Renewable Energy 117 (2018) 404e413

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Optimal design of water supply networks using an energy recovery

approach
Gustavo Meirelles Lima*, Bruno Melo Brentan, Edevar Luvizotto Jr.
Laboratory of Computational Hydraulics, School of Civil Engineering, Architecture and Urban Planning, University of Campinas, Av. Albert Einstein, 951,

ria, Campinas, Brazil
Cidade Universita

a r t i c l e i n f o a b s t r a c t

Article history: Water Distribution Networks (WDNs) represent a major investment for water supply systems devel-
Received 13 May 2017 opment. The standard procedure for their design is to search for the minimum cost, which is produced by
Received in revised form the smaller diameters capable of maintaining the minimum required pressure. However, some District
10 October 2017
Metered Areas (DMAs) have a significant topographic elevation difference, and even if the minimum
Accepted 23 October 2017
Available online 27 October 2017
diameters are reached, the pressure remains high, and a Pressure Reducing Valve (PRV) is necessary.
However, if the pipe diameters of the network are increased, distributed headloss will be reduced, and
this additional energy will be locally dissipated in PRVs to maintain the pressure in the DMA below the
Keywords:
Water distribution networks
maximum allowed value. If a turbine is installed instead, the dissipated energy can be used, creating a
Pumps as turbines benefit that can justify the additional investment due to the diameter increment. Therefore, this paper
Energy recovery presents a method for the optimal design of a WDN considering energy recovery. The use of Pumps as
Energy efficiency Turbines (PATs) is considered for energy production. The optimal design is obtained using a two-level
Optimization optimization procedure: the first is used to obtain pipe diameters, and the second is used for PAT se-
lection. Particle Swarm Optimization is used, and two case studies are presented.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction estimation of the gradient direction, which in some problems, is not

The pipes of water supply networks represent the largest in-
vestments for the deployment of these systems. Therefore, their
design is made to achieve the minimum costs, which is obtained
using the minimum diameters necessary to maintain the quality
standards of pressure and velocity. The calculation of these pa-
rameters in looped networks is performed using a simultaneous
solution of a system of equations, which are primarily non-linear.
The deterministic solution is difficult to obtain and usually re-
quires the linearization of hydraulic equations, which was virtually
impossible until recent computational advances [1]. Nevertheless,
even with the development of software for hydraulic modeling, the
design of water supply networks continued to depend on engi-
neer's experience. To avoid this dependence and the uncertainties
involved in this design process, the hydraulic model was associated
with optimization techniques to obtain the network design with
minimum costs. However, classical techniques usually require the
* Corresponding author.
E-mail addresses: limameirelles@gmail.com (G. Meirelles Lima), brunocivil08@
gmail.com (B.M. Brentan), edevar@fec.unicamp.br (E. Luvizotto).
simple and involves simplified numerical treatment. Therefore,
meta-heuristic methods have been widely applied as an alternative
for classical methods. Different techniques have been applied to the
optimal design problem with good response [2e7]. Due to its un-
restricted characteristic, the restrictions are evaluated separately
using penalty functions and are added to the objective function to
be minimized. According to [8], the penalty function can be neither
too hard, to avoid a wide search across the space, nor too soft,
which would lead to unfeasible solutions.
Pressure control is another hydraulic problem of water supply
networks that uses optimization techniques to improve system
efficiency [9,10]. developed methods for determining the optimal
location and operation of pressure reducing valves (PRVs). Inspired
by the energy recovery obtained in reverse osmosis for seawater
desalination [11,12], wastewater treatment [13] and industrial
process [14], several studies showed that PRVs can be replaced for
microturbines with this purpose [15e17]. Furthermore, due to the
low power potential observed in these valves, it is recommended to
use pumps as turbines (PATs), which has a low cost and good effi-
ciency [18e20]. The existing potential in water distribution net-
works can contribute significantly to energy production with low
environmental impact.

https://doi.org/10.1016/j.renene.2017.10.080
0960-1481/© 2017 Elsevier Ltd. All rights reserved.

https://freepaper.me/t/511303 : ‫خودت ترجمه کن‬

 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413 405

Based on the information from the Piemonte region - Italy, and

using a geographic information system [21], estimated 1447 energy
potentials in water supply systems, which can produce 14e26 MW
[22]. highlighted Switzerland with 90 micro-hydropower plants of
this type and Morocco, where 8% of hydroelectric energy is ob-
tained in water supply systems. Operation under a dynamic sce-
nario of consumption, with daily and seasonal variations, is the
major challenge for the advancement of this technology, as the
variation of available power produces both electrical and hydraulic
issues [23e26]. highlight the long-term variation in demand, as
population and economic growth leads to an increase in con-
sumption. However, the development of more efficient equipment,
such as dishwashers or washing machines, reduces it. This variation
can affect the PAT operation, reducing the produced energy and
even causing water shortage.
As previously described, the traditional procedure used for
network design searches the minimum diameters possible. This
approach reduces the available energy in the system, as small di-
ameters produce higher headloss. In this scenario, the pressure
becomes more sensitive to consumption variations, which in-
creases the available power amplitude and makes the PAT opera-
Fig. 1. Karman-Knapp circle [32].
tion even more difficult [27,28]. studied the energy recovery in
water mains, where a small addition to the investment of pipes
with larger diameters was significantly surpassed by the increase in (2).
energy production [29]. highlighted that the traditional design,
despite the reduced investment, cannot be assumed to be the best q ¼ atan1 ðx  pÞ (1)
due to the dynamic behavior of a water supply operation during its
life cycle. Therefore [30,31], proposed a multiobjective design,

taking into account energy consumption, demand variation and h ¼ WH$ q2 þ 1 (2)
pressure deficiency in addition to pipe costs to obtain more realistic
scenarios of network design. where q is the dimensionless flow coefficient, h is the dimension-
Considering the limitations of the traditional procedure for less head coefficient, x is the independent variable for the Suter
water network design and the energy potential that exists in PRVs, plan representation (varying from 0 to 2$p), and WH is the
this paper proposes the network design associated with the energy dimensionless head coefficient in the Suter plan. This procedure
recovery approach. Thus, the optimal design is obtained via the results in dimensionless curves, which are shown in Fig. 2-b. A
minimization of costs to install a micro-hydropower plant, which is power curve (Equation (3)) was adjusted for this new representa-
added to pipe costs while discounting the benefits obtained using tion. Since the specific speed of the machine is likely to be different
energy production during the plant life cycle. The costs of the micro for the 14 available curves, its characteristic curve can be obtained
hydropower plant are calculated according to the available power, via interpolation [35], polymorphism [36] or just adopting the
and the use of PATs is considered for energy recovery. Long-term curve with the nearest specific speed [37], as used in this paper.
forecasting is performed to predict both the demand and energy Therefore, using only the machine Best Efficiency Point (BEP), its
tariff variations through the plant life cycle. To solve the related characteristic curve in the turbine mode can be obtained.
optimization problem, the meta-heuristic technique Particle
Swarm Optimization (PSO) is used. Two case studies of Brazilian h ¼ a$qb (3)
District Metering Areas (DMAs) are presented, and the results ob-
tained using the energy recovery for optimal design are compared where a and b are the power curve adjustment coefficients.
with those obtained using the traditional approach of economical
design. 2. Optimal design

1.1. Pump as turbine modeling 2.1. Long-term forecasting

Hydraulic machines can operate under eight different operation
zones, characterized by head, flow, torque and speed direction.
Each zone is characterized by the direction of each parameter, as
presented by Ref. [32] in Fig. 1.
The mathematical representation of this circle is usually ach-
ieved by using curves in the Suter plan [33]. To obtain these curves,
as shown in Fig. 2-a, complex and expensive tests are necessary.
Therefore, in this paper, 14 curves obtained by Ref. [34] are used.
These curves are in the specific speed range from 0.46 to 4.94, and
according to affinity laws, they can be used for different machines
with the same specific speed. For each available curve, the opera-
tion as turbine zone (empty markers in Fig. 2-a) was identified, and
considering nominal speed, new curves relating the dimensionless
coefficients of flow, q, and head, h, were obtained via Eqs. (1) and
Social, economic and climatic factors can significantly affect the
water demand [39]. Considering that the life cycle of a PAT is
approximately 20 years [40], there is a great period for demand
changes, which can affect energy production, as described by
Ref. [26]. To create this dynamic scenario, three demand patterns
are necessary: a daily pattern, which represents the demand vari-
ations in a 24-h period; a monthly pattern, to evaluate the seasonal
behavior of the system; and an annual pattern, to estimate the
increase or decrease of demand on a long-term basis. Fig. 3 shows
the daily and monthly patterns used in case studies, which were
obtained from real water distribution systems of Brazil. The daily
pattern in Fig. 3-a shows the typical residential behavior, while the
monthly pattern in Fig. 3-b presents a reduction in consumption
during the winter.

https://freepaper.me/t/511303 : ‫خودت ترجمه کن‬

406 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
Fig. 2. a) Head (WH) and torque (WB) curves in the Suter plan; b) Dimensionless
curves in the turbine mode for different specific speeds [38].
For the annual forecasting, a correlation between water and
energy consumption was made using the available data for S~ ao
Paulo, from 1995 to 2014, which were obtained from Refs. [41,42].
Fig. 4-a shows a moderate correlation between water and energy
consumption for the period, with a Pearson coefficient of 0.454.
Considering the high uncertainties in long-term demand fore-
casting [43,44] and the fact that for residential sectors, there is a
good correlation between water and energy consumption [45], the
growth rate expected for energy, projected in Ref. [42], of 2.2% per
year is also applied for water demand. Even so, since dry seasons
are likely to have a drop in consumption due to rationing and
governmental policies, the probability of this event was also
considered, and a reduction rate of 1.5% per year (average observed
between 1997 and 2005) was used in this case. These dry seasons
created an initial discontinuity in the projected consumption and
four additional cases of reduction. Fig. 4-a shows the long-term
forecasting obtained using this approach.
Energy tariff is also an important parameter to be considered
when the benefits of energy recovery are calculated. Considering
the application in Brazil, where the electrical system is primarily
composed of hydropower plants, the readjustment of tariffs is very
https://freepaper.me/t/511303
Fig. 3. Demand curves: a) Daily pattern; b) Monthly pattern.
sensitive to climatic changes and can have major variations in short
periods, as shown in Fig. 4-b. Between 2010 and 2015, in addition to
the dry season, economic and political issues also affected the
tariffs [46,47]. examined this dubious scenario, highlighting that
the available forecasting methods are used only for a few years
ahead. Therefore, due to its high volatility, the forecasting of energy
tariffs is as complex as water demand forecasting. In this way, the
average inflation observed in the last 10 years, of 6.92% per year
[48], was used to readjust energy tariffs in simulations, which is a
conservative assumption because it is the base value for the annual
readjustment practiced by the companies.
2.2. Problem formulation
The optimal design problem combined with the energy recovery
approach can be solved using the minimization of an objective
function, defined as the sum of pipe and micro-hydropower plant
costs discounted by the benefits obtained via energy production, as
shown in Equation (4).
OF ¼ minðPC þ MC  EBÞ (4)
where OF is the objective function, PC is the pipe costs, MC is the
micro-hydropower plant cost, and EB is the energy recovery
: ‫خودت ترجمه کن‬
 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413 407

Fig. 4. Long term forecasting: a) Energy and water consumption correlation and their long-term forecasting; b) Historical energy tariff readjustments [42].

benefits. plant power were valid for the 5e224 kW range, allowing the
Pipe costs are a function of their diameter and length and can be calculation of its costs through Equation (6).
calculated using Equation (5).
MC ¼ 2:246$P 2 þ 338:23$P þ 8218:79 (6)

X
NP
where P is the available power.
PC ¼ DCi $Li (5)
Finally, the benefits obtained with the energy recovery through
i¼1
the PAT life cycle should be adjusted to present values. Therefore,
where NP is the number of pipes in the network, DCi is the unitary the Net Present Value (NPV) is used, per Equation (7).
cost of pipe i, and Li is the length of pipe i.
P 
i¼1 r$g$Qi;t $Hi;t $hi;t
8760
The micro-hydropower plant costs can be divided into three X
n $TEt
categories: project, civil construction and electromechanical EB ¼ NPV ¼ t
(7)
t¼1 ð1 þ iÞ
equipment costs. In this work, the costs obtained by Ref. [49],
adjusted by the period inflation, are used. The costs related to the

https://freepaper.me/t/511303 : ‫خودت ترجمه کن‬

408 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413

Fig. 5. Flowchart for network design using the energy recovery approach.

where n is the PAT life cycle, r is the water specific mass, g is the
gravitational acceleration, Qi,t is the PAT flow at hour i of year t, Hi,t
is the PAT head at hour i of year t, hi,t is the PAT efficiency at hour i of
year t, TEt is the energy tariff in year t, and i is the annual interest
rate.
To solve this optimization problem, the meta-heuristic tech-
nique PSO, developed by Ref. [50], is used [51]. studied four
different meta-heuristic techniques for the optimal design prob-
lem, and similar results were obtained. In addition [52], showed
that the penalty functions used to create the boundaries of the
problem are as important as the optimization technique used.
Nevertheless, PSO has been widely used to solve different problems
of water supply systems with good results [38,53,54]. This tech-
nique is based on the behavior of birds searching for food. Each
particle of the method represents a possible solution, with an initial
position (solution), X, and velocity, V, randomly defined. Over it-
erations, the velocity and position of each particle are updated
using Equations (8) and (9), taking into account its best position
ever found, P, the best position ever found by the swarm, G, and its
inertia, u. The coefficients c1 and c2, cognitive and social

Fig. 6. Laudissi DMA layout and the PAT location.

https://freepaper.me/t/511303 : ‫خودت ترجمه کن‬

 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
Fig. 7. Energy recovery feasibility for the Laudissi DMA design.
parameters, respectively, are adjusted through iterations to
perform a better global search for the initial iterations (higher value
of c1), and a refined local search at the end of the process (higher
value of c2).
Fig. 8. Optimization results for the standard scenario in Laudissi DMA: a) Objective function evolution; b) Pipe diameters; c) Pressure on the critical node.
https://freepaper.me/t/511303
409



Pik  Xik G  Xik
Vikþ1 ¼ u$Vik þ c1 $rand1 $ þ c2 $rand2 $ (8)
Dt Dt
Xikþ1 ¼ Xik þ Vikþ1 $Dt (9)
For the described design problem, two optimization stages are
performed: In the first, PSO particles are randomly initialized with
pipe diameters, and the first hydraulic simulation is carried out
without the micro-hydropower plant. Thus, knowing previously
the local for its installation, flow and pressure data obtained from
this first simulation are used to select an adequate PAT. In the
second stage, a frequency curve of power is created, and a second
optimization is performed to identify the PAT BEP, which maxi-
mizes the energy produced through its life cycle. The boundary of
this optimization is the minimum flow necessary for PAT operation.
In this study, a minimum of 50% of BEP is considered to avoid low
efficiency and the appearance of vibration problems [55]. After the
BEP is defined, the first stage continues, using the result to deter-
mine the characteristic curve of the PAT (Equation (3)). A new hy-
draulic simulation, now with the micro-hydropower plant already
installed, is performed, allowing the evaluation of the benefits
obtained with energy recovery and the violation of boundary
conditions, represented by the minimum pressure allowed in the
network, which will be added to the objective function as a penalty.
Finally, the costs are calculated, and the objective function is
: ‫خودت ترجمه کن‬
410 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
evaluated. Fig. 5 shows a flowchart that summarizes the procedure
for network design using the energy recovery approach. The
routine was developed in MatLab, using the Optimization Toolbox
to perform the PSO. It is important to highlight that leakage was not
considered in this work, which would make the optimization
problem even more complex due to the extra benefit/cost evalua-
tion of a reduction/increase in leakage volume.
3. Case study
3.1. Laudissi DMA
This case study presents a real DMA of the city of Santa Ba
rbara
do Oeste - SP, Brazil. This DMA has 25 nodes and 31 pipes, and a PRV
is already installed for pressure control. The average consumption
for the initial period of operation is 67.5 l/s. Fig. 6 shows the DMA
layout and the location of the PAT.
To evaluate the influence of the minimum pressure allowed and
energy tariff, which are the most sensitive to energy recovery
feasibility, different values were adopted for these parameters. For
minimum pressure, the range was defined between 10 m, which is
the minimum value defined for Brazilian systems [56], and 40 m.
For the initial energy tariff, a range of 0.15e0.60 $/kWh was
defined, as it encompasses the different energy charging systems
available in Brazil. A total of 110 scenarios were evaluated
combining different values for minimum pressure (10, 13, 16, 19, 22,
25, 28, 31, 34, 37 and 40 m) with the initial energy tariff (0.15, 0.20,
0.25, 0.30, 0.35, 0.40, 0.45, 0.50, 0.55 and 0.60 $/kWh). The PSO was
set using 50 particles, 1000 iterations and a tolerance for objective
function variation of 1012. Using the results obtained in all sce-
narios, Fig. 7 was created. On the left is a zone where the additional
investment to increase pipe diameters and install the micro-
Fig. 9. Campos do Conde II layout and DMA configuration for the PATs location.
https://freepaper.me/t/511303
hydropower plant is surpassed by the benefit obtained using en-
ergy recovery, while on the right, for minimum pressures above
37 m, the traditional design approach presents better results.
Considering a standard scenario, with the minimum pressure of
10 m and an initial energy tariff of 0.30 $/kWh, the additional in-
vestment necessary for the energy recovery is $42,480 (23.5%), but
the benefit obtained with the average power of 3.72 kW is
$297,804, seven times the additional value. Observing the differ-
ence between the average power produced and the nominal power
of the PAT, approximately 35 kW, it can be assumed that only for the
final years of the project is there an effective operation of the PAT.
Therefore, a second machine can be used for the initial years and for
the low consumption periods of the final years to increase the en-
ergy produced. If this new scenario is considered, with a smaller
PAT installed for the initial demand, with the addition of a second
machine after 10 years, the average power increases to 6 kW and
the benefit obtained increases to $480,156, with an additional cost
of only $51,899 (28.7%), which increases the energy recovery
feasibility even more. Due to the uncertainties in water demand
and energy tariff forecasting, a scenario with constant values for
these parameters was simulated, considering a minimum pressure
of 10 m. In this case, the net benefit obtained was significantly
improved to $534,280. This occurs because the PAT can operate on
its BEP during the entire life cycle, increasing the energy produced
and surpassing the drawback of a reduced energy tariff. considering
this result, the worst scenario possible occurs when the water de-
mand increase and the energy tariff remains constant. In this case,
the energy recovery still produces a better result, but the net
benefit is reduced to only $17,080.
Observing the evolution of the objective function for the stan-
dard scenario (Fig. 8-a), it can be noted that at the beginning, the
pipe costs are high due to the increase of their diameter to reduce
the headloss of the system and improve the available power for
: ‫خودت ترجمه کن‬
 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
energy recovery. After some iterations, the pipe diameters remains
practically constant, and the PAT operation point is adjusted to
maximize energy production, which increases the NPV value. At
final iterations, pipe diameters are reduced without major energy
losses since a small reduction cannot create a significant headloss,
while the PAT operation point is finally set to an optimal value. The
micro-hydropower plant costs remained insignificant compared
with others costs for all optimization procedure. Fig. 8-b shows the
increase of the first pipe, as it carries the total flow and is thus more
susceptible to headloss, which would significantly reduce the en-
ergy production. The other pipes show similar diameters, with the
exception of pipe 32, which is significantly larger using the tradi-
tional approach. Since it has one of the minor lengths in the
network, its diameter is increased to compensate the headloss of
other pipes with a low cost for the project. On average, the energy
recovery approach designed a network with 22.9% larger di-
ameters. Finally, Fig. 8-c shows the pressure on the critical node
(node with the minimum pressure) for the project life cycle. Due to
diameter differences, the critical node for the traditional procedure
is node 13, while with the energy recovery approach, node 23
presents minimum pressures. As expected, the pressure decays
over the years due to the increase in demand, and only in the final
year of operation is the minimum pressure allowed in the network
achieved. However, during the operation period, the energy re-
covery approach produced better pressure control. Therefore, even
using the traditional procedure, a pressure control device is
necessary, as during most of the time, the network operates at high
pressures.
Fig. 10. Optimization results for the standard scenario in Campos do Conde II DMA: a) Objective function evolution; b) Pipe diameters; c) Pressure at the critical node.
https://freepaper.me/t/511303
411
3.2. Campos do Conde II DMA
The second case study is also a real DMA, from Piracicaba - SP,
Brazil. This DMA is larger, with 121 nodes, 153 pipes and an initial
demand of 315 l/s. Observing the DMA layout and its elevation
distribution, four minor DMAs can be defined. Therefore, in this
case, four PATs are considered for the design optimization, one at
each new DMA entrance, as shown in Fig. 9.
For this case, only the standard scenario, with an initial energy
tariff of 0.30 $/kWh and a minimum pressure of 10 m, was evalu-
ated. Due to the higher number of variables in this problem, the PSO
was set with 200 particles and the maximum iterations to 5,000,
while the tolerance for the objective function variation was set to
1012. Different from the Laudissi DMA, objective function mini-
mization did not occur due to an increase in energy production over
iterations, as observed in Fig. 10-a. Instead, the energy production
remained practically constant, while the pipe costs were reduced.
This behavior occurs due to a greater relevance of pipe costs in the
objective function. Therefore, economically, it is more interesting to
reduce pipe costs than to increase energy production. Even so, the
additional costs to increase the available power ($305,331) were
surpassed by the energy benefits ($1,961,972) obtained with an
average power of 24.5 kW (sum of the average power of all four
PATs), indicating the feasibility of the energy recovery approach.
Observing the frequency of pipe diameters for each method
(Fig. 10-b), it is noticeable that an increase resulted from the energy
recovery approach. On average, pipe diameters increased by 58.3%,
which surprisingly is higher than the increase observed for the
: ‫خودت ترجمه کن‬
412 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
Laudissi DMA, even with the pipe costs being more relevant in this
case. This can be explained by the interaction of DMAs. Since the
pressure of all DMAs is connected with DMA 1, the pressure drop
caused by PAT 1 reduces the available power of the other three sites.
Therefore, the pipe diameters have to increase to maintain high
available power at these sites. The pressure behavior of the critical
nodes of each DMA was similar to that observed for Laudissi DMA,
as shown in Fig. 10-c, where a decay over the years is observed due
to the increase in demand, reaching the minimum value at the final
year, and the need for a pressure control device during the network
life cycle is once again highlighted.
4. Conclusions
The paper presented a new approach for WDN design, consid-
ering energy recovery. The procedure is based on the minimization
of global costs, represented by the pipes and micro-hydropower
plant, discounted by the benefit obtained using energy produc-
tion. The results showed that the additional costs to increase pipe
diameters are surpassed by the energy recovery benefit. The
sensitivity analysis of the energy tariff and minimum pressure
required performed for Laudissi DMA showed that even with un-
favorable conditions, i.e., low energy tariff and high minimum
pressure in the network, the energy recovery is feasible. In addition,
when two PATs are selected, one to operate with the initial con-
ditions of consumption and the second to be installed after 10 years
of operation, the observed benefits significantly increased. Finally,
it could be observed that for cases with low water demand varia-
tion during the PAT life cycle, the benefit is increased, since the PAT
can operate for longer periods close to its BEP. For the worst case
possible, with a constant energy tariff and a high variation in water
demand, the energy recovery approach remained feasible, but the
net benefit was significantly reduced. The study of Campos do
Conde II DMA confirmed the reliability of the proposed design
method, as the energy recovery was limited due to a higher econ-
omy obtained when the pipe diameters were reduced, maintaining
the commitment to global economic attractiveness. As expected,
due to the higher number of variables, the second case study pre-
sented a higher computational effort, requiring more than three
times the iterations of the first. Even so, the time spent to perform
the design using the energy recovery lasted no more than 10 h on a
regular computer, which is not relevant compared to the benefits
obtained. It is also important to highlight that for both case studies,
water losses were not considered, which can be a significant barrier
to the use of energy recovery. This is mainly due to the poor ca-
pacity of PATs to operate with a reasonable performance under
dynamic conditions, being ineffective for leakage control during
low consumption periods.
References
[1] T.M. Walski, A history of water distribution, J. Am. Water Works Assoc. 98 (3)
(2006) 110e121.
[2] A.R. Simpson, L.J. Murphy, G.C. Dandy, Pipe network optimization using Ge-
netic Algorithms, in: Proceedings of Specialty Conference of Water Resources
Planning and Management Division, American Society of Civil Engineers,
Seattle, Washington, USA, May 1993, pp. 392e395.
[3] Z.W. Geem, J.H. Kim, G.V. Loganathan, A new heuristic optimization algo-
rithm: harmony search, Simulation 76 (2) (2001) 60e68, https://doi.org/
10.1177/003754970107600201.
[4] H.R. Maier, A.R. Simpson, A.C. Zecchin, W.K. Foong, K.Y. Phang, H.Y. Seah, H.Y.,
C.L. Tan, Ant colony optimization for design of water distribution systems,
J. Water Resour. Plan. Manag. 139 (3) (2003) 200e209, https://doi.org/
10.1061/(ASCE)0733-9496(2003)129:3(200).
[5] C.R. Suribabu, T.R. Neelakantan, Design of water distribution networks using
particle swarm optimization, Urban Water J. 3 (2) (2006) 111e120, https://
doi.org/10.1080/15730620600855928.
[6] N. Moosavian, B.K. Roodsari, Soccer league competition algorithm: a novel
meta-heuristic algorithm for optimal design of water distribution networks,
https://freepaper.me/t/511303
Swarm Evol. Comput. 17 (2014) 14e24, https://doi.org/10.1016/
j.swevo.2014.02.002.

rez-García, M. Herrera, M, Water distribution
[7] I. Montalvo, J. Izquierdo, R. Pe
system computer-aided design by agent swarm optimization, Computer-
Aided Civ. Infrastructure Eng. 29 (6) (2014) 433e448, https://doi.org/10.1111/
mice.12062.
[8] Z.Y. Wu, A.R. Simpson, A self-adaptive boundary search genetic algorithm and
its application to water distribution systems, J. Hydraulic Res. 40 (2) (2002)
191e203, https://doi.org/10.1080/00221680209499862.
[9] L.F.R. Reis, R.M. Porto, F.H. Chaudhry, Optimal location of control valves in pipe
networks by genetic algorithm, J. Water Resour. Plan. Manag. 123 (6) (1997)
317e326, https://doi.org/10.1061/(ASCE)0733-9496(1997)123:6(317).
[10] J. Saldarriaga, C.A. Salcedo, Determination of optimal location and settings of
pressure reducing valves in water distribution networks for minimizing water
losses, in: 13th Computer Control for Water Industry Conference - CCWI,
Procedia Engineering, vol. 119, 2015, pp. 973e983, https://doi.org/10.1016/
j.proeng.2015.08.986.
[11] C.F. Wan, T.S. Chuang, Energy recovery by pressure retarded osmosis (PRO) in
SWROePRO, Appl. Energy 162 (2016) 687e698, https://doi.org/10.1016/
j.apenergy.2015.10.067.
[12] A. Altaee, P. Palenzuela, G. Zaragoza, A.A. Al-Anezi, Single and dual stage
closed-loop pressure retarded osmosis for power generation: feasibility and
performance, Appl. Energy 191 (2017) 328e345, https://doi.org/10.1016/
j.apenergy.2017.01.073.
[13] Y. Gu, Y. Li, X. Li, P. Luo, H. Wang, Z.P. Robinson, X. Wang, J. Wu, F. Li, The
feasibility and challenges of energy self-sufficient wastewater treatment
plants, Appl. Energy (2017), https://doi.org/10.1016/j.apenergy.2017.02.069.
[14] Q. Zhang, X. Zhao, H. Lu, T. Ni, Y. Li, Waste energy recovery and energy effi-
ciency improvement in China's iron and steel industry, Appl. Energy 191
(2017) 502e520, https://doi.org/10.1016/j.apenergy.2017.01.072.
[15] M.R. Vilanova, J.A.P. Balestieri, Hydropower recovery in water supply systems:
models and case study, Energy Convers. Manag. 84 (2014) 414e426, https://
doi.org/10.1016/j.enconman.2014.04.057.
[16] I. Samora, V. Hasmatuchi, C. Münch-Allign, M.J. Franca, A.J. Schleiss,
H.M. Ramos, Experimental characterization of a five blade tubular propeller
turbine for pipe inline installation, Renew. Energy 95 (2016) 356e366,
https://doi.org/10.1016/j.renene.2016.04.023.
[17] D. Jiyun, Y. Hongxing, S. Zhizheng, C. Jian, Micro hydro power generation from
water supply system in high rise buildings using pump as turbines, Energy
(2017), https://doi.org/10.1016/j.energy.2017.03.023.
[18] H.M. Ramos, A. Borga, M. Sima ~o, New design solutions for low-power energy
production in water pipe systems, Water Sci. Eng. 2 (4) (2009) 69e84, https://
doi.org/10.3882/j.issn.1674-2370.2009.04.007.
[19] A. Carravetta, G. Del Giudice, O. Fecarotta, H.M. Ramos, Energy production in
water distribution networks: a PAT design strategy, Water Resour. Manag. 26
(13) (2012) 3947e3959, https://doi.org/10.1007/s11269-012-0114-1.
[20] J. Gallagher, I.M. Harris, A.J. Packwood, A. McNabola, A.P. Williams, A strategic
assessment of micro-hydropower in the UK and Irish water industry: iden-
tifying technical and economic constraints, Renew. Energy 81 (2015)
808e815, https://doi.org/10.1016/j.renene.2015.03.078.
[21] C. Soffia, F. Miotto, D. Poggi, P. Claps, Hydropower potential from the drinking
water systems of the Piemonte region (Italy), in: In: 4th Sustainable Energy
and Environmental Protection Conference, 2010. Bari, Ita
lia.
[22] N. Byns, K. Leunis, K. Peeters, L. Tonnet, K.U. Leuven, The use of hydropower in
water supply, in: European Water and Sanitation Services Vs Sustainable
Development, 2011.
[23] N. Fontana, M. Giugni, D. Portolano, Losses reduction and energy production
in water-distribution networks, J. Water Resour. Plan. Manag. 138 (3) (2012)
237e244, https://doi.org/10.1061/(ASCE)WR.1943-5452.0000179.
[24] M. Marchis, G. Freni, Pump as turbine implementation in a dynamic numerical
model: cost analysis for energy recovery in water distribution network,
J. Hydroinformatics 16 (1) (2015) 1e14, https://doi.org/10.2166/
hydro.2015.018.
[25] R. Sitzenfrei, J. von Leon, Long-time simulation of water distribution systems
for the design of small hydropower systems, Renew. Energy 72 (2014)
182e187, https://doi.org/10.1016/j.renene.2014.07.013.
[26] L. Corcoran, A. McNabola, P. Coughlan, Predicting and quantifying the effect of
variations in long-term water demand on micro-hydropower energy recovery
in water supply networks, Urban Water 14 (7) (2017) 1e9, https://doi.org/
10.1080/1573062X.2016.1236136.
[27] A. Afshar, F.B. Jemaa, M.A. Marino, Optimization of hydropower plant inte-
gration in water supply system, J. Water Resour. Plan. Manag. 116 (5) (1990)
665e675, https://doi.org/10.1061/(ASCE)0733-9496(1990)116:5(665).
[28] G. Meirelles, A.N.C. Viana, R.S. Dias Jr., Optimal design of water mains
considering microgeneration” (in Portuguese), in: 25 Congresso Nacional de
Saneamento e Meio Ambiente, 2014. S~ ao Paulo.
[29] B. Coelho, A.A. Campos, Efficiency achievement in water supply systems-A
review, Renew. Sustain. Energy Rev. 30 (2014) 59e84, https://doi.org/
10.1016/j.rser.2013.09.010.
[30] I. Basupi, Z. Kapelan, Flexible water distribution system design under future
demand uncertainty, J. Water Resour. Plan. Manag. 141 (4) (2015), 04014067,
https://doi.org/10.1061/(ASCE)WR.1943-5452.0000416.
[31] J. Marques, M. Cunha, D. Savic, O. Giustolisi, Water network design using a
multiobjective real options framework, J. Optim. (2017) 1e13, https://doi.org/
10.1155/2017/4373952.
: ‫خودت ترجمه کن‬
 G. Meirelles Lima et al. / Renewable Energy 117 (2018) 404e413
[32] T.R. Knapp, Complete characteristics of centrifugal pumps and their use in the
prediction of transient behavior, Trans. ASME (1937) 683e689.
[33] E.B. Wylie, V.L. Streeter, Fluid Transients, EUA (1983) Thomson-Shore, Mich- [46]
igan, 1983.
[34] A.R.D. Thorley, M.H. Chaudry, Pump characteristics for transient flow analysis,
in: BHR Group Conference Series Publication vol. 19, Mechanical Engineering [47]
Publications Limited, Bedford, 1996, pp. 461e476.
[35] J. Izquierdo, P.L. Iglesias, Mathematical modelling of hydraulic transients in
simple systems, Math. Comput. Model. 35 (2001) 801e812, https://doi.org/ [48]
10.1016/S0895-7177(02)00051-1.
[36] G. Meirelles, E. Luvizotto Jr., Method to estimate complete curves of hydraulic
pumps through the polymorphism of existing curves, J. Hydraulic Eng. 143 (8) [49]
(2017), 04017017, https://doi.org/10.1061/(ASCE)HY.1943-7900.0001301.
[37] W. Wan, W. Huang, Investigation on complete characteristics and hydraulic
transient of centrifugal pump, J. Mech. Sci. Technol. 25 (10) (2011) [50]
2583e2590, https://doi.org/10.1007/s12206-011-0729-9.
[38] G. Meirelles, E. Luvizotto Jr., B.M. Brentan, Selection and location of pumps as
turbines substituting pressure reducing valves, Renew. Energy 109 (2017) [51]
392e405, https://doi.org/10.1016/j.renene.2017.03.056.
[39] M.A. Al-Zahrani, A. Abo-Monasar, Urban residential water demand prediction
based on artificial neural networks and time series models, Water Resour.
Manag. 29 (10) (2015) 3651e3662, https://doi.org/10.1007/s11269-015- [52]
1021-z.
[40] Europump, Hydraulic Institute and U.S. Department of Energy, Pump Life
Cycle Costs: a Guide to LCC Analysis for Pumping Systems, Hydraulic Institute,
Parsippany, NJ, 2001. [53]
[41] National System of Sanitation Information - SNIS, Diagnosis of Water and
Sewage Services e 2014, 2014. Technical Report (in Portuguese).
[42] Empresa De Pesquisa Energe
tica - EPE, Brazilian Energy Balance (In Portu- [54]
guese), Ministe
rio de Minas e Energia, 2015, p. 294.
[43] C.T. Osborn, J.E. Schefter, L. Shabman, The accuracy of water use forecasts:
evaluation and implications, JAWRA J. Am. Water Resour. Assoc. 22 (1) (1986)
101e109, https://doi.org/10.1111/j.1752-1688.1986.tb01865.x. [55]
[44] T.M. Fullerton, A.L. Molina, Municipal water consumption forecast accuracy,
Water Resour. Res. 46 (6) (2010), https://doi.org/10.1029/2009WR008450. [56]
[45] S. Jiang, J. Wang, Y. Zhao, S. Lu, H. Shi, H.F. He, Residential water and energy
nexus for conservation and management: a case study of Tianjin, Int. J.
https://freepaper.me/t/511303
413
Hydrogen Energy 41 (35) (2016) 15919e15929, https://doi.org/10.1016/
j.ijhydene.2016.04.181.
H.T. Pao, Forecasting electricity market pricing using artificial neural net-
works, Energy Convers. Manag. 48 (3) (2007) 907e912, https://doi.org/
10.1016/j.enconman.2006.08.016.
R. Weron, Electricity price forecasting: a review of the state-of-the-art with a
look into the future, Int. J. Forecast. 30 (4) (2014) 1030e1081, https://doi.org/
10.1016/j.ijforecast.2014.08.008.
Instituto Brasileiro De Geografia E Estatística - IBGE, http://www.ibge.gov.br/
home/estatistica/indicadores/precos/inpc_ipca/defaultseriesHist.shtm,
accessed in 01/2017.
M.R. Vilanova, Study of Pumps as Turbines and Motors as Generators for Micro
Hydropower Plants, Master Thesis (in portuguese), Universidade Federal de
Itajuba
, Itajub

a, Brazil, 2007.
R.C. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, in:
Proceedings of the Sixth International Symposium on Micro Machine and
Human Science, vol. 1, 1995, pp. 39e43.
D. Mora-Melia, P.L. Iglesias-Rey, F.J. Martinez-Solano, P. Ballesteros-Pe
rez,
Efficiency of evolutionary algorithms in water network pipe sizing, Water
Resour. Manag. 29 (13) (2015) 4817e4831, https://doi.org/10.1007/s11269-
015-1092-x.
I.N. Marchiori, G. Meirelles, B.M. Brentan, E. Luvizotto Jr., J. Izquierdo, Penalty
functions in the optimal design of new water distribution networks” (in
Portuguese), in: Congreso de Me
todos Nume
ricos en Ingeniería, 2017.
Valencia, Espanha.
D. Kang, K. Lansey, Real-time demand estimation and confidence limit anal-
ysis for water distribution systems, J. Hydraulic Eng. 135 (10) (2009) 825e837,
https://doi.org/10.1061/(ASCE)HY.1943-7900.0000086.
B.M. Brentan, E. Campbell, G. Meirelles, E. Luvizotto, J. Izquierdo, Social
network community detection for DMA creation: criteria analysis through
multilevel optimization, Math. Problems Eng. (2017), https://doi.org/10.1155/
2017/9053238.
A. Williams, Pumps as Turbines Users Guide, International Technology Pub-
lications, London, 1995, p. 59. ISBN l-85339-285-5.
Brazilian Association of Technical Standards - ABNT, NBR 12218-Project of
Water Distribution Network for Public Supply, 1994. Technical report (in
Portuguese).
: ‫خودت ترجمه کن‬