You are on page 1of 9

PROOF COPY [IR/2003/022566] 006503QIR

Pumping Selection and Regulation


PR
for Water-Distribution Networks
Patricio Planells Alandi1; Pedro Carrión Pérez2; José Fernando Ortega Álvarez3;
O
Miguel Ángel Moreno Hidalgo4; and José María Tarjuelo Martín-Benito5
OF

Abstract: Because of the increasing importance of on-demand irrigation systems, a support system for general use has been developed
to aid in selecting and regulating pumping stations. This innovation will improve the balance between total costs 共project and energy兲 and
operation quality. The procedure first determines the maximum and minimum system head curves, followed by the evolution of demand
CO
curves to obtain the maximum discharge needed. Once this discharge is determined, it is possible to carry out the dimensioning and
regulation of the pumping station. An easy tool to select the number of variable and fixed speed pumps has also been developed Excel and
Visual Basic can be used. The results demonstrate the importance of selecting pumps that are best adapted to the system head curve. The
minimum total cost solution has been obtained by using one variable-speed pump in conjunction with another operating at fixed speed.
PY

DOI: XXXX
CE Database subject headings: Pumping stations; Optimization; Water demand; Hydraulic networks.
[IR

Introduction hydrants. The problem with this solution is the large number of
possible supply failures to the network 共Lamaddalena and Sagar-
Currently, in many parts of the world, pressure irrigation net- doy 2000; Pulido-Calvo et al. 2003兲. This situation often forces a
/2

works are becoming more common in order to improve and con- change in the rotational delivery system during the peak period of
solidate irrigation systems. However, in an on-demand pressur- water consumption. One of the reasons is that the probability
hypotheses do not really guarantee hydrant equity for any given
00

ized irrigation network, the determination of the discharge and


upstream head that produce the best balance between the total time during the irrigation season. Distortions usually take place
cost of pumping and operation quality is not a completely solved because of aspects related either to the irrigation system 共e.g.,
using only nocturnal sprinkler irrigation兲, to energy tariffs 共e.g.,
3/

problem. The operating quality can be measured through the fail-


different daytime energy costs for different days of the week兲, or
ure probability of the network supply.
to other factors. For optimal regulation, a pumping station must
In many cases, Cléement’s methodology 共Clément and Galand
02

be adapted as closely as possible to the system head curve


1979兲 has been applied for an upstream discharge of the network
共García-Serra and Madalena 1992; López-Luque et al. 1995; Mar-
with a determined head. This head has been calculated without
tínez and Vela 1993兲, thereby minimizing the excess of head for
taking into account the different spatial distribution of the open
25

each demanded discharge.


Various researchers have studied the operation of pumping sta-
1
Professor, Centro Regional de Estudios del Agua, Castilla-La tions in urban water districts 共Duan et al. 1990; Brion and Mays
Mancha Univ., Campus Universitario, s/n. E02071, Albacete, Spain.
66

1991兲. Although Ormsbee and Lansey 共1994兲 reviewed methods


E-mail: Patricio.Planells@uclm.es.
2 used to minimize operation costs, few researchers have studied
Professor, Centro Regional de Estudios del Agua, Castilla-La
Mancha Univ., Campus Universitario, s/n. E02071, Albacete, Spain. the problem of pumping stations in on-demand irrigation net-
works 共Pulido-Calvo et al. 2003兲
]0

E-mail: Pedro.Carrion@uclm.es. Tel.: ⫹34 967 599200. Fax: ⫹34 967


599238. This work presents a support tool for proper dimensioning and
3
Researcher, Centro Regional de Estudios del Agua, Castilla-La regulating pumping stations that supply water to an on-demand
Mancha Univ., Campus Universitario, s/n. E02071, Albacete, Spain. irrigation network. This tool attempts to determine the best bal-
06

E-mail: Jose.Ortega@uclm.es.
4
ance between total pumping costs 共project and energy兲 and opera-
Researcher, Centro Regional de Estudios del Agua, Castilla-La tion quality.
Mancha Univ., Campus Universitario, s/n. E02071, Albacete, Spain.
50

E-mail: mamoreno@alu-ab.uclm.es.
5
Professor, Centro Regional de Estudios del Agua, Castilla-La
Mancha Univ., Campus Universitario, s/n. E02071, Albacete, Spain. Methodology
E-mail: Jose.Tarjuelo@uclm.es.
3Q

Note. Discussion open until November 1, 2005. Separate discussions The initial step in the process is analyzing the maximum and
must be submitted for individual papers. To extend the closing date by
minimum system head curves that limit the possible operating
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos-
points of the pumping station. Next, a process of daily random
IR

sible publication on May 28, 2003; approved on June 21, 2004. This generation of demand curves that consider the specific conditions
paper is part of the Journal of Irrigation and Drainage Engineering, of the network 共hydrant discharge, irrigation time, etc.兲 is devel-
Vol. 131, No. 3, June 1, 2005. ©ASCE, ISSN 0733-9437/2005/3-1– oped. The evolution of these curves determines the maximum
XXXX/$25.00. upstream discharges, which in turn form the basis for dimension-

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005 / 1

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR

PR

Fig. 1. Diagram of Q-H curve of a pump; maximum and minimum


O
demand curves
OF
Fig. 2. Representation of maximum and minimum system head and
daily random demand curves
ing the pumping station. The last step determines the type and
number of variable- and fixed-speed pumps that must work at
each discharge interval to minimize project and operating costs
CO

冉兺 冊
while maintaining proper operation quality in the network.
2
Hi = Han − Ri a jd j 共4兲
j苸Ci
Maximum and Minimum System Head Curves
where Hi = head downstream i line; Han = upstream head i line.
PY

The system head curve is defined by several points Hc-Qc 共head-


• Pressure head excess in the nodes 共Pi / ␥兲 with respect to the
discharge兲 in the upstream node of the irrigation network. It must
setting pressure in the reducing valve 共Pti / ␥兲, defined for each
guarantee the requirements for discharge and pressure in all the
hydrant as
hydrants of a branched network.
The possible operating points of the network are located be-
[IR

tween the most unfavorable 共Hmax兲 and the most favorable system Pi Pti
head curve 共Hmin兲 共Fig. 1兲. = Hi − Hgi − 共5兲
␥ ␥
To determine the maximum and minimum heads that corre-
/2

spond to a certain upstream flow, the discharge of each hydrant in and the function of minimum is applied
the network 共di兲 is multiplied by a coefficient 共ai兲. This coefficient

冉冊
can be 0 or 1, depending on whether the hydrant is closed or open
00

at a given time. In this way, the real discharge of each hydrant 共qi兲 Pi
Min ⇒0 共6兲
for each of the operating conditions supposed is obtained as ␥
3/

q i = a id i 共1兲 where Hg = hydrant elevation; 共Pti / ␥兲 = reducing valve setting


pressure, according to minimum pressure necessary in the hy-
Since the same identification number is assigned to each line
02

drant. It can be different for each hydrant depending on the


and to the node located immediately downstream, the flow in line size and shape of the plot that it supplies and the type of
i共Qi兲 is the sum of discharges for the hydrants that the i line emitter used 共sprinkler or dripper; for example, if, in the same
supplies network, both irrigation methods are used in different plots兲.
25

• The coefficients ai.


Qi = 兺 qi 共2兲 • The upstream discharge Qc:
66

i苸Ci

n
where Ci = number of hydrants fed by line i.
To calculate friction head losses in the network pipes, the Qc = 兺 a id i 艋 Q p 共7兲
]0

i=1
Darcy–Weisbach equation is used
where Q p = pre-fixed upstream discharge; and n = number of
8LiQ2i hydrants in network.
= RiQ2i 共3兲
06

Hri = f i In this way, the values ai that maximize the upstream head are
␲2gD5i
identified, indicating the open or closed hydrants. In a similar
where Hr = friction head losses 共m/m兲; f = friction factor; L = pipe way, the minimum of Hc is obtained, with the same restrictions as
50

length 共m兲; g = acceleration of gravity 共9.81 m / s2兲; D = internal previously mentioned, except the last one since Qc 艌 Q p is neces-
diameter 共m兲; Q = flow 共m3 / s兲; and R = drag coefficient 共s2 / m5兲. sary for the process to converge.
The coefficients ai that maximize or minimize the upstream The method allows for the possibility of fixing nodes with
3Q

head needed to guarantee minimum pressure in each node for constant discharge that do not take part in the optimization pro-
different assumed discharges can be identified by using nonlinear cess together with the cancellation of hydrants when there is no
integer programming 共Rios 1988兲. This identification can be done irrigation in the plot.
by analyzing all the possible localizations of open hydrants, The maximum and minimum system head curves in the Hc-
IR

which, when taken together, sum the upstream discharge. Qc plane are the result of the possible operating points in the
To find the maximum of the upstream head 共Hc兲, the method irrigation network. The pumping station should be dimensioned
indicated above is used with the following restrictions: for the maximum system head curve to ensure sufficient pressure
• Energetic 共for all lines of the network兲 in the hydrants under all the possible conditions.

2 / JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR

PR
Fig. 3. Representation of the operating plots during OT along with
required discharges Fig. 4. Outline of normal distribution
O OF
Daily Random Generation of Demand Curves
sectors irrigated per day兲, maximum number of open hydrants,
To know the upstream discharge in an on-demand irrigation net- existence of subnetworks, interaction between variables, and so
work at any time, it is possible to carry out a random generation on.
of open hydrants, which is in fact the basis for dimensioning the The standard deviation of the DRDC changes depending on
pumping station. the time interval in which OT is divided. Fig. 5 represents the
CO

The set of open hydrants at a given time will affect the distri- effect of long and short intervals on upstream discharge distribu-
bution of the flow in the network, generating a discharge and an tion. There is a higher possibility of discharge concentration when
upstream head, which constitute the system head curve 共Fig. 2兲. using longer intervals, thus leading to DRDCs that present
The evolution of these curves enables us to estimate the maxi- sharper discharge peaks at certain moments of the day.
PY

mum upstream discharge 共Qmax兲. Clément’s methodology for determining design discharge 共Qd兲
To estimate the evolution of the daily discharge variation in uses the average irrigation time of the plots, which corresponds to
the network, the volume of water that must be distributed accord- the situation of longer intervals.
ing to the crop requirements should be taken as a starting point. Although all system head curves supposedly have the same
Additional data such as hydrant discharge 共qd兲, irrigation time for
[IR

probability, the curves generated can present subsets that are more
each plot, irrigation interval, and operating time of the network probable 共Pulido et al. 1998兲, depending on tariffs 共periods of
共OT兲 should be known. To obtain daily random generation of lower electric energy cost during the day兲 or on irrigation habits
demand curves 共DRDC兲, the operating time of the network, OT, is 共e.g., night or weekend use of sprinkler irrigation, or on the week-
/2

divided into short intervals 共e.g., 15 min.兲. Then a random simu- ends兲.
lation of open hydrants is carried out, with deliverance of the A way to implement this possibility in this process is to settle
same volume 共Vd兲 and average discharge 共Qa = Vd / OT兲 per day
00

the condition for distributing a proportion of the daily volume in


used as conditions, by randomly selecting the starting time for some hours of the day. Then the operation probability will not be
each one of the hydrants, in a period of time between the initial the same for all hours of the day. In this case, the obtained system
OT and the OT minus the time necessary to irrigate each plot. The head curves will not fit a normal distribution, which requires the
3/

required upstream discharge is obtained, almost continuously, by use of other probability density functions, such as the Weibull
adding the flows generated in each interval 共⌺qd兲, as represented distribution 共Mavropoulos 1996兲.
02

in 共Fig. 3兲. Since the location of the irrigated plots is known in


each moment, the upstream head can be determined.
For each day, the same amount of upstream discharge data will Application to Network Analysis
25

be observed as intervals in OT. Using the same procedure to In the simulation of the system head curves, for each OT interval,
generate numerous DRDCs, a large database of upstream dis- it can be obtained from the flows in the pipes of the network.
charges that supposedly fit on a normal distribution will be ob- From them, the maximum flows can be determined in as many
tained 共Clément and Galand 1979兲. With these data, Clément’s
66

situations as the number of generated DRDCs. These flows can be


methodology can be used to calculate the design of upstream used for the network analysis by using software like EPANET
discharge with a preestablished operating quality 共EPANET 1997兲 and obtaining the pressure evolution in the hy-
]0

drants of the network once all the pipes have been dimensioned.
Qd = ␮ + U␴, 共8兲
It allows the verification of correct selection of pipe diameters
where U = percent of the normal distribution function associated and pressure control valves, as well as the identification of areas
with a definite operating quality, 共OQ兲 共Table 1兲; ␮ = mean of with excess or deficit pressure in the network.
06

normal density function; and ␴ = standard deviation. See Fig. 4.


In this way, Qd represents the upper limit in the confidence
Reliability of Pumping Station
interval. The confidence coefficient is determined by the OQ pa-
50

rameter. The U values are the percent of the normal distribution The study concerning reliability of the pumping station 共PS兲 is
function for a given OQ. usually carried out by choosing a set of open hydrants at random.
To generate the DRDC, several variables can be considered: This procedure leads to an upstream discharge 共Qc兲 and an asso-
3Q

crop, average application rate, number of sectors, automation ciated head to guarantee minimum pressure in the hydrants. Once
grade 共because of its influence on irrigation time and number of the pumping station has been dimensioned, the Qc and Hc require-
IR

Table 1. U-Values in Function of Operation Quality 共OQ兲


CF 90.0 91.0 92.0 93.0 94.0 95.0 96.0 97.0 98.0 99.0 99.5
U 1.28 1.34 1.41 1.48 1.56 1.65 1.75 1.88 2.05 2.33 2.58

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005 / 3

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR
Pump Selection
The type and number of variable- and fixed-speed pumps to in-
stall in a pumping station must be the combination that minimizes
the total annual costs, which are the sum of project costs plus
energy costs. A two-phase procedure is carried out:
PR
Phase One
1. Choose several solutions that satisfy the Q-H requirements
O
Fig. 5. Effect of interval size on upstream discharge design of the network using different commercial pumps.
2. Determine the combination of the number and type of
OF
variable- and fixed-speed pumps for maximum efficiency
during the working interval. To do so, SOLVER software
ments can be higher than the pumping station can provide, 共included in EXCEL兲 can be used. The procedure consists of
thereby producing a supply failure in the system. the following:
Fig. 1 shows a common dimensioning for a pumping station a. Select a set of Q-H values representative of the work-
CO
using a Q1 equivalent to Clément’s design discharge. The high ing interval in which the pumping station can func-
proportion of possible failures which can result when designing a tion 共e.g., 15–20兲. It is assumed that these values have
pumping station of reduced dimensions can be observed. Thus, a
the same probability. With additional available infor-
Qc lower than Q1 does not lead to failures. A Qc between Q1 and
mation for more probable values, certain Q values
PY

Q2 can cause failures, depending on the required Hc. For a Qc


can be considered. With these data, a proportional
higher than Q2, there will always be failures in the network. The
number of supply failures to the network shows the reliability of average can be calculated.
the pumping station dimensioned for that network. b. Using the characteristic curves of each type of pump
The reliability of the pumping station can also be studied by calculate the efficiency of each Q-H value, with using
[IR

using the DRDCs generated for each one of the OT intervals. For the necessary set of variable- and fixed-speed pumps.
a determined Qc, the percentage of DRDCs that do not exceed In addition to pump efficiency, engine efficiency has
this discharge can be found. In this way, the failure distribution also been considered 共0.90–0.93 depending on the
function associated with Qc can be determined. power, which is between 20 and 150 kW兲. For
/2

Fig. 6 emphasizes the three most interesting points that usually variable-speed pumps, a frequency converter effi-
appear in the numerous on-demand sprinkler irrigation networks ciency of 0.947–0.96 has also been considered.
00

studied. Point a corresponds to the upstream discharge obtained c. When using SOLVER, the process for selecting the
by applying Clément’s methodology 共Clément and Galand 1979兲. number of variable- and fixed-speed pumps that
Point b corresponds to the maximum flow obtained from the maximize pumping station efficiency is as follows:
3/

DRDC generated, and point c corresponds to the total discharge 共1兲 Use the number of variable- and fixed-speed
when all the hydrants are opened. pumps, as well as the relative number of revolu-
02

To maximize reliability in new irrigation networks, the pump- tions of variable-speed pumps, as changing vari-
ing station should be dimensioned with the maximum discharge ables in the optimization process.
obtained from the DRDC generated, along with the head corre- 共2兲 Introduce the necessary restrictions in coherence.
sponding to the Hmax. 共3兲 Set out the objective function as maximum aver-
25

age pumping station efficiency 共or minimum


Estimation of Daily Energy Consumption consumption of average power兲 for the set of
66

Q-H values where it can work. This efficiency is


The net energy consumed by the network for each one of the the arithmetic average 共or the proportional aver-
DRDCs during the OT can be calculated, since the upstream dis- age, in the case of more probable discharges兲 of
charge and the associated head are known.
]0

the efficiencies matching the set of Q-H values


By summing the energies consumed in each OT interval, the
representative of the working interval.
daily net energy can be determined. If these energy data fit a
normal distribution, as will happen with the discharges, both the
Phase Two
06

average and the standard deviation of the distribution can be de-


termined. To calculate the total annual costs for each type of pump, the
procedure consists of the following:
1. Calculating the daily net energy consumption from the
50

DRDC, while deducing the annual net energy consumption


according to the requirements of the crops during the irriga-
tion season;
3Q

2. Calculating the annual gross energy consumption by dividing


net energy by optimal average efficiency; and
3. Selecting the number and type of variable- and fixed-speed
IR

pumps that produce the lowest total annual cost.


In the application, the project’s annual cost has been calculated
by multiplying the initial project cost by the capital recovery fac-
Fig. 6. Discharge cumulative frequency tor 共CRF兲

4 / JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR

PR
O OF

Fig. 7. Enumeration of lines and nodes in on-demand irrigation net-


Fig. 8. Probability density functions
work
CO

r共1 + r兲n If the pumping station is composed of n P equal pumps ar-


CRF = = 0.13 共9兲
共1 + r兲n − 1 ranged in parallel, where nVP are variable-speed pumps with
shared regulation and nFP are fixed-speed pumps, the power ab-
PY

where n = useful life of the project 共10 years兲; and r = interest rate
sorbed by the pumping station 共Pabs in kW兲 is
considered 共5%兲. Operating cost is determined for an energy cost
of 0.064 Euros/ kWh.
0.00981QVPH 0.00981QFPH
Pabs = + 共12兲
G F G F
QBV + 2 2 Q2VP QFP + 2 Q2FP
[IR

Regulation of Pumping Stations


nVP␣ nVP␣ nFP nFP
In a pumping station composed of nVP equal variable-speed
pumps and shared regulation, which means that all the pumps where QVP = total discharge of variable-speed pumps; and QFP
spin at the same speed 共García-Serra and Madalena 1992兲, the = total discharge of fixed-speed pumps.
/2

head provided 共H, in m兲 and the pump efficiency 共␩兲, in terms of Consequently, the total discharge of the pumping station is
discharge 共Q, in L/s兲, are QC = QVP + QFP.
00

When the pumping station is dimensioned, its regulation by


E 2 means of a programmable proportional, integral, and derivative
H = C␣2 − Q 共10兲
n2VP 共PID兲 automaton, as described in Creurs 1981, can be carried out
3/

in multiple ways:
G F 1. With only one pressure transducer upstream. A minimum
␩= Q + 2 2 Q2 共11兲 pressure is established, while the automaton regulates the
nVP␣
02

nVP␣
operation of the different variable-speed pumps in an attempt
where C , E , F , G = pump coefficients, obtained by regression to maintain this pressure. This regulation does not fit prop-
analysis based on characteristic curves of commercial pumps; ␣ erly with the requirement of the system head curve in the
25

= pump relative revolution number 共␣ = N P / No兲 at a given time; network, leading to pressure excesses for low discharges and
No = pump nominal revolution number; and N P = number of pump insufficient pressure for higher discharges.
revolutions at a given time. 2. With information on the discharge and pressure. In this case,
66

Table 2. Areas and Discharges Designated for Nodes


]0

Node 1 2 3 4 5 6 7
Area 共ha兲 0.00 5.04 0.00 8.64 5.91 0.00 4.38
Discharge 共L/s兲 0.00 16.80 0.00 28.80 19.70 0.00 14.60
06

Node 8 9 10 11 12 13 14
Area 共ha兲 0.00 8.37 5.16 3.84 3.93 2.55 0.00
Discharge 共L/s兲 0.00 27.90 17.20 12.80 13.10 8.50 0.00
50

Node 15 16 17 18 19 20 21
Area 共ha兲 0.00 4.92 2.58 5.49 8.07 5.22 0.00
Discharge 共L/s兲
3Q

0.00 16.40 8.60 18.30 26.90 17.40 0.00


Node 22 23 24 25 26 27 28
Area 共ha兲 7.02 5.82 4.98 5.40 5.22 4.47 0.00
Discharge 共L/s兲 23.40 19.40 16.60 18.00 17.40 14.90 0.00
IR

Node 29 30 31 32 33 34
Area 共ha兲 1.89 0.33 0.00 5.70 5.76 7.02
Discharge 共L/s兲 6.30 1.10 0.00 19.00 19.20 23.40

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005 / 5

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR

PR
O OF

Fig. 9. Example of daily random demand curve during July and


evolution of maximum discharges, as well as daily distributed water Fig. 11. Maximum and minimum system head curves
volume
CO

several possibilities exist, although only two are considered irrigation time and OT have been reduced in an attempt to main-
here: tain similar average discharges for all the months.
a. Following the maximum system head curve. This op- For July, the requirements are 4.2 L m−2 day−1, which entails
distributing a water volume of 5,360 m3 / day. OT has been set for
PY

tion has the advantage of guaranteeing minimum


pressure under all operating conditions of the net- 18 h, with an average application rate of 1.2 L m−2 h−1 for the
work, and it has the disadvantage of generating pres- system 共6 emitters⫻ 4 L / s, for the area corresponding to each
sure excesses in some hydrants. tree兲, thus obtaining an average discharge of 82.7 L / s.
b. Following a dynamic system head curve with at least By dividing OT into 15-min intervals and generating 4,000
[IR

one hydrant free from excess pressure. For this op- DRDCs, the average flow of these curves is 82.7 L / s with a stan-
tion, a certain number of pressure sensors three to dard deviation of 45.8 L / s 共Curve 2 in Fig. 8兲, hence obtaining a
five, depending on the network兲 located in the most design upstream discharge of 189.2 L / s. The irrigation time
should be a multiple of the OT intervals to ensure the same vol-
/2

unfavoruable nodes are needed. These nodes should


be previously identified, for example, with EPANET ume for each day. As can be observed in Fig. 9, the evolution of
software. The upstream pressure will be adjusted until these DRDCs gives a maximum discharge of 256 L / s. A repre-
00

there is no excess pressure in the one of the sensors. sentation of one of these curves has also been indicated, as well as
the daily evolution of the discharge and the distributed water vol-
ume. When all the hydrants are opened, the total discharge in the
3/

network is 425.7 L / s.
Results
If OT is divided into 2-h intervals, a standard deviation of
42.67 L / s and an upstream discharge of 181.7 L / s are obtained
02

The methodology has been applied to an on-demand irrigation


for the same average discharge 共82.7 L / s兲. With intervals of
network that irrigates 127.7 ha of citrus fruits in Valencia, Spain,
3.5 h, which is the average irrigation time in the zone, the stan-
with a spacing between trees of 5 m ⫻ 4 m. In the agronomic
dard deviation is 35.4 L / s and the upstream discharge is
25

design, it is necessary to have at least six emitters/plant, with a


165.1 L / s, as shown in Curve 1 in Fig. 8 for the same average
nominal discharge of 4 L / s and a pressure of 10 m. These con-
discharge 共82.7 L / s兲.
ditions originate with minimum pressure of 25 m in the hydrants
This analysis can also be performed with the maximum dis-
共Fig. 7兲.
66

charge of each one of the DRDCs generated. When OT is divided


Table 2, indicaties the areas and discharges of the hydrants.
into intervals of 15 min, the average discharge is 172.5 L / s,
Discharges have been obtained for an irrigation time of 3.5 h in
thereby obtaining a standard deviation of 19.2 L / s and an up-
the most unfavorable month 共July兲. In the remaining months, the
]0

stream discharge of 218 L / s, as shown in Curve 3 in Fig. 8.


06
50
3Q
IR

Fig. 10. Discharge distribution function Fig. 12. Characteristic curve Q-H for the set of pumps selected

6 / JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR
Table 3. Required Number and Power for Each Type of Pump
Pump type B C D III I IV II
Number 6 5 5 4 2 3 3
Power (kW) 27 32 40 48 112 101 126
PR
For selecting the pumps, it was necessary to take into account
their power and characteristic curve. Type A is the least powerful;
whereas Types B, C, and D are similar to A but more powerful.
O
The Type III pump is similar to C, but has a flatter Q-H curve.
Pump II is more powerful, and Type I is intermediate but has the
OF
flattest Q-H curve. As for efficiency curves, three groups can be
distinguished. The first group, with similar efficiency curves, is
Fig. 13. Characteristic curve Q-␩ for the set of pumps selected composed of Type A, B, C, and D pumps. A second group is
composed of Types III and IV. The third group includes the pump
Type I, with a very open Q-␩ curve; and type II, which is inter-
CO

The reliability of the pumping station is studied by means of mediate.


the DRDCs generated. The failure distribution function, which To determine the best solution for a minimum total cost, we
contains the percentage of DRDCs with a maximum discharge applied the methodology described. In the interval of discharges 共
higher than the design discharge, Qd, is also calculated. Fig. 10 共0 – 218 L / s兲, 17 uniformly distributed values have been selected.
PY

shows the distribution function obtained where OT is divided into SOLVER has been used for identifying the variable- and fixed-
15-min intervals. The maximum discharge 共256 L / s兲, design dis- pump combination that maximizes efficiency for the 17 dis-
charge 共218 L / s兲, and Clément’s discharge 共166 L / s兲 are high- charges. Table 3 shows the total required number of each type of
lighted. pump, plus the power consumed for the maximum efficiency
The maximum 共Hmax兲 and minimum 共Hmin兲 system head point.
[IR

curves obtained by using this methodology are shown in Fig. 11. To calculate energy consumption during the irrigation season,
In this case, the maximum system head curve adjusts to the the network is assumed to cover the daily irrigation requirements
straight line H = 25.84+ 0.081Q. and the daily irrigation volume, as indicated in Table 4. In this
table the daily, monthly, and annual net energy consumed are also
/2

considered. The daily average energy has been calculated as the


Pump Selection
average of the energy consumption distribution of the set of gen-
00

For dimensioning the pumping station, an upstream discharge of erated DRDCs. To simplify this calculation, the maximum head
218 L / s has been considered. This design discharge is obtained has been assigned to every upstream discharge, instead of being
with the distribution of maximum discharges of the generated determined according to the distribution of open hydrants for each
3/

DRDCs, dividing OT into intervals of 15 min. With respect to one of the 4,000 DRDC generated.
regulation, only the maximum head curve has been considered. Fig. 14 shows the average and standard deviation of the
02

To satisfy the demand conditions of 218 L / s at 44 m, eight monthly energy consumption distribution.
types of pumps 共A, B, C, D, I, II, III, and IV兲 have been consid- In Fig. 15, the annual total cost evolution is shown according
ered, in addition to their characteristic curves, Q-H and Q-␩, to the maximum number of variable speed pumps and type of
shown in Figs. 12 and 13. The two types of nomenclature are used pump. To simplify the calculations, all the pumps are considered
25

solely to distinguish pumps from different manufacturers. as equal, thereby creating an interesting situation in which any
66

Table 4. Water Requirements and Annual Energy Consumption


Daily
]0

irrigation Daily average


requirements Daily volume of net energy Days in the Monthly net
Month 共L m−2 day−1兲 共m3 / day兲 media 共kWh兲 month energy 共kWh兲
06

January 0.520 664.092 74.0 31 2,294.1


February 0.300 383.313 35.8 28 1,003.5
March 0.720 919.512 115.4 31 3,576.2
50

April 1.190 1,519.749 145.7 30 4,372.4


May 1.930 2,464.803 262.0 31 8,122.5
June 3.230 4,125.033 405.4 30 12,161.6
3Q

July 4.200 5,360.000 556.8 31 17,259.6


August 3.800 4,852.980 481.8 31 14,935.9
September 2.210 2,822.391 299.1 30 8,971.7
October 0.790 1,008.909 113.0 31 3,502.6
IR

November 0.550 702.405 74.0 30 2,221.1


December 0.540 689.634 74.0 31 2,295.2
Total annual 80,716.0

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005 / 7

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR

PR
O OF

Fig. 14. Average and standard deviation of monthly energy con- Fig. 16. Cost of pumps and frequency converter
sumption distribution
CO
important differences in the annual total cost or in the global
pump can be connected to the frequency converter. This would efficiency of the pumping station.
avoid overworking the one pump that would be continuously con- • Introducing a second pump with variable speed produces a
nected to the frequency converter. The total cost has been calcu- significant increase in efficiency in many situations, compen-
sating for the increase in investment cost with reduced energy
PY

lated as the sum of the cost of pumping, the motor, and the fre-
quency converter, plus the energy cost. This last has been consumption.
calculated by dividing the annual net energy 共Table 4兲 by the • The availability of two variable-speed pumps facilitates adap-
average optimal efficiency for each combination of pumps. tation to smaller discharges, thereby improving the connection
For fixed-speed pumps, we have considered the cost of an to fixed-speed pumps. Furthermore, the probability that the
[IR

electronic starter, necessary for a start-up similar to the one ob- pumping station will not break down is increased, since it can
tained with a frequency converter. The cost of the pumps consid- work perfectly with just one frequency converter if the others
ered and the frequency converter, using average market values, break down.
are shown in Fig. 16.
/2

As can be observed in Fig. 15, the least expensive solution is


normally found by using one or two variable-speed pumps and Conclusions
00

fixed-speed pumps, for the others. The best combination for this
situation is the use of less powerful pumps, even though more An easily-implemented procedure that uses Excel Visual Basic
pumps are necessary to cover the requirements of the pumping has been presented for dimensioning and regulating pumping sta-
3/

station. Nevertheless, this solution has a lower risk in case one of tions with a determined operation quality for the on-demand irri-
the pumps, breaks down, as well as a lower cost for pumps in gation network. This procedure guarantees the lowest total-cost
solution while using equipment available on the market and iden-
02

reserve, which are necessary in any pumping station.


Maximum efficiencies for each type of pump in the pumping tifies the type and number of fixed- and variable-speed pumps
station are shown in Fig. 17 for different numbers of variable- necessary.
speed pumps. The methodology proposed can be adapted to any network,
25

On analyzing Figs. 15 and 17, the following can be high- whatever its specific characteristics and restrictions may be 共fluc-
lighted: tuation in demand, different irrigation system, crops, and so on兲.
• Pump Type A is not introduced because of its lower power, Methods for defining the maximum and minimum system head
66

thereby eliminating it as an acceptable economical solution. curves and DRDC in the network have been developed. These
• Pumps with a flatter Q-H 共Types I and III兲 do not produce methods are the basis for dimensioning the pumping station and
for its regulation and control.
]0
06
50
3Q
IR

Fig. 15. Annual total cost evolution according to maximum number Fig. 17. Global efficiencies of pumping station for different numbers
of variable-speed pumps and type of pump of variable-speed pumps for each type of pump studied

8 / JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005

PROOF COPY [IR/2003/022566] 006503QIR


PROOF COPY [IR/2003/022566] 006503QIR
In the network studied, the best solution for dimensioning and Mecánica de Fluidos, Servicio de Publicaciones de la Universidad
regulating the pumping station is with one or two variable-speed Politécnica de Valencia, Valencia, Spain 共in Spanish兲.
pumps, with the remaining pumps being of fixed speed. The avail- Lamaddalena, N., and Sagardoy, J. A. 共2000兲. “Performance analysis of
ability of two variable-speed pumps can also improve the regula- on-demand pressurized irrigation systems.” Irrigation and Drainage
tion and the probability that the pumping station will not break Paper 59, Food and Agricultural Organization of the United Nations
PR
down. 共FAO兲.
López-Luque, R., Reca, J., Camacho, E., Roldán, J., and Alcaide, M.
共1995兲. “Valoración energética y diseño óptimo de impulsiones con
caudal variable.” Ingeniería del Agua, 2, 19–24 共in Spanish兲.
O
References Martínez, F., and Vela, A. 共1993兲. “Regulación de estaciones de bombeo.
Curso de diseño hidráulico de redes de riego.” U.D. de Mecánica de
Brion, L. D., and Mays, L. W. 共1991兲. “Methodology for optimal opera-
OF
Fluidos, Servicio de Publicaciones de la Universidad Politécnica de
tion of pumping stations in water distribution systems.” J. Hydraul. Valencia, Valencia, Spain 共in Spanish兲.
Eng., 117共11兲, 1551–1569. Mavropoulos, T. I. 1996. “Sviluppo di una nuova formula per il calcolo
Clément, R., and Galand, A. 共1979兲. “Irrigation par aspersión et réseaux delle portate di punta nelle reti irrigue con esercizio alladomanda.”
collectifs de distribution sous pression.” Editions Eyrolles, Paris 共in Riv. di Irr. E Dren, 44共2兲: 27–35 共in Italian兲.
CO

French兲. Ormsbee, L. E., and Lansey, K. E. 共1994兲. “Optimal control of water-


Creurs, A. 共1981兲. Instrumentación industrial, Marconabo, Barcelona 共in supply pumping systems.” J. Water Resour. Plan. Manage., 120共2兲,
Spanish兲. 237–252.
Duan, N., Mays, L. W., and Lansey, K. E. 共1990兲. “Optimal reliability- Pulido, I., López, R., and Roldán, J. 共1998兲. “Caracterización horaria y
PY

based design of pumping and distribution systems.” J. Hydraul. Eng., estacional de la demanda en una red de distribución de agua para
116共2兲, 249–286. riego.” XVI Congreso Nacional de Riegos. AERYD, Murcia 共in Span-
EPANET. 共1997兲. Manual de Usuario de EPANET. Risk Reduction En- ish兲.
gineering Laboratory, Office of Research and Development, U.S. En- Pulido-Calvo, I., Roldan, J., López-Luque, R., and Gutiérrez-Estrada, J.
vironmental Protection Agency, Cincinnati, Ohio. C. 共2003兲. “Water delivery system planning considering irrigation si-
[IR

García-Serra, J., and Madalena, M. 共1992兲. “La inyección directa a red multaneity.” J. Irrig. Drain. Eng., 129共2兲, 247–255.
con grupos de velocidad fija y variable.” Curso de ingeniería hidráu- Rios, S. 共1988兲. Investigación operativa. Optimización, Centro de Estu-
lica aplicada a los sistemas de distribución de agua, U.D. de dios Ramón Areces S. A., Madrid 共in Spanish兲.
/2
00
3/
02
25
66
]0
06
50
3Q
IR

JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / MAY/JUNE 2005 / 9

PROOF COPY [IR/2003/022566] 006503QIR

You might also like