-
1
Optimal Location of Booster Chlorination Stations in
Water Distribution Networks using Genetic Algorithms.
Hernndez Cervantes, Daniel
1; Mora Rodrguez, Jess
2; Delgado Galvn, Xitlali
2;
Ortz Medel, Josefina2; Jimnez Magaa, Martn Rubn
3
1 Hydraulics Engineering Student. Universidad de Guanajuato. Av.
Jurez No. 77, Centro, 36000,
Guanajuato, Mexico. [email protected] 2 Geomatics
and Hydraulics Engineering Department. Universidad de Guanajuato.
Av. Jurez No. 77,
Centro, 36000, Guanajuato, Mexico. [email protected],
[email protected], [email protected] 3 Hydraulics Department. Facultad
de Estudios Superiores de Aragn, Universidad Nacional Autnoma
de
Mxico. Av. Rancho Seco S/N, Colonia Impulsora, Nezahualcyotl,
Edo. de Mxico, 57130.
[email protected]
Abstract
The water distribution networks (WDN) through time, have
suffered damage due to wear and
normal operation. WDN can present problems of intrusion of
pollutants, particularly, pathogen
microorganisms might affect consumers health, one solution for
those problems is attended by adding more chlorine on the Drinking
Water Treatment Plant (DWTP), but this alternative could
generate the formation of Trihalomenthanes (THMs) when the
chlorine reacts with natural organic
matter. To reduce the risk of the presence of such
microorganisms, the installation of stations
chlorine reinjection takes place in the water distribution
network. The main objective is to define
the most appropriate sites, focusing on proposing the least
chlorine supply of stations providing the
optimum amount of chlorine. The application of this work intends
to minimize the risk of health
problems of persons linked to the consumption of contaminated
water or excess disinfectant in
drinking water having unfavorable operating conditions and
maintenance.
Keywords: Optimal concentration; water quality; chlorination
booster location.
Corresponding author: Mora Rodrguez, Jess.
1. INTRODUCTION
Nowadays, a large number of water distribution networks (WDN)
have reached their lifetime;
through time, they have suffered damage due to wear and normal
operation. WDN can present
problems of intrusion of pollutants, according to the type of
operation and maintenance.
Particularly, pathogen microorganisms might affect consumers
health (Figure 1). The preservation
of water quality in WDN is one of the most complex technological
issues for water suppliers.
Optimal water quality related to microorganism is achieved when
the disinfection process treats the
water in the Drinking Water Treatment Plant (DWTP). Once the
optimal quality is achieved, the
water flows to the WDN. Disinfectants are used primarily to
ensure the inactivation of
microorganisms that may be present in the water from supply
sources and the re-growing during the
network. The objective is to prevent gastrointestinal diseases
due to contaminated drinking water.
The effectiveness of disinfection and microorganism resistance
depend on the concentration of
disinfectant and the contact time.
-
2
Figure 1. Presence of microorganisms in water pipes (Based on
Knobelsdorf, 1997).
The industry of bottle water is the most important in the world
and the people have assumed to
consume the drinking water from this industry. Nevertheless, the
Municipalities Water Distribution
Systems (MWDS) require maintain the disinfection according to
the Official Mexican Standard
(NOM-127-SSA1-1994 or NOM-127) established by the Ministry of
Health. In the NOM-127 the
range of the chlorine must be on the points of consumption of
0.20 mg/L and 1.50 mg/L. with this
consideration is going to propose the analysis of this
paper.
The principal disinfection process includes free chlorine,
chloramines, ozone, chlorine dioxide,
ultraviolet light (Propato et al., 2004). The free chlorine is
one of the most effective agents to
inactive bacteria and other pathogens due to its residual effect
of disinfection along the entire
network (Propato et al., 2004). However, when chlorine gets in
contact with water, it reacts in
different processes and tends to diminish its concentration
(Geldreich, 1996).
The booster chlorination stations (BCS) are install in places
where the free chlorine concentration is
under the minimum level; with the CBS the operators of the MWDS
warrant the disinfection on the
entire network with the minimum concentration. The technique
used to find the optimal scenario of
minimum chlorination on the WDN is the genetic algorithms (GA).
The paper proposes a
decrement of the concentrations on the MWDS in order to uniform
the concentrations of free
chlorine along the entire network considering diminish the use
of chlorine and with this savings in
the income costs of the MWDS. On the other hand, the second
objective to uniform concentrations
is to avoid higher concentrations on the zones near the Drinking
Water Treatment Plant (DWTP)
diminishing the possibility of the generation of
Trihalomenthanes (THMs). The aim is to define the
most appropriate sites, focusing on proposing the least amount
of stations providing the optimal
concentration of free chlorine. The application of this tool
aims to minimize the risk of health
problems of the consumers associated to the consumption (in the
cases that the consumers drink or
cook with this water) of contaminated water on the zones where
the concentrations are under the
inferior limit and because of the consumption of higher
concentrations above to the superior limits
near to the DWTP.
The use of high concentrations inner the limits of disinfectant
in drinking water it could be produce
by having unfavorable operating conditions and maintenance. For
example in the case of Mexico, a
lot of small cities supply water in an intermittent way and to
warranty the disinfection on the
networks due to the entrance of pathogens during the hours
without service, the operator increment
the amount of chlorine on the DWTP.
-
3
In this paper, it is proposed the optimal location of BCS using
the GA. The main objective is to
warranty the minimum permissible limit of free residual chlorine
of 0.20 mg/L, analyzing the
economical cost that implies install a BCS and considering the
less use of disinfectant in the WDN.
2. CHLORINATION IN WDN
Most of the disinfectant use in Mexico in DWTP is the free
chlorine, due to its effectiveness along
the WDN. Therefore the analysis of the optimal BCS is made with
this disinfectant.
2.1 Decay mechanism of chlorine
When chlorine gets in contact with water, it reacts in different
processes and tends to degrade its
concentration. Loss of chlorine concentration is a function of
the characteristics of microorganisms,
their state and their mixture with dissolved material, besides
to other factors such as temperature
and pH (Geldreich, 1996). According to Castro (2003), loss of
residual chlorine concentration
throughout WDN is due to several separate mechanisms, on the
Table 1, it is shows the diverse type
of reaction and some reaction coefficients related to them.
Those values depend on multiple
variables and they could vary according to the local conditions
of every study. Ozdemizer and
Erkan (2005) relate the decay of chlorine to the residence time
of water in the network, the quality
of the treated water and the age of the pipes. Alcocer et al.
(2004) mentioned that the lowest
concentration could occur in zones with low velocity and in
storage tanks, no necessary in the
farthest zones from the DWTP.
Table 1. Mechanisms chlorine decay
Type of reaction
Typical values for
Reaction Coefficients
(CNA, 2007b)
Typical values for
Reaction Coefficients
(UBA, 2000) By chlorine reaction in the bulk
water, bacteria and other
microorganisms. 0.102 1/day 1.68 1/day 0.1 1/day 1.5 1/day
By chlorine reaction with pipe wall. 0.132 m/day 2.072 m/day
0.06 m/day 1.52 m/day
The Chlorine decay curve describes the evolution of chlorine in
contact with water (Figure 2).
When chlorine contact with water, generates reaction with
reducing compounds, these substances
can be dissolved or suspended. The compounds that act with
chlorine are hydrogen sulfide,
manganese, iron and nitrites. The additional chlorine begins to
react with organic matter, the
organic chlorine compounds are produced from this reaction. The
organic chlorine does not have
the ability to disinfect and generates an odor and flavor
characteristic. The chlorine continues
reacting with reducing substances, organic matter and ammonia.
Finally, the additional chlorine will
remain as free chlorine available that is a very active
disinfectant. Once reached this point, all the
nitrogen compounds have been destroyed and therefore, any
further addition of chlorine causes an
increase in the level of free chlorine in the water (AEAAS,
1984). Therefore, chlorine decays once
introduced into the WDN and exist the risk that in certain zones
the network could be unprotected
with the corresponding risk to the health of the consumers. The
quality of the drinking water
-
4
depends on the integrity of the WDN. Maintaining appropriate
levels of quality becomes a primary
task due to the impact on the health of consumers.
Figure 2. Chlorine decay curve.
2.2 Booster chlorination stations
Normally, this BCS are incorporated to the WDN in order to
maintain the disinfection in zones
where the free chlorine is not enough the minimum concentration
limit according to the standards
(Islam et al., 2013). To reduce the risk of the presence of
pathogen microorganisms, it is propose
the installation of CBS (Figure 3) in strategic locations in the
WDN to maintain the minimum
permissible chlorine concentration in the entire network during
all the day in the conditions
mentioned on the chapter one.
Figure 3. Typical booster chlorination station.
-
5
3. OPTIMAL BOOSTER DISINFECTION MODEL
The model to obtain an optimal location of BCS is analyzed with
the heuristic technic of GA. The
optimization propose a regular concentration in the entire
network inner the permissible limits of
free chlorine, both minimum (0.20 mg / L) and maximum (1.50 mg /
L) specified in the Official
Mexican Standard NOM-127. Every node of the network is analyzed
during 24 hours of
consumption to ensure efficient use of disinfectant based on the
NOM-127. Depending on the range
established, it will be establish the optimal scenario for the
efficient use of disinfectant to be within
the limits throughout the distribution network and does not
affect the health of the consumers by the
consumption of water with high concentration and to consumer
water without disinfectant.
3.1 Genetics algorithms
The GA are adaptive methods that can be used to solve
specialized problems of search and
optimization (Beasley et al., 1993). The AG are based on the
genetic processes of biological
organisms. For many generations, natural populations tend to
evolve according to the principles of
Natural Selection, in the standard: "The survival of the
fittest", established by Charles Darwin in his
work: "The origin of species". GA (originally called "genetic
reproductive plans") were developed
by John H. Holland in the early 1960s in order to solve problems
of machine learning.
The basic algorithm considers the following steps:
1.- Generate (randomly) an initial population.
2.- Calculate the fitness of each individual.
3.- Select (sample) on the basis of aptitude.
4.- Apply genetic operators (crosses and mutation) to generate
the next population.
5.- Cycle until some condition is satisfied.
GA uses a direct analogy with the natural behavior. The GA work
with a population of individuals,
each individual represents a feasible solution to a given
problem. Each individual obtain a score
depending on how good is the solution that represents for the
given problem. In the nature, the score
of each individual is equivalent to the effectiveness of an
organism to compete for certain resources.
The higher the adaptation of an individual to the problem, the
greater the probability to be selected
to reproduce, crossing their genetic material with another
individual selected in the same way. This
crossing will produce new individuals, which share some of the
characteristics of their parents. The
lower the adaptation of an individual, the less probability that
the individual be selected for
reproduction, and therefore its genetic material is not spread
over successive generations and then
die.
In this way, it is produce a new population of possible
solutions. This population replaces the
previous one and the properties of this new generation must
contain a higher proportion of good
features in comparison with the previous population. If the GA
has been well designed, the
population will converge toward an optimal solution of the
problem.
3.2 Optimal locations propose by GA
The main focus of this paper is to propose the minimum number of
BCS necessary to maintain the
optimal levels of free chlorine on the standard limits. The
objective is to economize the operation of
the disinfection maintaining the concentrations in all the
networks near to the minimum level.
-
6
Besides, the health of the consumer must be warranted
considering that the level of disinfection
never is going to be under the low limit and the concentration
near to the DWTP is going to be far
from the superior limit of free chlorine.
The GA process requires multiple iterations for optimal results.
This specific algorithm is program
in MATLAB environment running sequences of GA. The analysis of
the free chlorine was
simulated in extended period with the computer program EPANET
created by the United States
Environmental Protection Agency (Rossman, 1996).
EPANET requires the following data: A) length, diameter and
roughness coefficient of the pipe
network. B) Demands and elevations of nodes. C) Characteristics
of tanks and pumps. D) The curve
of demands represented by the multipliers demands on the
consumers. E) The initial quality of the
tanks and nodes. F) Reaction coefficients of chlorine in the
flow and the pipe wall (Rossman, 1996).
The algorithm of the GA proposed for the analysis considers that
every node of the WDN simulates
a BCS providing a value of additional supply concentration of
chlorine. The concentration values
provide from the BCS are the variables for the GA. The
simulation time depends on three factors: 1)
the number of variables for each individual, 2) The methods
including on the GA process: crosses,
selection, mutation and others, and 3) Number of generations to
evaluate. In this case, it is propose
8 variables for the free chlorine concentration between 0.2 and
1.5 mg/L (0.2, 0.4, 0.6, 0.8, 1.0, 1.2
and 1.5 mg/L). It is consider that this number is appropriate
for the time of simulation of the
algorithm proposed on the GA, including the search for the
investment for the hydraulic and quality
function.
The variables are coded in binary numbers from 0 to 7. With
these 8 different variables the binary
numbers use the base of 2 and the power of 3, in order to
contain the 8 variables, without exception
(23 = 8). Therefore, it will have 3 bits for each variable in
binary code (Table 2). If the variable on
GA obtain a zero concentration for any node means that this node
does not supply chlorine,
consequently in that node it does not requires a booster
chlorine disinfection.
Table 2: Binary code for the variables of chlorine
concentration
Variable (mg/L)
Binary
Value
0.0 000
0.2 001
0.4 010
0.6 011
0.8 100
1.0 101
1.2 110
1.5 111
-
7
3.3 Aptitude function
The Aptitude function is applied to an individual in order to
determine the effectiveness of the
solution proposed by that individual. The higher value of the
aptitude function, the best solution of
the individual for the use of BCS. Three main aspects to obtain
an optimal solution are: A) to
maintaining the free chlorine concentration in the range
established by the NOM-127. B) to
maintain a low number of BCS. A low number of BCS implies a low
investment cost. C) the BCS
proposed by the algorithm of GA the minimum quantity of chlorine
in the range of standard limits
in order to warranty the minimum values of free chlorine in
every node of the network in any hour
of the day. According to these considerations the aptitude
function is propose in the equation [1].
[1]
Where:
= Minimum chlorine concentration. = Maximun chlorine
concentration. = Mean chlorine concentration for the node i. =
Booster chlorine disinfection installation cost = Penalization cost
due to the range concentration out of the standard
limits cmin, cmax.
= Concentration of chlorine out of range of the standard limits
of the node i (cmin, cmax)
= Number of nodes in the network.
The values of the aptitude function improves when the average
concentration in each node gains on
to the value of the minimum concentration during the analysis.
The objective of the algorithm is to
maintain the minimum concentration of the nodes in order to
obtain the optimal dosage of chlorine
from the BCS. On the other hand, the aptitude function tends to
decrease according to the behavior
of two aspects: 1) with a large number of BCS proposed by an
individual and 2) when the
concentrations in the nodes are out of range of the standard for
the NOM-127. Finally, the standard
deviation implemented in the aptitude function is focused on the
mean concentrations of the nodes
near to the minimum permissible value of 0.2 mg/L.
4. APPLICATION EXAMPLES
4.1 Hanoi WDN
To illustrate the performance of the algorithm will stage the
network of literature given by HANOI,
with the following characteristics (figure 4):
Number of nodes: 31
Number of wter pipes: 34
Pipe diameters: 12 - 40
Global friction coeff (H-W): 125
-
8
Figure 4. General layout of Hanoi network.
To simulate extended period we used demand multipliers that are
commonly used to simulate WDN
of Mexico City, taken from the book "Datos bsicos" of Comisin
Nacional del Agua (CNA,
2007b), these factors are detailed in the follow figure.
Figure 5. Demand multipliers reference.
For chlorine reaction effects during the 24 hours, the initial
concentration was defined in the tank
outlet point with a value of 0.65 mg/L and initial
concentrations on all nodes of 0.26 mg/L. The
chlorine reaction coefficients were defined by -0.52 and -0.87
for bulk and wall respectively, and
were elected by the type of material and water residence times
on the network to form a network
with characteristics of quality problems due to their high decay
chlorine.
At certain times of the day, an excessive chlorine supply is
used and some nodes do not reach the
minimum concentration required by the rules (figure). A very
frequently solution, in order to meet
the pre established norms, is to increase the supply of chlorine
to the tank outlet and quite possibly
all nodes are within the established range, but this means that
all nodes provide water with high
chlorine levels for most nodes.
-
9
Figure 6. Simulation of concentrations on some nodes.
PROPOSAL (Example 1):
Due to this, we propose the diminished supply of chlorine into
the tank and place the fewest number
of booster chlorination stations within the network in order to
keep the network with minimal use
of the disinfectant.
Program runs with 600 individuals and 180 generations, and the
best result is to use 2 booster
stations as follows (figure 5, table3):
Table 3: Chlorine supply in each booster station.
Figure 7. Location of booster chlorination stations
Booster station
Location Supply (mg/L)
1 Node 10 0.33
2 Node 29 0.25
Chlorine station 1
Node 10
Chlorine station 2
Node 29
-
10
Using chlorine booster stations, we keep the network with
minimal concentration meet the
standards for optimal use of chlorine in water (table 4).
Table 4: Mean concentration values in both simulations
Mean concentration throughout the simulation (mg/L)
Normal
simulation
Simulation using booster chlorination
stations
Tank supply 0.65 0.33
Node 13 0.42 0.26 Node 30 0.46 0.24 Node 16 0.44 0.25 Node 31
0.5 0.25
We reduce chlorine supply across the network regulating high
consumption to all nodes using
booster chlorination stations.
Figure 8. Simulation of concentrations on some nodes from
proposal.
4.2 Example 3 of Epanet manual WDN
The network used in the simulations to this scenario was an
example network from the EPANET
program manual, net3.net (Fig. 8), consists of the following
components:
2 reservoirs (river and lake)
3 tanks
2 pumps
117 pipes
91 nodes
1 general demand pattern and other 4 to certain nodes.
-
11
Figure 9. General layout of network.
The added input values for the water quality were bulk decay and
wall decay coefficients and initial
chlorine concentrations for reservoirs, tanks and nodes (Table
5). The latter were adopted the latter
were adopted to ensure that the network has a stage with
conditional quality require the use of this
optimization tool for the improvement and reduction of chlorine
present in the network.
Table 5: input values to this simulation
Description Value
Initial chlorine concentration in river 0.89 mg/L
Initial chlorine concentration in lake 1.02 mg/L
Initial chlorine concentration in nodes (general) 0.56 mg/L
Initial chlorine concentration in Tank 1 0.48 mg/L
Initial chlorine concentration in Tank 2 0.48 mg/L
Initial chlorine concentration in Tank 3 0.48 mg/L
Global Bulk decay coefficient (1st grade) -0.45 1/day
Global Wall decay coefficient (general) -0.28 m/day
Chlorine reaction coefficient in tank 1 -0.34 1/day
Chlorine reaction coefficient in tank 2 -0.41 1/day
Chlorine reaction coefficient in tank 3 -0.11 1/day
Lake
River
Tank 1
Tank 2
Tank 3
-
12
Because of the long lengths of pipe that network, the farthest
nodes to reservoirs (lake, river) lead
only sufficient concentration to maintain above the minimum
level of chlorine concentration in the
standards. To do this, need to provide a high concentration in
the tank to satisfy these
concentrations. Assuming network years ages of age more, the
reactions will be more severe, so that
an extra increase in dosage is required in the reservoir and
this causes excess chlorine consumption
in the points near the reservoirs.
Figure 11. Average reaction rates.
Figure 10. Contour plot at peak consumption hour. PROPOSAL
(Example 2):
Each node has 8 different possibilities to deploy a station
chlorination (Table 2), so we have a total
of 8 ^ (91 nodes + 3 tanks) = 7.77*1084
alternatives to find a suitable solution using booster
chlorine
stations.
Start the program with 180 generations and 1200 individuals,
which are 216,000 assessments, that
represent 2.78 * 10-78
% of alternatives to choose from, obtaining the following
proposal:
Table 6: Chlorine supply in each booster station.
Chlorine booster station
Location Supply (mg/L)
1 Node 131 0.23
2 Node 145 0.35
3 Node 209 0.4
4 Node 215 0.38
-
13
Figure 12. Location of booster chlorination stations.
To achieve low chlorine consumption, supply in the 2 reservoirs
is decreased and with the addition
of chlorine by reinjection stations, we have a network of more
balanced behavior in compliance
with the limits set by the rules, whereby each node will have
only enough chlorine concentration
(fig. 9)
Figure 13. Contour plot at peak consumption hour with booster
chlorination stations
-
14
5. RESULTS
On economic issues, the use of two stations is more expensive
than using just one, but discards the
effect of high concentrations still have to get the chlorine at
the outermost points of the reservoirs,
thereby using two stations nodes with interior points in the
network allows not allocate extra
concentration on nodes where it is not required.
In Hanoi network we reduce 50.77% the use of chlorine in the
tank and therefore also the
concentration decreased by about 48.2% on most network nodes.
Average concentrations in the
most notorious nodes is kept near the minimum limit down in this
network so consumers have less
chlorinated water. The original proposals by the program are
0.25 for season 1 and 0.4 for, it was
changed last, because the program determines the best location
of stations and the lower
consumption of chlorine, but having a limited range of
possibilities (Table 2) can further reduce this
dosage.
Also for example network 3 manual EPANET, supply tank fell and
remained a more balanced
network as a low-chlorine. The concentration in reservoirs was
reduced Most of the network nodes
were supplied with the required amount of chlorine (table
7).
Table 6: Chlorine supply in each booster station.
Decreased chlorine supply in reservoirs (mg/L)
Reservoir Normal
simulation
Simulation using booster chlorination
stations
River 0.89 0.33
Lake 1.02 0.26
In the development of (Figure 13) get better reaching proposals
as they increase the generations.
When you have a large number of variables, it tends to increase
the number of individuals, to
maintain the diversity of individuals and can perform searches
on those who are improving in their
fitness. Using AG significantly reduces the number of
simulations to find a better option in terms of
limited use of chlorine.
Figure 14. Evolution of Genetic algorithms in a scenario (eg
application 2)
-
15
CONCLUSIONS
Maintain the chlorine concentrations in the standards for
drinking water becomes a complex
concern, that requires an optimal infrastructure of the WDN and
operation of the MWDS, besides a
lot of samples in the network, conditions that does not have
many small cities in Mexico. Therefore,
in this paper is propose a numerical algorithm using heuristic
techniques to verify required optimal
location of BCS.
The validation of the GA in two networks to obtain the optimal
number and locations of BCS to
maintain the minimum chlorine concentration specified by the
NOM-127.
The propose of a BCS with a minimum dosage led the MWDS to
maintained inner the minimum
limit of 0.20mg/L avoiding the excessive use of disinfectant but
enough to combat pathogenic
microorganisms at the time. The GA applied to finding the
optimum BCS have alternative solutions
to use the least amount of residual chlorine in the network and
at the same time to provide a quality
service to consumers of drinking water.
Genetic algorithms have greater effectiveness depending on the
fitness function used. Basically, this
function is carefully chosen to assign a better measure of
fitness to those networks with
concentrations close to the minimum and the minimum number of
stations.
In this paper is propose an alternative aptitude factor that
considerate the specific conditions of
quality and chlorine concentration outside the permissible
limits and the use of drinking water on
Mexico.
REFERENCES
AEAAS, 1984. Manual de la cloracin, Asociacin Espaola de
Abastecimientos de
Agua y Saneamiento, Editorial A. E. A. A. S., 32pp.
Alcocer Yamanaka, V. H., & Velitchko, T. G. (2004). Modelo
de calidad del agua en
redes de distribucin.
Beasley, D., Martin, R. R., & Bull, D. R. (1993). An
overview of genetic algorithms:
Part 1. Fundamentals. University computing, 15, 58-58.
Castro, P., & Neves, M. (2003). Chlorine decay in water
distribution systems case
studylousada network. Electronic Journal of Environmental,
Agricultural and Food Chemistry, 2, 261-266.
Comisin nacional del agua. (2007)., Modelacin Hidrulica y de
Calidad del Agua en
Redes de Agua Potable. Manual de Agua Potable y Alcantarillado y
Saneamiento.
Mxico.
Comisin nacional del agua. (2007b). Datos Bsicos. Manual de Agua
Potable y
Alcantarillado y Saneamiento. Mxico,
Geldreich, E. E. (1996). Microbial quality of water supply in
distribution systems. CRC
Press.
Holland, J. H. (1992). Algoritmos genticos. Investigacin y
Ciencia, 192, 38-45.
-
16
Islam, N., Sadiq, R., & Rodriguez, M. J. (2013). Optimizing
booster chlorination in
water distribution networks: a water quality index approach.
Environmental monitoring
and assessment, 1-16.
NOM, N. O. M. (1996). 127-SSA1-1994. Salud ambiental, agua para
uso y consumo
humano. Lmites permisibles de calidad y tratamientos a que debe
someterse el agua
para su potabilizacin. Mxico, DF: Diario Oficial de la
Federacin, 18.
Ozdemir, O. N., & Erkan Ucaner, M. (2005). Success of
booster chlorination for water
supply networks with genetic algorithms. Journal of Hydraulic
Research, 43(3), 267-
275.
Propato, M., & Uber, J. G. (2004). Vulnerability of water
distribution systems to
pathogen intrusion: How effective is a disinfectant residual?.
Environmental science &
technology, 38(13), 3713-3722.
Rossman, L. A. (2000). EPANET 2: users manual.