality between these geographic locations, a similar
system configuration could meet the water demands. Indicating that for widespread adoption,
large-scale manufacturing of the systems would be possible.
8.2 Conclusions
This thesis provides several significant conclusions and contributions towards resolving
the challenge of supplying clean drinking water to off-grid, resource-constrained communities.
This was accomplished through the experimental characterization of membrane permeability
decline in intermittently operated systems and the development of a design optimization tool that
considers this membrane fouling. The research objectives outlined in Section 1.3 were addressed
in this thesis and resulted in the following conclusions:
1. Development of a new experimental system equipped with autonomous control, pre-
treatment, rinsing and continuous data collection to characterize the effect of
intermittent operation on membrane fouling. Prior to this thesis, it was widely
132
acknowledged in the literature, without experimental validation, that intermittent
operation with a daily extended shut down period, common of solar powered reverse
osmosis systems, increased membrane fouling. To address the lack of evidence for
increased membrane fouling caused by intermittent operation, this thesis contributed
the first-in-literature experimental characterization of membrane permeability
decline observed during intermittent operation compared to continuous operation.
The experimental system was designed, built, commissioned and operated as a part
of this thesis. The preliminary experimental results showed that there was minimal
difference in the membrane permeability decline of continuously operated systems
and intermittently operated systems when no anti-scalant was used.
2. Membrane fouling caused by intermittent operation can be minimized by using anti-
scalant pre-treatment and rinsing with permeate water prior to shut down for
experimental lab-scale systems operated with the experimental MilliQ-based matrix.
The experimental characterization of membrane fouling from intermittent operation
was performed on the improved experimental system for longer duration experiments
(up to six days of operation). The improved experimental system was also operated
under more realistic conditions of a real operating desalination system, consistent
operating pressure (20.7 bar) and consistent recovery ratio (75%). This was the first-
report in literature of the detailed experimental characterization of the effects of anti-
scalant and rinsing of the membranes prior to shut down for intermittent operation
compared to continuous operation. These experimental results highlighted the
importance of using anti-scalant pre-treatment for these solar powered reverse
osmosis systems since they are typically operated at high recovery ratios to minimize
the brine which needs to be discharged to the environment. As well, the results
133
demonstrated the efficacy of using a permeate water rinse prior to shut down to
maintain high membrane permeability. An analytical model of the membrane
permeability decline was derived for use in the design optimization framework to
ensure the systems could adequately consider the membrane permeability.
3. Membrane fouling caused by intermittent operation can be minimized by using anti-
scalant pre-treatment and rinsing with permeate water prior to shut down for the
experimental lab-scale and pilot-scale systems operated with the experimental
groundwater-based matrix composed of a local groundwater augmented with lab-
grade chemicals. The results showed that intermittent operation did not have a
significant negative impact on the membrane permeability in the short-term (seven
days of operation in the cross-flow unit) when using the experimental groundwater-
based matrix with organic content. The membrane permeability decline was the least
when the system was operated intermittently with anti-scalant and a daily permeate
water rinse prior to the extended shutdown period. The pilot-scale experimental
system results were consistent with the improved experimental lab-scale system
results. Therefore, the lab-scale system adequately represented the normalized
membrane permeability decline of a full-scale spiral wound reverse osmosis
membrane module.
4. A novel design optimization framework considering membrane fouling was
developed to address the need for a computer-based design tool for configuring solar
powered reverse osmosis systems for resource-constrained communities. The design
optimization considers membrane permeability decline common of intermittently
operated solar reverse osmosis systems and consisted of a simulation model coupled
with a genetic algorithm. It was shown that higher reliability systems have a higher
134
water cost. The case studies for Mexico, Ghana, India and Bangladesh showed that
despite variations in solar insolation and water salinity, a similar system
configuration could meet the community’s water demands. This indicates for
widespread adoption of these systems, large-scale manufacturing would be possible.
The optimization demonstrated that considering membrane fouling during the design
is essential. Without considering membrane fouling in the system design, the system
would be under-sized system and unable to produce the daily drinking water
requirements for the community. It was concluded that intermittent operation of solar
powered water treatment systems effectively decreases the energy storage
requirements and reduces the system cost to an achievable range (one USD/m3) for
resource-constrained communities.
8.3 Recommendations for Future Work
The experimental investigations presented in this thesis can be extended to further
understand the mechanisms of fouling seen by PVRO systems under intermittent operation. This
could involve studying the behavior of membrane fouling for a wider range of water sources to
see the impact of different scaling compounds and organic foulant concentrations. A database of
fouling models for this wide range of water sources could be developed to create a set of models
for use in the design optimization framework. As well, the effect of the recovery ratio could also
be studied. There could be interesting trade-offs between membrane fouling and brine volume.
Furthermore, detailed experimental studies on the various steps during intermittent operation
could be studied to visually quantify the rate of scale growth, biological growth, scale dissolution
and foulant removal. These studies could inform the development of continuous on-line
monitoring for the early identification of membrane fouling and autonomous implementation of
remedial actions during operation.
135
The most immediate extension of this thesis would be to use the design optimization tool
to configure a real system for a new community. This would involve working with project
partners PVPure to find community partners and funding organizations to deploy the custom-
configured water treatment system. Lab-scale analyses could be conducted on the community’s
water using the experimental system and the results could be inputted into the design
optimization framework and subsequently built in the community.
As well, the design optimization framework could be expanded to include multiple
sources of renewable energy for example, wind, tidal or hydropower resources to expand the
application of the water treatment systems to communities with low solar insolation. For
example, Canada’s Northern communities are facing drinking water crises with the end-of-life
of water infrastructure. Adapting the design optimization tool for a broader range of power
sources and a broader range of water treatment technologies could help address the growing need
for small-scale water treatment in resource-constrained communities.
Future work could entail taking this optimization algorithm and adding uncertainty into
the design algorithm for the design of water treatment systems that are robust to increases in
community water demands, annual fluctuations in local groundwater quality, and resilient to
extreme weather events. Further studies are required to identify the dominating fouling
mechanisms for a wide range of water qualities to elucidate methods to effectively minimize
membrane fouling. For example, membrane fouling could be reduced by investigating multiple
forms of pre-treatment, and potentially coupling multiple stages of water treatment technologies
(e.g. Ultrafiltration and RO, Microfiltration and RO).
136
8.4 List of Published Papers from this Thesis
This thesis resulted in one journal article in-preparation and three peer-reviewed journal articles:
Freire-Gormaly, M., Bilton, A., (2017) An Experimental System for Characterization of
Membrane Fouling of Solar Photovoltaic Reverse Osmosis Systems under Intermittent
Operation, Desalination and Water Treatment, Volume 73, pp.54-63.
Freire-Gormaly, M., Bilton, A., (2018) Experimental Quantification of the Effect of
Intermittent Operation on Membrane Performance of Solar Powered Reverse Osmosis
Desalination Systems, Desalination, Volume 435, pp.188-197.
Freire-Gormaly, M., Bilton, A., (In-Preparation) Experimental Lab-scale and Pilot-scale
Characterization of the Effect of Intermittent Operation on Membrane Fouling for Solar
Powered Reverse Osmosis Desalination Systems, Desalination.
Freire-Gormaly, M., Bilton, A., (Under review, 2018) Design of Solar Powered Reverse
Osmosis Desalination Systems Considering Membrane Fouling Caused by Intermittent
Operation, Desalination, DES_2018_301.
8.5 List of Conference Proceedings from this Thesis
The work in this thesis resulted in several conference papers, oral presentations and posters:
Freire-Gormaly, M., Bilton, A., (2017) Optimization of Solar Powered Reverse Osmosis
Water Treatment Systems, CSCE 15th International Conference on Environmental
Engineering, May 31-June 3, Vancouver, BC. (Oral Presentation and Paper)
Freire-Gormaly, M., Bilton, A., (2016) Degradation of Photovoltaic Reverse Osmosis
Systems under Intermittent Operation, EDS Desalination for the Environment: Clean
Water and Energy, May 22-26, Rome, Italy. (Oral Presentation and Paper)
Freire-Gormaly, M., Bilton, A., (2015) Multi-Objective Optimization of Renewable
Power Systems for Remote Communities, ASME 2015 International Design Engineering
Technology Conf., August 2-5, Boston, MA. (Oral Presentation and Paper)
Freire-Gormaly, M., Bilton, A., (2015) Multi-Objective Design of PV-Wind-Battery
Power Systems for Remote Communities, 6th Annual Mechanical and Industrial
Engineering Research Symposium, June 4, University of Toronto, Toronto, ON. (Oral
Presentation)
Freire-Gormaly, M., Bilton, A., (2017) Optimization of Solar Powered Reverse Osmosis
Systems Design Considering Membrane Fouling, 8th Ann. Mech. & Ind. Eng. Research
Symposium, June 9, Toronto, ON. (Poster Presentation)
137
Freire-Gormaly, M., Bilton, A., (2016) Development of Experimental System for
Characterization of Membrane Degradation in Photovoltaic Reverse Osmosis Systems,
7th Ann. Mech. & Ind. Eng. Research Symposium, June 9, Toronto, ON. Awarded Best
Poster Prize. (Poster Presentation)
Freire-Gormaly, M., Bilton, A., (2016) Experimental Characterization of Degradation in
Photovoltaic Reverse Osmosis Systems for Remote Communities, University of
Toronto’s Inst. of Sustainable Energy Symposium, March 22, Toronto, ON. Awarded
Best Poster Prize. (Poster Presentation)
Freire-Gormaly, M., Mahmoud, A., Bilton, A., (2015) Water and Energy Systems for
Remote Communities, University of Toronto CEIE Building Ground Breaking Research,
June 24, Toronto, ON. (Poster Presentation)
Freire-Gormaly, M., Bilton, A., (2014) Optimization of a Renewable Power System for
a Remote Off-Grid Community in Marsabit, Kenya, 5th Ann. MIE Research Symposium,
June 2, Toronto, ON. (Poster Presentation)
138
References
[1] World Health Organization (WHO) and Unicef Joint Monitoring Program (JMP), Progress
on Sanitation and Drinking Water: 2014 Update, Geneva, Switzerland, 2014. doi:ISBN
9789 24 150724 0.
[2] G. Legros, I. Havet, N. Bruce, S. Bonjour, The energy access situation in developing
countries, New York, 2009. http://www.undp.org/content/dam/undp/library/Environment
and Energy/Sustainable Energy/energy-access-situation-in-developing-countries.pdf.
[3] H. Morais, P. Kádár, P. Faria, Z.A. Vale, H.M. Khodr, Optimal scheduling of a renewable
micro-grid in an isolated load area using mixed-integer linear programming, Renew.
Energy. 35 (2010) 151–156. doi:10.1016/j.renene.2009.02.031.
[4] A.A. Setiawan, Y. Zhao, C. V. Nayar, Design, economic analysis and environmental
considerations of mini-grid hybrid power system with reverse osmosis desalination plant
for remote areas, Renew. Energy. 34 (2009) 374–383. doi:10.1016/j.renene.2008.05.014.
[5] B. Peñate, L. García-Rodríguez, Current trends and future prospects in the design of
seawater reverse osmosis desalination technology, Desalination. 284 (2012) 1–8.
doi:10.1016/j.desal.2011.09.010.
[6] K. Khanafer, K. Vafai, Applications of Nanomaterials in Solar Energy and Desalination
Sectors, in: Adv. Heat Transf., 1st ed., Elsevier Inc., 2013: pp. 303–329.
doi:10.1016/B978-0-12-407819-2.00005-0.
[7] L. Henthorne, B. Boysen, State-of-the-art of reverse osmosis desalination pretreatment,
Desalination. 356 (2015) 129–139. doi:10.1016/j.desal.2014.10.039.
139
[8] J.L. Lienhard, M.A. Antar, A. Bilton, A. Blanco, G. Zaragoza, Solar Desalination, Annu.
Rev. Heat Transf. 15 (2012) 277–347.
[9] F. Vince, F. Marechal, E. Aoustin, P. Bréant, Multi-objective optimization of RO
desalination plants, Desalination. 222 (2008) 96–118. doi:10.1016/j.desal.2007.02.064.
[10] M.M. El-Halwagi, Synthesis of reverse-osmosis networks for waste reduction, AIChE J.
38 (1992) 1185–1198. doi:10.1002/aic.690380806.
[11] A. Bilton, A modular design architecture for application to community-scale photovoltaic-
powered reverse osmosis systems, PhD, Massachusetts Institute of Technology, 2013.
http://dspace.mit.edu/bitstream/handle/1721.1/79337/845073888.pdf?sequence=1.
[12] I. de la Nuez Pestana, F.J.G. Latorre, C.A. Espinoza, A.G. Gotor, Optimization of RO
desalination systems powered by renewable energies. Part I: Wind energy, Desalination.
160 (2004) 293–299. doi:10.1016/S0011-9164(04)90031-8.
[13] A. Bilton, L. Kelley, F. Mazzini, Design Optimization of Sustainable Off-Grid Power
Systems for the Developing World, in: ASME 2011 5th Int. Conf. Energy Sustain., 2011:
pp. 1531–1540. doi:10.1115/ES2011-54815.
[14] P. Poovanaesvaran, M.A. Alghoul, K. Sopian, N. Amin, M.I. Fadhel, M. Yahya, Design
aspects of small-scale photovoltaic brackish water reverse osmosis (PV-BWRO) system,
Desalin. Water Treat. 27 (2011) 210–223. doi:10.5004/dwt.2011.1824.
[15] L.F. Greenlee, D.F. Lawler, B.D. Freeman, B. Marrot, P. Moulin, P. Ce, Reverse osmosis
desalination: Water sources, technology, and today’s challenges, Water Res. 43 (2009)
2317–2348. doi:10.1016/j.watres.2009.03.010.
140
[16] B. Peñate, V.J. Subiela, F. Vega, F. Castellano, F.J. Domínguez, V. Millán, Uninterrupted
eight-year operation of the autonomous solar photovoltaic reverse osmosis system in Ksar
Ghilène (Tunisia), Desalin. Water Treat. (2014) 1–8. doi:10.1080/19443994.2014.940643.
[17] B.G. Keefer, R.D. Hembree, F.C. Schrack, Optimized matching of solar photovoltaic
power with reverse osmosis desalination, Desalination. 54 (1985) 89–103.
doi:10.1016/0011-9164(85)80008-4.
[18] L. Kelley, H. Elasaad, S. Dubowsky, Autonomous operation and maintenance of small-
scale PVRO systems for remote communities, Desalin. Water Treat. (2014) 1–13.
doi:10.1080/19443994.2014.940650.
[19] A. Al-Karaghouli, D. Renne, L.L. Kazmerski, Technical and economic assessment of
photovoltaic-driven desalination systems, Renew. Energy. 35 (2010) 323–328.
doi:10.1016/j.renene.2009.05.018.
[20] S. Dallas, N. Sumiyoshi, J. Kirk, K. Mathew, N. Wilmot, Efficiency analysis of the
Solarflow - An innovative solar-powered desalination unit for treating brackish water,
Renew. Energy. 34 (2009) 397–400. doi:10.1016/j.renene.2008.05.016.
[21] A.I. Schafer, C. Remy, B.S. Richards, Performance of a small solar-powered hybrid
membrane system for remote communities under varying feedwater salinities, Water Sci.
Technol. Water Supply. 4 (2004) 233–243.
[22] B. Durham, A. Walton, Membrane pretreatment of reverse osmosis: Long-term experience
on difficult waters, Desalination. 122 (1999) 157–170. doi:10.1016/S0011-
9164(99)00037-5.
141
[23] S.P. Chesters, Innovations in the inhibition and cleaning of reverse osmosis membrane
scaling and fouling, Desalination. 238 (2009) 22–29. doi:10.1016/j.desal.2008.01.031.
[24] NREL, National Solar Radiation Data Base (NSRDB), (2017). http://maps.nrel.gov/nsrdb-
viewer/.
[25] K. Bourouni, T. Ben M’Barek, a. Al Taee, Design and optimization of desalination
reverse osmosis plants driven by renewable energies using genetic algorithms, Renew.
Energy. 36 (2011) 936–950. doi:10.1016/j.renene.2010.08.039.
[26] S. Lee, S. Myung, J. Hong, D. Har, Reverse osmosis desalination process optimized for
maximum permeate production with renewable energy, Desalination. 398 (2016) 133–
143. doi:10.1016/j.desal.2016.07.018.
[27] D.G. Harrison, G.E. Ho, K. Mathew, Desalination using renewable energy in Australia,
Renew. Energy. 8 (1996) 509–513. doi:10.1016/0960-1481(96)88909-9.
[28] A. Serna, F. Tadeo, D. Torrijos, Rule-Based Control of Off-Grid Desalination Powered by
Renewable Energies, J. Renew. Energy Sustain. Dev. (2015) 205–213. http://apc.aast.edu.
[29] N. Ghaffour, J. Bundschuh, H. Mahmoudi, M.F. a. Goosen, Renewable energy-driven
desalination technologies: A comprehensive review on challenges and potential
applications of integrated systems, Desalination. 356 (2015) 94–114.
doi:10.1016/j.desal.2014.10.024.
[30] F. Calise, A. Cipollina, M. Dentice d’Accadia, A. Piacentino, A novel renewable
polygeneration system for a small Mediterranean volcanic island for the combined
production of energy and water: Dynamic simulation and economic assessment, Appl.
142
Energy. 135 (2014) 675–693. doi:10.1016/j.apenergy.2014.03.064.
[31] F.J. García Latorre, S.O. Pérez Báez, A. Gómez Gotor, Energy performance of a reverse
osmosis desalination plant operating with variable pressure and flow, Desalination. 366
(2015) 146–153. doi:10.1016/j.desal.2015.02.039.
[32] H. Sharon, K.S. Reddy, A review of solar energy driven desalination technologies,
Renew. Sustain. Energy Rev. 41 (2015) 1080–1118. doi:10.1016/j.rser.2014.09.002.
[33] H. Elasaad, A. Bilton, L. Kelley, O. Duayhe, S. Dubowsky, Field evaluation of a
community scale solar powered water puri fi cation technology: A case study of a remote
Mexican community application, Desalination. 375 (2015) 71–80.
doi:10.1016/j.desal.2015.08.001.
[34] A. Bilton, S. Dubowsky, Modular Design of Community-Scale Photovoltaic Reverse
Osmosis Systems Under Uncertainty, in: ASME 2014 Int. Des. Eng. Tech. Conf. Comput.
Inf. Eng. Conf., ASME, Buffalo, New York, 2014: p. 10.
http://proceedings.asmedigitalcollection.asme.org/proceeding.aspx?articleid=2090522.
[35] M. Thomson, D. Infield, Laboratory demonstration of a photovoltaic-powered seawater
reverse-osmosis system without batteries, Desalination. 183 (2005) 105–111.
doi:10.1016/j.desal.2005.03.031.
[36] M. Thomson, D. Infield, A photovoltaic-powered seawater reverse-osmosis without
batteries, Desalination. 153 (2002) 1–8.
http://www.sciencedirect.com/science/article/pii/S0011916403800048.
[37] G. Xia, Q. Sun, X. Cao, J. Wang, Y. Yu, L. Wang, Thermodynamic analysis and
143
optimization of a solar-powered transcritical CO2 (carbon dioxide) power cycle for
reverse osmosis desalination based on the recovery of cryogenic energy of LNG (liquefied
natural gas), Energy. 66 (2014) 643–653. doi:10.1016/j.energy.2013.12.029.
[38] E. Koutroulis, D. Kolokotsa, A. Potirakis, K. Kalaitzakis, Methodology for optimal sizing
of stand-alone photovoltaic/wind-generator systems using genetic algorithms, Sol. Energy.
80 (2006) 1072–1088. doi:10.1016/j.solener.2005.11.002.
[39] F. Majali, H. Ettouney, N. Abdel-Jabbar, H. Qiblawey, Design and operating
characteristics of pilot scale reverse osmosis plants, Desalination. 222 (2008) 441–450.
doi:10.1016/j.desal.2007.01.169.
[40] M.D. Afonso, J.O. Jaber, M.S. Mohsen, Brackish groundwater treatment by reverse
osmosis in Jordan, Desalination. 164 (2004) 157–171. doi:10.1016/S0011-
9164(04)00175-4.
[41] H.E.S. Fath, F.M. El-Shall, G. Vogt, U. Seibert, A stand alone complex for the production
of water, food, electrical power and salts for the sustainable development of small
communities in remote areas, Desalination. 183 (2005) 13–22.
doi:10.1016/j.desal.2005.03.028.
[42] D.E. Moudjeber, H. Mahmoudi, M. Djennad, D.C. Sioutopoulos, S.T. Mitrouli, A.J.
Karabelas, Brackish water desalination in the Algerian Sahara—Plant design
considerations for optimal resource exploitation, Desalin. Water Treat. 52 (2014) 4040–
4052. doi:10.1080/19443994.2013.875947.
[43] A.M. Bilton, S. Dubowsky, A Computer Architecture for the Automatic Design of
144
Modular Systems With Application to Photovoltaic Reverse Osmosis, J. Mech. Des. 136
(2014) 101401. doi:10.1115/1.4027879.
[44] A.M. Gilau, M.J. Small, Designing cost-effective seawater reverse osmosis system under
optimal energy options, Renew. Energy. 33 (2008) 617–630.
doi:10.1016/j.renene.2007.03.019.
[45] Y.Y. Lu, Y.D. Hu, D.M. Xu, L.Y. Wu, Optimum design of reverse osmosis seawater
desalination system considering membrane cleaning and replacing, J. Memb. Sci. 282
(2006) 7–13. doi:10.1016/j.memsci.2006.04.019.
[46] K.M. Sassi, I.M. Mujtaba, Effective design of reverse osmosis based desalination process
considering wide range of salinity and seawater temperature, Desalination. 306 (2012) 8–
16. doi:10.1016/j.desal.2012.08.007.
[47] W. He, Y. Wang, M.H. Shaheed, Stand-alone seawater RO (reverse osmosis) desalination
powered by PV (photovoltaic) and PRO (pressure retarded osmosis), Energy. 86 (2015)
423–435. doi:10.1016/j.energy.2015.04.046.
[48] M. Elshafei, A.K. Sheikh, N. Ahmad, Directly Driven RO System by PV Solar Panel
Arrays, Open J. Appl. Sci. 3 (2013) 35–40. doi:10.4236/ojapps.2013.32B007.
[49] C. Tzotzi, T. Pahiadaki, S.G. Yiantsios, a. J. Karabelas, N. Andritsos, A study of CaCO3
scale formation and inhibition in RO and NF membrane processes, J. Memb. Sci. 296
(2007) 171–184. doi:10.1016/j.memsci.2007.03.031.
[50] A. Cipollina, E. Tzen, V. Subiela, M. Papapetrou, J. Koschikowski, R. Schwantes,
Renewable energy desalination: performance analysis and operating data of existing RES
145
desalination plants, Desalin. Water Treat. 55 (2015) 3126–3146.
doi:10.1080/19443994.2014.959734.
[51] C. Fritzmann, J. Löwenberg, T. Wintgens, T. Melin, State-of-the-art of reverse osmosis
desalination, Desalination. 216 (2007) 1–76.
[52] S.R. Suwarno, X. Chen, T.H. Chong, V.L. Puspitasari, D. McDougald, Y. Cohen, S.A.
Rice, A.G. Fane, The impact of flux and spacers on biofilm development on reverse
osmosis membranes, J. Memb. Sci. 405–406 (2012) 219–232.
doi:10.1016/j.memsci.2012.03.012.
[53] E. El-Zanati, S. Eissa, Development of a locally designed and manufactured small-scale
reverse osmosis desalination system, Desalination. 165 (2004) 133–139.
doi:10.1016/j.desal.2004.06.015.
[54] K.M. Sassi, I.M. Mujtaba, Optimal operation of RO system with daily variation of
freshwater demand and seawater temperature, Comput. Chem. Eng. 59 (2013) 101–110.
doi:10.1016/j.compchemeng.2013.03.020.
[55] M.G. Marcovecchio, P.A. Aguirre, N.J. Scenna, Global optimal design of reverse osmosis
networks for seawater desalination: Modeling and algorithm, Desalination. 184 (2005)
259–271. doi:10.1016/j.desal.2005.03.056.
[56] A. Alkhatib, Reverse-Osmosis Desalination of Water Powered by Photo-Voltaic Modules,
Comput. Water, Energy, Environ. Eng. 3 (2014) 22–29. doi:10.4236/cweee.2014.31003.
[57] A.G. Pervov, Scale formation prognosis and cleaning procedure schedules in reverse
osmosis systems operation, Desalination. 83 (1991) 77–118. doi:10.1016/0011-
146
9164(91)85087-B.
[58] E. Lyster, M.M. Kim, J. Au, Y. Cohen, A method for evaluating antiscalant retardation of
crystal nucleation and growth on RO membranes, J. Memb. Sci. 364 (2010) 122–131.
doi:10.1016/j.memsci.2010.08.020.
[59] S. Sobana, R.C. Panda, Modeling and control of reverse osmosis desalination process
using centralized and decentralized techniques, Desalination. 344 (2014) 243–251.
doi:10.1016/j.desal.2014.03.014.
[60] G. Greenberg, D. Hasson, R. Semiat, Limits of RO recovery imposed by calcium
phosphate precipitation, Desalination. 183 (2005) 273–288.
doi:10.1016/j.desal.2005.04.026.
[61] T. Nguyen, F.A. Roddick, L. Fan, Biofouling of water treatment membranes: a review of
the underlying causes, monitoring techniques and control measures., Membranes (Basel).
2 (2012) 804–40. doi:10.3390/membranes2040804.
[62] A.A. Alawadhi, Pretreatment plant design — Key to a successful reverse osmosis
desalination plant, Desalination. 110 (1997) 1–10. doi:10.1016/S0011-9164(97)00079-9.
[63] Z. Hu, A. Antony, G. Leslie, P. Le-Clech, Real-time monitoring of scale formation in
reverse osmosis using electrical impedance spectroscopy, J. Memb. Sci. 453 (2014) 320–
327. doi:10.1016/j.memsci.2013.11.014.
[64] A.I. Radu, J.S. Vrouwenvelder, M.C.M. van Loosdrecht, C. Picioreanu, Modeling the
effect of biofilm formation on reverse osmosis performance: Flux, feed channel pressure
drop and solute passage, J. Memb. Sci. 365 (2010) 1–15.
147
doi:10.1016/j.memsci.2010.07.036.
[65] R.A. Al-Juboori, T. Yusaf, Biofouling in RO system: Mechanisms, monitoring and
controlling, Desalination. 302 (2012) 1–23. doi:10.1016/j.desal.2012.06.016.
[66] K.P. Lee, T.C. Arnot, D. Mattia, A review of reverse osmosis membrane materials for
desalination—Development to date and future potential, J. Memb. Sci. 370 (2011) 1–22.
doi:10.1016/j.memsci.2010.12.036.
[67] M. Herzberg, M. Elimelech, Biofouling of reverse osmosis membranes: Role of biofilm-
enhanced osmotic pressure, J. Memb. Sci. 295 (2007) 11–20.
doi:10.1016/j.memsci.2007.02.024.
[68] P. Bacchin, P. Aimar, R.W. Field, Critical and sustainable fluxes: Theory, experiments
and applications, J. Memb. Sci. 281 (2006) 42–69. doi:10.1016/j.memsci.2006.04.014.
[69] F. a. Abd El Aleem, K. a. Al-Sugair, M.I. Alahmad, Biofouling problems in membrane
processes for water desalination and reuse in Saudi Arabia, Int. Biodeterior.
Biodegradation. 41 (1998) 19–23. doi:10.1016/S0964-8305(98)80004-8.
[70] A. Matin, Z. Khan, S.M.J. Zaidi, M.C. Boyce, Biofouling in reverse osmosis membranes
for seawater desalination: Phenomena and prevention, Desalination. 281 (2011) 1–16.
doi:10.1016/j.desal.2011.06.063.
[71] V. Kochkodan, N. Hilal, A comprehensive review on surface modified polymer
membranes for biofouling mitigation, Desalination. 356 (2015) 187–207.
doi:10.1016/j.desal.2014.09.015.
148
[72] A.I. Radu, J.S. Vrouwenvelder, M.C.M. van Loosdrecht, C. Picioreanu, Effect of flow
velocity, substrate concentration and hydraulic cleaning on biofouling of reverse osmosis
feed channels, Chem. Eng. J. 188 (2012) 30–39. doi:10.1016/j.cej.2012.01.133.
[73] O. Gibert, B. Lefèvre, M. Fernández, X. Bernat, M. Paraira, M. Calderer, X. Martínez-
Lladó, Characterising biofilm development on granular activated carbon used for drinking
water production., Water Res. 47 (2013) 1101–10. doi:10.1016/j.watres.2012.11.026.
[74] J.S. Vrouwenvelder, C. Picioreanu, J.C. Kruithof, M.C.M. van Loosdrecht, Biofouling in
spiral wound membrane systems: Three-dimensional CFD model based evaluation of
experimental data, J. Memb. Sci. 346 (2010) 71–85. doi:10.1016/j.memsci.2009.09.025.
[75] S.H. Joo, B. Tansel, Novel technologies for reverse osmosis concentrate treatment: A
review, J. Environ. Manage. 150 (2015) 322–335. doi:10.1016/j.jenvman.2014.10.027.
[76] T. Chen, A. Neville, M. Yuan, Calcium carbonate scale formation - Assessing the initial
stages of precipitation and deposition, J. Pet. Sci. Eng. 46 (2005) 185–194.
doi:10.1016/j.petrol.2004.12.004.
[77] L.F. Greenlee, F. Testa, D.F. Lawler, B.D. Freeman, P. Moulin, The effect of antiscalant
addition on calcium carbonate precipitation for a simplified synthetic brackish water
reverse osmosis concentrate, Water Res. 44 (2010) 2957–2969.
doi:10.1016/j.watres.2010.02.024.
[78] Z. Amjad, Applications of antiscalants to control calcium sulfate scaling in reverse
osmosis systems, Desalination. 54 (1985) 263–276. doi:10.1016/0011-9164(85)80022-9.
[79] E. Lyster, J. Au, R. Rallo, F. Giralt, Y. Cohen, Coupled 3-D hydrodynamics and mass
149
transfer analysis of mineral scaling-induced flux decline in a laboratory plate-and-frame
reverse osmosis membrane module, J. Memb. Sci. 339 (2009) 39–48.
doi:10.1016/j.memsci.2009.04.024.
[80] S. Salvador Cob, C. Beaupin, B. Hofs, M.M. Nederlof, D.J.H. Harmsen, E.R. Cornelissen,
a. Zwijnenburg, F.E. Genceli Güner, G.J. Witkamp, Silica and silicate precipitation as
limiting factors in high-recovery reverse osmosis operations, J. Memb. Sci. 423–424
(2012) 1–10. doi:10.1016/j.memsci.2012.07.016.
[81] G. Braun, W. Hater, C. Zum Kolk, C. Dupoiron, T. Harrer, T. Götz, Investigations of
silica scaling on reverse osmosis membranes, Desalination. 250 (2010) 982–984.
doi:10.1016/j.desal.2009.09.086.
[82] M. Badruzzaman, A. Subramani, J. DeCarolis, W. Pearce, J.G. Jacangelo, Impacts of
silica on the sustainable productivity of reverse osmosis membranes treating low-salinity
brackish groundwater, Desalination. 279 (2011) 210–218.
doi:10.1016/j.desal.2011.06.013.
[83] S.P. Chesters, M.W. Armstrong, M. Fazel, R. Wilson, D.A. Golding, RO Membrane
Cleaning, Past, Present, Future – Innovations for Improving Ro Plant Operating
Efficiency, Int. Desalin. Assoc. World Congr. Desalin. Water Reuse. (2013).
[84] N. Prihasto, Q.F. Liu, S.H. Kim, Pre-treatment strategies for seawater desalination by
reverse osmosis system, Desalination. 249 (2009) 308–316.
doi:10.1016/j.desal.2008.09.010.
[85] E.E.A. Ghafour, Enhancing RO system performance utilizing antiscalants, Desalination.
150
153 (2003) 149–153. doi:10.1016/S0011-9164(02)01117-7.
[86] C. Gabelich, T. Yun, J. Green, I. Suffet, W. Chen, Evaluation of precipitative fouling for
Colorado River Water desalination using reverse osmosis, Los Angeles, California, 2002.
http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:EVALUATION+OF+P
RECIPITATIVE+FOULING+FOR+COLORADO+RIVER+WATER+DESALINATION
#0.
[87] J.S. Vrouwenvelder, J.W.N.M. Kappelhof, S.G.J. Heijman, J.C. Schippers, D. van der
Kooij, Tools for fouling diagnosis of NF and RO membranes and assessment of the
fouling potential of feed water, Desalination. 157 (2003) 361–365. doi:10.1016/S0011-
9164(03)00417-X.
[88] A.C.M. Franken, Prevention and control of membrane fouling : practical implications and
examining recent innovations, 2009. http://www.ispt.eu/media/CS-01-02-Final-report-
Prevention-and-reduction-of-membrane-fouling.pdf.
[89] C. Liu, S. Caothien, J. Hayes, T. Caothuy, T. Otoyo, T. Ogawa, Membrane Chemical
Cleaning: From Art to Science, Am. Water Work. Assoc. (2001) 25.
[90] Koch Membrane Systems Inc., Anti-scalants reduce membrane cleaning, Membr.
Technol. (2005) 2–3.
[91] I. Zaslavschi, H. Shemer, D. Hasson, R. Semiat, Rotating cylinder technique for assessing
the effectiveness of anti-scalants, Water Res. 47 (2013) 3716–3722.
doi:10.1016/j.watres.2013.04.022.
[92] H. Wang, G. Liu, J. Huang, Y. Zhou, Q. Yao, S. Ma, K. Cao, Y. Liu, W. Wu, W. Sun, Z.
151
Hu, Performance of an environmentally friendly anti-scalant in CaSO4 scale inhibition,
Desalin. Water Treat. 53 (2015) 8–14. doi:10.1080/19443994.2013.837007.
[93] S.P. Chesters, F. Del Vigo, E.G. Darton, Theoretical and practical experience of calcium
phosphate inhibition in RO waters, IDA World Congr. Canar. (2007) 1–14.
[94] E.G. Darton, S. Gallego, Simple laboratory techniques improve the operation of RO pre-
treatment systems., in: IDA World Congr. Canar., 2007: pp. 1–11.
http://www.genesysro.com/Distributors_docs/5.Genesys Research
Papers07/Simple_laboratory_techniques.pdf%5Cnpapers://54d15a03-26bc-4494-9a1b-
dac933b164ea/Paper/p5866.
[95] B.K. Pramanik, Y. Gao, L. Fan, F.A. Roddick, Z. Liu, Antiscaling effect of polyaspartic
acid and its derivative for RO membranes used for saline wastewater and brackish water
desalination, Desalination. 404 (2017) 224–229. doi:10.1016/j.desal.2016.11.019.
[96] D. Lisitsin, Q. Yang, D. Hasson, R. Semiat, Inhibition of CaCO3 scaling on RO
membranes by trace amounts of zinc ions, Desalination. 183 (2005) 289–300.
doi:10.1016/j.desal.2005.10.002.
[97] A. Antony, J.H. Low, S. Gray, A.E. Childress, P. Le-Clech, G. Leslie, Scale formation and
control in high pressure membrane water treatment systems: A review, J. Memb. Sci. 383
(2011) 1–16. doi:10.1016/j.memsci.2011.08.054.
[98] Y.O. Rosenberg, I.J. Reznik, S. Zmora-Nahum, J. Ganor, The effect of pH on the
formation of a gypsum scale in the presence of a phosphonate antiscalant, Desalination.
284 (2012) 207–220. doi:10.1016/j.desal.2011.08.061.
152
[99] B. Peñate, F. Castellano, P. Ramírez, PV-RO desalination stand-alone system in the
village of Ksar Ghilène (Tunisia), in: IDA World Congr., Maspalomas, Gran Canaria -
Spain, 2007: pp. 1–12.
[100] D.J. London, Reverse Osmosis Membrane Fouling Control for an Integrated membrane
water reclamation system, MASc, University of Toronto, 2005.
http://search.library.utoronto.ca/details?5739771&uuid=1baf8de5-6b5d-4689-af8f-
3c5223fac62d.
[101] E.R. Cornelissen, J. Vrouwenvelder, S. Heijman, X. Viallefont, D. Vanderkooij, L.
Wessels, Periodic air/water cleaning for control of biofouling in spiral wound membrane
elements, J. Memb. Sci. 287 (2007) 94–101. doi:10.1016/j.memsci.2006.10.023.
[102] M.A. Saad, Early discovery of RO membrane fouling and real-time monitoring of plant
performance for optimizing cost of water, Desalination. 165 (2004) 183–191.
doi:10.1016/j.desal.2004.06.021.
[103] M.A. Saad, Silent Alarm software optimizes membrane plants, Int. Desalin. Water Reuse
Q. 13 (2003) 45–49.
[104] M. Uchymiak, A. Rahardianto, E. Lyster, J. Glater, Y. Cohen, A novel RO ex situ scale
observation detector (EXSOD) for mineral scale characterization and early detection, J.
Memb. Sci. 291 (2007) 86–95. doi:10.1016/j.memsci.2006.12.038.
[105] R.K. McGovern, G.P. Thiel, G. Prakash Narayan, S.M. Zubair, J.H. Lienhard V,
Performance limits of zero and single extraction humidification-dehumidification
desalination systems, Appl. Energy. 102 (2013) 1081–1090.
153
doi:10.1016/j.apenergy.2012.06.025.
[106] E.M. V Hoek, J. Allred, T. Knoell, B.H. Jeong, Modeling the effects of fouling on full-
scale reverse osmosis processes, J. Memb. Sci. 314 (2008) 33–49.
doi:10.1016/j.memsci.2008.01.025.
[107] A.S. Kahdim, S. Ismail, A. Abdulrazaq, Modeling of reverse osmosis systems, Water. 158
(2003) 323–329.
[108] H.-J. Oh, Y.-K. Choung, S. Lee, J.-S. Choi, T.-M. Hwang, J.H. Kim, Scale formation in
reverse osmosis desalination: model development, Desalination. 238 (2009) 333–346.
doi:10.1016/j.desal.2008.10.005.
[109] J. Thompson, N. Lin, E. Lyster, R. Arbel, T. Knoell, J. Gilron, Y. Cohen, RO membrane
mineral scaling in the presence of a biofilm, J. Memb. Sci. 415–416 (2012) 181–191.
doi:10.1016/j.memsci.2012.04.051.
[110] A. Abbas, N. Al-Bastaki, Performance decline in brackish water Film Tec spiral wound
RO membranes, Desalination. 136 (2001) 281–286. doi:10.1016/S0011-9164(01)00191-6.
[111] Y.G. Lee, Y.S. Lee, J.J. Jeon, S. Lee, D.R. Yang, I.S. Kim, J.H. Kim, Artificial neural
network model for optimizing operation of a seawater reverse osmosis desalination plant,
Desalination. 247 (2009) 180–189. doi:10.1016/j.desal.2008.12.023.
[112] S.S. Madaeni, M. Shiri, A.R. Kurdian, Modeling, Optimization, and Control of Reverse
Osmosis Water Treatment in Kazeroon Power Plant Using Neural Network, Chem. Eng.
Commun. 202 (2014) 6–14. doi:10.1080/00986445.2013.828606.
154
[113] A.M. Aish, H.A. Zaqoot, S.M. Abdeljawad, Artificial neural network approach for
predicting reverse osmosis desalination plants performance in the Gaza Strip,
Desalination. 367 (2015) 240–247. doi:10.1016/j.desal.2015.04.008.
[114] M.A. Alghoul, P. Poovanaesvaran, M.H. Mohammed, A.M. Fadhil, A.F. Muftah, M.M.
Alkilani, K. Sopian, Design and experimental performance of brackish water reverse
osmosis desalination unit powered by 2 kW photovoltaic system, Renew. Energy. 93
(2016) 101–114. doi:10.1016/j.renene.2016.02.015.
[115] A.M. Helal, S.A. Al-Malek, E.S. Al-Katheeri, Economic feasibility of alternative designs
of a PV-RO desalination unit for remote areas in the United Arab Emirates, Desalination.
221 (2008) 1–16. doi:10.1016/j.desal.2007.01.064.
[116] B.A. Qureshi, S.M. Zubair, A.K. Sheikh, A. Bhujle, S. Dubowsky, Design and
performance evaluation of reverse osmosis desalination systems: An emphasis on fouling
modeling, Appl. Therm. Eng. 60 (2013) 208–217.
doi:10.1016/j.applthermaleng.2013.06.058.
[117] A. Antony, T. Chilcott, H. Coster, G. Leslie, In situ structural and functional
characterization of reverse osmosis membranes using electrical impedance spectroscopy,
J. Memb. Sci. 425–426 (2013) 89–97. doi:10.1016/j.memsci.2012.09.028.
[118] K. Ishida, R. Bold, D. Phipps, Identification and evaluation of unique chemicals for
optimum membrane compatibility and improved cleaning eficiency, 2005.
[119] A.H. Taheri, L.N. Sim, T.H. Chong, W.B. Krantz, A.G. Fane, Prediction of reverse
osmosis fouling using the feed fouling monitor and salt tracer response technique, J.
155
Memb. Sci. 475 (2015) 433–444. doi:10.1016/j.memsci.2014.10.043.
[120] L. Seminario-vidal, E.R. Lazarowski, S.F. Okada, Assessment of Extracellular ATP
Concentrations, in: P.B. Rich, C. Douillet (Eds.), Biolumin. Methods Mol. Biol., 2009: pp.
25–36. doi:10.1007/978-1-60327-321-3.
[121] A. Sweity, T. Rezene, I. David, S. Bason, Y. Oren, Z. Ronen, M. Herzberg, T.R. Zere, I.
David, S. Bason, Y. Oren, Z. Ronen, M. Herzberg, Side effects of antiscalants on
biofouling of reverse osmosis membranes in brackish water desalination, J. Memb. Sci.
481 (2015) 172–187. doi:10.1016/j.memsci.2015.02.003.
[122] S.M. Hunt, E.M. Werner, B. Huang, M.A. Hamilton, P.S. Stewart, Hypothesis for the
Role of Nutrient Starvation in Bio lm Detachment, Appl. Environ. Microbiol. 70 (2004)
7418–7425. doi:10.1128/AEM.70.12.7418.
[123] . BWA Water Additives UK, Product Information Flocon 885 - Biodegradable and
Phosophorous Free Antiscalant for Reverse Osmosis, (2014) 2.
http://www.wateradditives.com/images/Updated_WF/Flocon_885_GP_WF.pdf (accessed
April 14, 2015).
[124] A.M. Bilton, L.C. Kelley, S. Dubowsky, Photovoltaic Reverse Osmosis – Feasibility and a
Pathway to Develop Technology, Desalin. Water Treat. 31 (2011) 24–36.
[125] P. Belotti, C. Kirches, S. Leyffer, J. Linderoth, J. Luedtke, A. Mahajan, Mixed-integer
nonlinear optimization, Acta Numer. 22 (2013) 1–131. doi:10.1017/S0962492913000032.
[126] D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning, 1st
ed., Addison-Wesley Longman Publishing Co., Inc., Boston, MA, USA, 1989.
156
https://dl.acm.org/citation.cfm?id=534133.
[127] B.S. Richards, D.P.S. Capão, W.G. Früh, A.I. Schäfer, Renewable energy powered
membrane technology: Impact of solar irradiance fluctuations on performance of a
brackish water reverse osmosis system, Sep. Purif. Technol. 156 (2015) 379–390.
doi:10.1016/j.seppur.2015.10.025.
[128] V.J. Subiela, B. Penate, Autonomous Desalination and Cooperation. The Experience in
Morocco Within the ADIRA Project, in: Water Secur. Mediterr. Reg., 2011: pp. 319–337.
doi:10.1007/978-94-007-1623-0.
[129] A. Habte, M. Sengupta, A. Lopez, Evaluation of the National Solar Radiation Database
(NSRDB): 1998-2015, NREL (National Renew. Energy Lab. (2017) 1998–2015.
http://www.nrel.gov/docs/fy17osti/67722.pdf.
[130] P. Yang, A. Nehorai, Joint Optimization of Hybrid Energy Storage and Generation
Capacity with Renewable Energy, IEEE Trans. Smart Grid. 5 (2014) 1566–1574.
http://arxiv.org/abs/1309.5832 (accessed October 3, 2014).
[131] M. Freire-Gormaly, A.M. Bilton, Experimental quantification of the effect of intermittent
operation on membrane performance of solar powered reverse osmosis desalination
systems, Desalination. (2017).
doi:http://www.sciencedirect.com/science/article/pii/S0011916417311104.
[132] M. Freire-Gormaly, A.M. Bilton, An experimental system for characterization of
membrane fouling of solar photovoltaic reverse osmosis systems under intermittent
operation, Desalin. Water Treat. 73 (2017) 54–63. doi:10.5004/dwt.2017.20391.
157
[133] CS, Civic Solar, (2017). https://www.civicsolar.com/product/solarworld-plus-sw-280-
mono-280w-mono-slvwht-1000v-us-solar-panel (accessed July 13, 2017).
[134] R. Fu, D. Chung, T. Lowder, D. Feldman, K. Ardani, R. Fu, D. Chung, T. Lowder, D.
Feldman, K. Ardani, U.S. Solar Photovoltaic System Cost Benchmark: Q1 2016, Golden,
CO, 2016. http://www.nrel.gov/docs/fy16osti/66532.pdf.
[135] P.T. Moseley, D.A.J. Rand, J. Garche, Lead-acid batteries for future automobiles: Status
and prospects. Status and prospects., Elsevier B.V., 2017. doi:10.1016/B978-0-444-
63700-0.00021-0.
[136] Water Anywhere, (2017). http://www.wateranywhere.com (accessed August 20, 2017).
[137] FWS, Fresh Water Systems, (2017). https://www.freshwatersystems.com/ (accessed
August 23, 2017).
[138] EA, Evoqua Advantage, (2017). https://www.evoquaadvantage.com/ (accessed August 23,
2017).
[139] A, Aquatell, (2017). https://www.aquatell.ca/products/standard-water-filter-housing-kit-
20-blue (accessed August 15, 2017).
[140] WP, Williamson Pumps, (2017). http://www.williamson-shop.co.uk/100-series-with-dc-
powered-motors-3586-p.asp (accessed July 21, 2017).
[141] Canadian Tire, (2017). http://www.canadiantire.ca/en/pdp/reliance-rectangular-aqua-pak-
water-container-0854035p.html (accessed August 31, 2017).
[142] BWA Water Additives UK, BWA Water Additives, (2017).
158
http://www.wateradditives.com/ (accessed August 12, 2017).
[143] R. Meeks, Water works: The economic impact of water infrastructure, 2012.
http://heep.hks.harvard.edu/files/heep/files/dp35_meeks.pdf.
[144] PM, Plastic Mart, (2017). http://www.plastic-mart.com/category/9/plastic-water-tanks
(accessed September 5, 2017).
[145] AltEsource, Solar Panels, (2017). http://www.altestore.com/store/Solar-Panels/c541/.
[146] SunXtenderInc, AGM Deep Cycle Solar Battery Datasheet, West Covina, CA, 2014.
https://s3.amazonaws.com/ecodirect_docs/CONCORDE/Sun_Xtender.pdf.
[147] M. Sarai Atab, A.J. Smallbone, A.P. Roskilly, An operational and economic study of a
reverse osmosis desalination system for potable water and land irrigation, Desalination.
397 (2016) 174–184. doi:10.1016/j.desal.2016.06.020.
[148] B.M. Jha, Ground water quality in shallow aquifers of India, 2010.
http://www.cgwb.gov.in/documents/Waterquality/GW_Quality_in_shallow_aquifers.pdf.
[149] C. Chandramouli, District Census Handbook of Raichur, Raichur, Karnataka, 2011.
http://www.censusindia.gov.in/2011census/dchb/2905_PART_B_DCHB_RAICHUR.pdf.
[150] S.M. Yidana, A. Yidana, An assessment of the origin and variation of groundwater
salinity in southeastern Ghana, Environ. Earth Sci. 61 (2010) 1259–1273.
doi:10.1007/s12665-010-0449-y.
[151] B.Y. Bruce, S.M. Yidana, Y. Anku, T. Akabzaa, D. Asiedu, Water quality characterization
in some birimian aquifers of the Birim Basin, Ghana, KSCE J. Civ. Eng. 13 (2009) 179–
159
187. doi:10.1007/s12205-009-0179-4.
[152] J.P. Milési, P. Ledru, P. Ankrah, V. Johan, E. Marcoux, C. Vinchon, The metallogenic
relationship between Birimian and Tarkwaian gold deposits in Ghana, Miner. Depos. 26
(1991) 228–238. doi:10.1007/BF00209263.
[153] P. Nyarko, 2010 Population and Housing Census, District Analytical Report - Prestea-
Huni Valley District, 2014.
http://www.statsghana.gov.gh/docfiles/2010_District_Report/Western/Prestea Huni-
Valley.pdf.
[154] A.E. Khan, A. Ireson, S. Kovats, S.K. Mojumder, A. Khusru, A. Rahman, P. Vineis,
Drinking water salinity and maternal health in coastal Bangladesh: Implications of climate
change, Environ. Health Perspect. 119 (2011) 1328–1332. doi:10.1289/ehp.1002804.
[155] L. Hossain, M. Kamal Hossain, Impact of Sea-Level Rise on Land Use Suitability and
Adaptation Options: Coastal Land Zoning in the Southwest, Dhaka, Bangladesh, 2006.
http://www.academia.edu/1224958/CLIMATE_CHANGE_SEA_LEVEL_RISE_AND_C
OASTAL_VULNERABILITIES_OF_BANGLADESH_WITH_ADAPTATION_OPTIO
NS.