processes
Article
The Estimation of Centrifugal Pump Flow Rate Based on the
Power–Speed Curve Interpolation Method
Yuezhong Wu 1 , Denghao Wu 1, * , Minghao Fei 1 , Gang Xiao 2 , Yunqing Gu 1 and Jiegang Mou 1
Citation: Wu, Y.; Wu, D.; Fei, M.;
Xiao, G.; Gu, Y.; Mou, J. The
Estimation of Centrifugal Pump Flow
Rate Based on the Power–Speed
Curve Interpolation Method.
Processes 2022, 10, 2163. https://
doi.org/10.3390/pr10112163
Academic Editors: Rey-Chue Hwang
and Huixin Tian
Received: 12 September 2022
Accepted: 20 October 2022
Published: 22 October 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Processes 2022, 10, 2163. https://doi.org/10.3390/pr10112163
1 College of Metrology and Measurement Engineering, China Jiliang University, Hangzhou 310018, China
2 College of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310014, China
* Correspondence: wdh@cjlu.edu.cn; Tel.: +86-13588395807
Abstract: During the global energy crisis, it is essential to improve the energy efficiency of pumps by
adjusting the pump’s control strategy according to the operational states. However, monitoring the
pump’s operational states with the help of external sensors brings both additional costs and risks
of failure. This study proposed an interpolation method based on PN curves (power–speed curves)
containing information regarding motor shaft power, speed, and flow rate to achieve high accuracy
in predicting the pump’s flow rates without flow sensors. The impact factors on the accuracy of the
estimation method were analyzed. Measurements were performed to validate the feasibility and
robustness of the PN curve interpolation method and compared with the QP and back-propagation
neural network (BPNN) methods. The results indicated that the PN curve interpolation method has
lower errors than the other two prediction models. Moreover, the average absolute errors of the PN
curve interpolation method in the project applications at 47.5 Hz, 42.5 Hz, 37.5 Hz, and 32.5 Hz are
0.1442 m3 /h, 0.2047 m3 /h, 0.2197 m3 /h, and 0.1979 m3 /h. Additionally, the average relative errors
are 2.0816%, 3.2875%, 3.6981%, and 2.9419%. Hence, this method fully meets the needs of centrifugal
pump monitoring and control.
Keywords: centrifugal pump; PN curves interpolation; flow estimation; sensorless
1. Introduction
According to the International Energy Agency, electric motor systems consume approx-
imately 53% of all electricity worldwide, representing 5.5 Gton of CO2 emissions, which
accounts for 70% of the total electricity consumption of industry [1,2]. As per the European
Commission report [3], pumping systems occupy nearly 22% of the energy consumed by
electric motors worldwide and consume approximately 8.7% of global electricity [4]. Hence,
a feasible way to alleviate the energy crisis and reduce carbon emissions is to improve
the energy efficiency of pumps [5]. According to statistics, even a 1% increase in pump
efficiency could reduce CO2 emissions by at least 572 tons per day in the European re-
gion [6]. For this reason, centrifugal pumps, as a vital part of pump systems, consume 72%
of the pump system’s energy and have great potential for energy and carbon savings [7].
Meanwhile, this also helps to extend the life of equipment by improving the pumps’ energy
efficiency [8,9]. A practical method to reduce the energy consumption of centrifugal pumps
is to optimize the control strategy using frequency converters [10], and the optimization pro-
cess typically requires real-time monitoring of the centrifugal pump’s operational state [11].
However, monitoring by external sensors not only requires more installation space but
also brings more risk of centrifugal pump failure [12]. The estimation of motor operating
parameters by modern frequency converters makes it possible to monitor a centrifugal
pump’s operational status without sensors [13].
The current methods for monitoring the flow rate of centrifugal pumps driven by
frequency converters mainly include QP methods [13–15], QP extension methods [16–18],
and data-driven artificial neural network-based estimation methods [19–21].
https://www.mdpi.com/journal/processes
Processes 2022, 10, 2163 2 of 16

The intelligent pump control (IPC) program developed by ABB can calculate the flow
rate based on the performance curves of the QH and QP programmed into the frequency
converter parameters data. Hammo et al. [14] compared the flow rate calculated by the
program with the measured flow rate and concluded that the error mostly comes from
both the accuracy of the QP and QH curves and the position of the operating point on the
system curve. Ahonen et al. [15] developed the QP model, which can predict the flow rate
and head of a centrifugal pump based on the measured QP and QH curves. However, the
shape of the QP curve (flat or humped) can severely impact estimation accuracy. Hence, a
method based on centrifugal pump process characteristics was proposed [16]. Since the
centrifugal pump operates on the intersection of the process characteristic curve and the
QH curve, the flow rate and head can be predicted by the affinity law along with these
two curves.
While the QP and process characteristics methods have been validated in the labo-
ratory and extended to engineering applications, there is still significant opportunity to
improve the estimation accuracy of the flow rate. Therefore, Tamminen et al. [17] proposed
a hybrid QH/QP estimation method. The hybrid QH/QP model showed approximately
5% flow rate error compared to the standard QP model, which showed approximately
10% standard deviation. By identifying the parameters of the process characteristic curve
using the QP method, Ahonen et al. [18] established an improved model for predicting the
operating status of centrifugal pumps by the process characteristic method.
With the development of data-driven technologies [19], the neural network has become
a new solution for estimating centrifugal pump performance due to its good nonlinear
fitting ability and capability to solve multi-input problems [20]. A dual neural network pre-
diction method based on motor speed and power was proposed for predicting centrifugal
pump flow and head [21]. Han et al. [22] developed a dual hidden layer BP neural network
to predict the head and efficiency of centrifugal pumps using the Levenberg–Marquardt
(LM) training algorithm. Based on measurements of air temperature and solar radiation
collected every hour, the model can precisely estimate the hourly flow rate of photovoltaic
water pumps [23]. Moreover, an improved neural network prediction model based on
Bayesian regularized back propagation (BRBP) was proposed, and it is more accurate than
the QP estimation model [24].
Most existing studies are concerned with estimating the operating conditions of
centrifugal pumps. However, there misses a highly accurate model for predicting the
small flow rates of centrifugal pumps [25]. This paper proposed an interpolation method
based on PN curves to solve the above problem. Meanwhile, the factors affecting the
accuracy of the estimation method were analyzed. Moreover, we compared the PN curve
interpolation method with the QP and BPNN methods, and its feasibility was verified in
practical engineering applications.

2. Experimental Setup
We measured the test data in a pump closed-loop test facility. It consists of an inlet tank
(Yongli, Beichen, Tianjin, China, 2019), an outlet tank (Yongli, Beichen, Tianjin, China, 2019),
electric valves (Dunbang, Songjiang, Shanghai, China, 2019), electromagnetic flow meters
(Endress + Hauser, Minhang, Shanghai, China, 2019), pressure sensors (Endress + Hauser,
Minhang, Shanghai, China, 2019), an electrical control cabinet (Ganlang, Wenzhou, Zhe-
jiang, China, 2019), and a computer (Hasee, Shenzhen, Guangzhou, China, 2021) (see
Figure 1). Figure 2 presents the configuration of the test equipment, and Table 1 shows the
technical parameters of the measurement instruments.
 lin, Brandenburg, Germany, 2020). The frequency converter communicates with the PLC
through the
through the Modbus
Modbus protocol.
protocol. The
The software
software was
was written
written in
in LabVIEW
LabVIEW (Version
(Version 18.0,
18.0, Na-
Na-
tional Instruments, Austin, Texas, United States, 2021), and the PLC communicates
tional Instruments, Austin, Texas, United States, 2021), and the PLC communicates with with
the computer
the computer through
through Ethernet.
Ethernet.
To reduce
To reduce measurement
measurement errors,
errors, we
we repeated
repeated the
the test
test five
five times
times and
and recorded
recorded the
the
Processes 2022, 10, 2163 measured data. Furthermore, we used the average of five experiments as the
measured data. Furthermore, we used the average of five experiments as the final data tofinal data
3 ofto
16
enhance the reliability of the data.
enhance the reliability of the data.

Inlettank
Inlet tank
Outlettank
Outlet tank

Magnetic flow
Magnetic flowtransducers
transducers

Test object
valves Test object
Motorizedvalves
Motorized

Flowdirection
Flow direction

AC380V
Ground AC380V
Ground
PLC Frequencyconverter
Frequency converter
PLC Speedtransducer
transducer
Speed

Figure
Figure 1.1.A
Figure1. AAclosed-loop
closed-loop pump
closed-looppump test
pumptest system.
testsystem.
system.

Figure
Figure2.
Figure 2.2.Experimental
Experimental
Experimentalfacilities: (a)closed-loop
facilities:
facilities: (a) closed-loop testbed,
(a) closed-loop
test bed, (b)
test(b) vertical
bed, centrifugal
(b) vertical
vertical pump,and
centrifugal
centrifugal pump, and (c)con-
pump,
(c) con-
and
trol
trol system.system.
(c) system.
control

Table
Table 1.
Table1. Test
1.Test sensor
Testsensor information.
sensorinformation.
information.
Sensor
Sensor
Sensor
Type
Type
Type
Measurement
Measurement
Measurement
Range
Range Precision
Range Precision
Precision
Manufacturers
Manufacturers
Manufacturers
Electromagnetic
Electromagnetic
Electromagnetic EndressSwitzerland
+ Hauser,
0–150
0–150 mm3m
0–150
3/h
/h3 /h 0.2% 0.2%Endress
0.2% Endress ++ Hauser,
Hauser, Switzerland
flowmeter
flowmeter
flowmeter Switzerland
Power
PowerPower meter
meter
meter Current,
Current, voltage,
voltage,
Current, voltage, power
power
power 0.5% 0.5% Qingzhi
0.5% QingzhiQingzhi
instrument,
instrument, China
China
instrument, China
Tachometer
Tachometer 0–20,000
0–20,000 rpmrpm ±1 rpm±1 rpm Onosokki,Onosokki,
Japan Japan
Tachometer 0–20,000 rpm ±1 rpm Onosokki, Japan

3. Estimation
3. Estimation Methods
The testMethods
object is a multistage centrifugal pump (Nanyuan, Hangzhou, Zhejiang,
China, 2019) with a 5.5 kW induction motor (Nanyuan, Hangzhou, Zhejiang, China, 2019).
The nominal operating point is Qnom = 15 m3 /h, Hnom = 70.5 m, and nnom = 2900 rpm. The
frequency converter (Siemens, Berlin, Brandenburg, Germany, 2020) adjusts the speed
of the pump, and a motorized valve controls the flow rates of the pump. The controller
is a Siemens S7-200smart ST30 (Siemens, Berlin, Brandenburg, Germany, 2020), and the
analog signals are collected through the expansion module Siemens EM AM06 (Siemens,
Berlin, Brandenburg, Germany, 2020). The frequency converter communicates with the
PLC through the Modbus protocol. The software was written in LabVIEW (Version 18.0,
National Instruments, Austin, Texas, United States, 2021), and the PLC communicates with
the computer through Ethernet.
Processes 2022, 10, 2163 4 of 16

Processes 2022, 10, x FOR PEER REVIEW 4 of 16

To reduce measurement errors, we repeated the test five times and recorded the
measured data. Furthermore, we used the average of five experiments as the final data to
3.1. QP Method
enhance the reliability of the data.
The QP and QH curves, typically determined at the nominal speed of the pump, are
transformed
3. Estimationinto the current rotational speed by the affinity law shown in Equations (1)–
Methods
(3).
3.1. QP Method
The QP and QH curves, typically determined
Q N at the nominal speed of the pump, are
transformed into the current rotational speed = the affinity
by , (1)
law shown in Equations (1)–(3).
Q N nom nom

Q N
= ,2 (1)
Qnom
H  Nnom
N 
=  , (2)
H nom  N nom 
H N 2
= , (2)
Hnom Nnom
3
P  N  
P =  N  3, (3)
P= N nom  , (3)
Pnomnom Nnom
where N is the actual speed, rpm; nnom nomisisthe
therated
ratedspeed,
speed,rpm;rpm;Q Q isis the
the actual
actual flow
flow rate,
rate,
m33/h; Q is the rated flow rate, m 3
3/h; H is the actual head, m; H
/h; Qnom is the rated flow rate, m /h; H is the actual head, m; Hnom is the rated head,
nom nom is the rated head, m;m;P
is thethe
P is actual
actual power,
power, kW; kW; Pnom is the
Pnom rated
is the rated power,
power, kW.kW.
As
As shown
shown in in Figure
Figure 3, in the current QP and QH curves curves transformed
transformed by by the
the affinity
affinity
law, the
the flow
flowraterateQQ canbebe
estestcan calculated
calculated using
using thethe
motormotor
shaftshaft
powerpower
Pact , P act, then
and and then the
the head
head
Hest canHestbecan be computed
computed by the by the flow
above above flow
rate Qestrate
. In Q est.QP
the In method,
the QP method, the transfor-
the transformation of
the QP curve
mation of the QPbased on the
curve basedmotor speed
on the motor andspeed
the calculation of the flow
and the calculation of rate usingrate
the flow theusing
shaft
power
the are
shaft two of
power arethe
two critical
of thesteps.
criticalHence,
steps. the speed
Hence, theand shaft
speed andpower of the centrifugal
shaft power of the cen-
pump are
trifugal vitalare
pump variables in the flow
vital variables rateflow
in the prediction process. process.
rate prediction

(a) (b)

Pnom=f(Qnom ) Hnom=f(Qnom )

Affinity law
Affinity law
P (kW)

H (m)

Pact
Pact=f(Qact ) Hest
Hact=f(Qact )
Qest Qest

Q (m3/h) Q (m3/h)

Figure 3.
Figure QP characteristic
3. QP characteristic curve
curve method:
method: (a)
(a) QP
QP curves
curves and
and (b)
(b) QH
QH curves.
curves.

3.2. PN Curves Interpolation Method

3.2. PN Curves Interpolation Method
As seen in Figure 4a, it is evident that changing the valve opening has a more signifi-
cant As seen
effect oninthe
Figure
shaft 4a, it is of
power evident
systemsthatwith
changing the valve
high input opening
frequencies hasthose
than a more signif-
with low
icant effect on the shaft power of systems with high input frequencies than
input frequencies. As the input frequency is 25 Hz, the variation of shaft power is already those with low
input frequencies.
very tiny. Figure 4bAsindicates
the inputthefrequency
flow rateis 25 Hz, thewith
variation variation
valveofopening.
shaft power is already
Changing the
very tiny. Figure 4b indicates the flow rate variation with valve opening.
valve opening has a more significant effect on the flow rate for systems with a high input Changing the
valve opening
frequency thanhas a more
those withsignificant
a low input effect on the flow
frequency. rate for when
Meanwhile, systemsthewith
valvea high
openinginput is
frequency than those
small, changing with aopening
the valve low inputhasfrequency. Meanwhile,
a more significant when
impact onthe
thevalve
flow opening
rate of the is
small, changing
centrifugal pumpthe valve opening
compared hasvalve
to a large a more significant
opening. When impact on the
the input flow rate
frequency of the
is 50 Hz,
centrifugal pumpbycompared
the speed varies to a large when
40 rpm. However, valve the
opening.
input When the input
frequency frequency
is 25 Hz, the motor is 50 Hz,
speed
the speed
varies by varies
merelyby 3 40 rpm.
rpm, However,
which when thethat
demonstrates input
thefrequency
motor speed is 25affected
Hz, the motor
by the speed
valve
varies by merely 3 rpm, which demonstrates that the motor speed
opening is consistent with the motor shaft power and the pump flow rate (see Figureaffected by the valve
5).
opening is consistent with the motor shaft power and the pump flow rate (see Figure 5).
Processes 2022,10,
Processes2022, xxFOR
FORPEER
10,2163 PEERREVIEW
REVIEW 555of
of 16
16

(a)
(a) (b)
(b)
77
50
50Hz
Hz 45
45Hz
Hz 40
40Hz
Hz 50
50Hz
Hz 45
45Hz
Hz 40
40Hz
Hz
35
35Hz
Hz 30
30Hz
Hz 25
25Hz
Hz 35
35Hz
Hz 30
30Hz
Hz 25
25Hz
Hz
66 20
20

55
15
15

/h)
(m33/h)
(kW)
PP (kW) 44

Q (m
Q
33 10
10

22
55
11

00 00
00 33 66 99 12
12 15
15 00 33 66 99 12
12 15
15
Valve
Valveopening
opening(%)
(%) Valve
Valveopening
opening(%)
(%)

Figure
Figure4.4.Shaft
Shaftpower
Shaft powerand
andflow
flowvariation
variationwith
withvalve
valveopening:
opening:(a)
(a)shaft
shaftpower
powerand
and(b)
(b)flow
flowrate.
rate.

00 55 10
10 15
15 00 55 10
10 15
15
2982
2982 50
50Hz
Hz 2079.0
2079.0 35
35Hz
Hz
(rpm)

(rpm)
2076.9
N (rpm)

N (rpm)
2968
2968 2076.9

2074.8
2074.8
N

N
2954
2954

2072.7
2072.7
2940
2940
2675.4
2675.4 1780.8
1780.8
45
45Hz
Hz 30
30Hz
Hz
2671.5
2671.5 1779.2
1779.2
(rpm)

(rpm)
N (rpm)

N (rpm)

2667.6
2667.6 1777.6
1777.6
N

2663.7
2663.7 1776.0
1776.0

2375.9
2375.9
40
40Hz
Hz 25
25Hz
Hz
1480.68
1480.68
2373.6
2373.6
(rpm)

(rpm)
N (rpm)

N (rpm)

1479.87
1479.87
2371.3
2371.3
N

1479.06
1479.06
2369.0
2369.0

1478.25
1478.25
00 55 10
10 15
15 00 55 10
10 15
15
Valve
Valveopening
opening(%)
(%) Valve
Valveopening
opening(%)
(%)

Figure
Figure5.5.Speed
Speedvaries
Speed varieswith
withvalve
valveopening.
opening.

Weknow
We knowthe
know the
theQPQP characteristic
QP characteristic
characteristic curve
curveonlyonly
curve reflects
only the shaft
reflects
reflects the power
the shaft and flow
shaft power
power and
andrate flowchange
flow rate
rate
and
change
changeindirectly
and reflects the
and indirectly
indirectly motor
reflects
reflects thespeed
the motor
motor change.
speed The
speed change.
change.affinity
Thelaw
The affinityusually
affinity lawassumes
law usually
usually assumesthat the
assumes
pump’s
that theefficiency
that the pump’s remains constant
pump’s efficiency
efficiency remainswhen
remains transforming
constant
constant when the QP curve,
when transforming
transforming thewhich
the QP
QP curve,is notwhich
curve, affected
which isis
by
not changes
notaffected
affectedby in motor
bychanges
changesinspeed [15].
inmotor
motorspeedHowever,
speed[15]. it cannot
[15].However, ignore
However,ititcannot the
cannotignoreeffect
ignoretheon pump
theeffect
effecton efficiency
onpump
pump
and shaft and
efficiency
efficiency power
andshaft when
shaft powerthewhen
power motorthe
when speed
the motor
motor changes
speed on
speedchanges a large
changes onscale.
on aalarge
large This
scale.is This
scale. aThis
critical factor
isisaacritical
critical
that
factor
factoraffects
that the prediction
thataffects
affects the accuracy
theprediction
prediction of the QP
accuracy
accuracy of
ofthe method.
the QP
QPmethod.
method.
Therefore,
Therefore,
Therefore,thisthis paper
thispaper proposed
paperproposed a
proposedaacurve curve based
curvebasedbasedon on motor
onmotor speed
motorspeed
speedand and shaft
andshaft power,
shaftpower,
power,which which
which
can
can visually reflect the variation of motor speed, power, and
can visually reflect the variation of motor speed, power, and flow rate. After setting the
visually reflect the variation of motor speed, power, and flow
flow rate.
rate. After
After setting
setting the
the
input
input frequency,
inputfrequency,
frequency,we we obtained
weobtained
obtainedthethe speed
thespeed and
speedand power
andpower data
powerdata of
dataof different
ofdifferent flow
differentflow rates
flowrates by
ratesby changing
bychang-
chang-
the
ingvalve
ing the opening,
the valve
valve opening,and and
opening, then those
and then data data
then those
those were fitted
data were to a PN
were fitted
fitted tocurve
to aa PN for
PN curve thefor
curve same
for theflow
the same
same rate
flowat
flow
different
rate
rateat input
atdifferent frequencies.
different input Figure
inputfrequencies. 6a
frequencies. Figure shows
Figure6a that
6ashows at a
showsthat specific
thatat flow
ataaspecific rate,
specificflow the power
flowrate,rate,the increases
thepower
power
as the pump speed increases; at a specific
increases
increasesas asthe
the pump
pump speed
speedincreases;
increases; ataspeed,
at aspecific
specificthespeed,
flow rate
speed, theincreases
the flow
flowrateratewith the power;
increases
increases with
withthe at
the
a specific
power; power, the flow rate decreases with speed.
power;at ataaspecific
specificpower,
power,the theflow
flowrate
ratedecreases
decreaseswith withspeed.
speed.
As
As shown
shownin in Figure
Figure 6b,6b, the
the Q Q00and
and Q Qnn PN
PN curves
curves divided
divided the the graph
graph into into three
three parts.
parts.
Q0
Q0 isis the
the PN
PNcurve
curve with
with the
the lowest
lowest flow
flow rate,
rate, generally
generally takentaken as as 00 m
m33/h.
/h. Therefore,
Therefore, when when
Processes 2022, 10, x FOR PEER REVIEW 6 of 16

Processes 2022, 10, 2163 the centrifugal pump works below the Q0 curve, namely the S1 region, the estimation6 flow
of 16
rate is 0 m /h.
3

(a) (b)

0 m3/h 2 m3/h 4 m3/h 4

4 2.8 S3
6 m3/h 8 m3/h 10 m3/h
2.4 Pt
3 3 t
2.0 Nt S2

P (kW)
P (kW) 2
2500 2600 t
2
QK: P=f2(N)

1 1 S1
QM Q0
QJ: P=f1(N) QJ QK
0 0
1500 1800 2100 2400 2700 3000 1500 1800 2100 2400 2700 3000

N (rpm) N (rpm)

Figure 6.6. PN
Figure PNcurves
curvesand
andPNPNcurves interpolation
curves methods:
interpolation (a) (a)
methods: PNPN
curves for different
curves flowflow
for different ratesrates
and
(b)
andPN
(b)curve-based flow flow
PN curve-based estimation method.
estimation method.

As
Theshown
S2 regionin Figure
contains 6b,athe lot Qof0 PN Qn PN the
andcurves, curvesflowdivided
rate from theQgraph
0, Q1,… into
Qj, three
Qj+1,…parts.
Qk-1,
Q 3
Q0k is the PN
covers curverange
a wide with of thecentrifugal
lowest flow rate, generally
pumps with smalltaken flowas 0 m rated
rates, /h. Therefore,
conditions, when
and
the centrifugal pump works below the Q curve, namely the
high flow rates. If t lies between the Qj and Qj+1 flow curves in the S2 region, the flow rate
0 S1 region, the estimation flow
rate is 0 m 3 /h.
at point t (N t, Pt) can be obtained by the internal interpolation method, which is shown in
2 region contains a lot of PN curves, the flow rate from Q0 , Q1 , . . . Qj , Qj+1 , . . .
The S(4).
Equation
Qk−1 , Qk covers a wide range of centrifugal pumps with small flow rates, rated conditions,
and high flow rates. If t lies Q between the Q Pt −and f j (N ) flow curves in the S2 region, the flow
Qtj+1
est = Q j +  (Q j +1 − Q j ) ,
j
(4)
rate at point t (Nt , Pt ) can be obtained byf j +the 1 ( N t ) −
internal f j Ninterpolation
( t) method, which is shown
in Equation (4).
where fj is the power–speed function at Pt the
− fflow
j ( Nt )rate Qj, and fj+1 is the power–speed func-
tion at the flow rate QQ j+1.est
= Qj + · ( Q j +1 − Q j ), (4)
f j+1 ( Nt ) − f j ( Nt )
The S3 region represents the centrifugal pump working above the Qn curve. Typically,
pumpsfjoperate
where is the power–speed
in the S2 and Sfunction at the flow rate Qj , and f of
1 regions. When the valve opening
is the
j+1the power–speed
system is so large
function at the flow rate Q j+1 .
that the pump is in extreme operating condition, the flow rates are in the S 3 region. This
paper The S3 region
provides anrepresents the centrifugal
external interpolation pump to
method working
calculate above
the theflowQnrates
curve.
in Typically,
the S 3 re-
pumps operate in
gion, as shown in Equationthe S 2 and S
(5). 1 regions. When the valve opening of the system is so large
that the pump is in extreme operating condition, the flow rates are in the S3 region. This
paper provides an external interpolation method Pt − fk (N tot )calculate the flow rates in the S3 region,
Qest = Qk +  (Qk − Qk −1 ) (5)
as shown in Equation (5). fk (Nt ) − fk −1 (Nt )

Where fk-1 is the power–speed function Pt −at f kthe

( Ntflow
) rate Qk-1, and fk is the power–speed
Q =Q + · ( Q k − Q k −1 ), (5)
function at the flow rateestQk. k f k ( Nt ) − f k−1 ( Nt )

where f k−1 is the

3.3. Estimation power–speed
Model Curves at the flow rate Qk−1 , and fk is the power–speed
function
Based on PN
function at the flow rate Qk .
Figure 7 presents the power variations ΔP of each 2 m3/h flow rate at 50 Hz in the QP
curve.
3.3. As the flow
Estimation Modelrate increases,
Based dP/dQ gradually decreases, causing ΔP to be smaller.
on PN Curves
The QP curves for the other input frequencies have similar3power variations to 50 Hz. This
Figure 7 presents the power variations ∆P of each 2 m /h flow rate at 50 Hz in the QP
means that to make the power difference between PN curves approximately equivalent,
curve. As the flow rate increases, dP/dQ gradually decreases, causing ∆P to be smaller.
the flow rate difference between adjacent PN curves should be increased appropriately
The QP curves for the other input frequencies have similar power variations to 50 Hz. This
while in the high flow rate region.
means that to make the power difference between PN curves approximately equivalent, the
flow rate difference between adjacent PN curves should be increased appropriately while
in the high flow rate region.
Processes 2022, 10, x FOR PEER REVIEW 7 of 16

6
Processes 2022,
Processes 2022, 10,
10, 2163
x FOR PEER REVIEW 77 of 16
of 16
5

46

P (kW)
35

24

P (kW) 13 50 Hz 45 Hz
40 Hz 35 Hz
30 Hz 25 Hz
02
0 4 8 12 16 20
1 3 50 Hz 45 Hz
Q (m /h) 40 Hz 35 Hz
30 Hz 25 Hz
Figure
0 7. Power variation of QP curve.
0 4 8 12 16 20
The input frequencies of 50 Hz, 45 Hz, 40 Hz, 35 Hz, and 30 Hz are fitted to the curves
Q (m3/h)
with flow rates of Q0 = 0 m3/h, Q1 = 3 m3/h, Q2 = 6 m3/h, Q3 = 9 m3/h, Q4 = 15 m3/h, and Q5 =
Figure
21 Power
7. The
m3/h. variation
fitted of QP are
equations curve.
shown in Equations (6)–(11).
The
The input Q0 : P =of
input frequencies
frequencies 50
50 Hz,
Hz,- 45
0.57482
of 45 Hz,
Hz, 40
0.000915 Hz,
40212
Hz,N35+Hz,
35 and
and 30  Nfitted
Hz are
0.000000453394
Hz, 30 Hz are 2
, to
fitted the curves
to the curves
(6)
with flow rates of Q = 0 m 3 /h, Q = 3 m3 /h, Q = 6 m3 /h, Q = 9 m3 /h, Q = 15 m3 /h,
with flow rates of Q00= 0 m3/h, Q1 =13 m3/h, Q2 = 6 2m3/h, Q3 = 9 m33/h, Q4 = 15 m43/h, and Q5 =
3 /h. The fitted equations are shown in Equations (6)–(11).
21 mQ
and 5 =The
3/h. 21 mfitted Q1 : P = are
equations shown
0.60412  N + 0.000000704725
in Equations
-0.00105 (6)–(11).  N2 , (7)
Q0 :QP0 = = 0.57482
0.57482
: P − 0.000915212 · N N++0.000000453394
- 0.000915212 0.000000453394·N N22,, (6)
Q2 : P = 0.53861-0.000886451 N + 0.000000513669  N2 , (8)
Q1 : P = 0.60412 − 0.00105 · N + 0.000000704725 · N 22 , (7)
Q1 : P = 0.60412-0.00105  N + 0.000000704725  N , (7)
Q2 : P = : P = 0.63796-0.001
Q30.53861 − 0.000886451 N·+N0.000000615403  N2 ·, N 2 ,
+ 0.000000513669 (9)
(8)
Q23 :: PP==0.63796 − 0.001 · N + N + 0.000000513669
· N 2,N ,
2
Q 0.53861-0.000886451 0.000000615403 (8)
(9)
Q4 : P = 0.88357-0.00105  N + 0.000000952153 N2 , (10)
Q4 : P = 0.88357 − 0.00105 · N + 0.000000952153 · N2 2 , (10)
Q3 : P = 0.63796-0.001 N + 0.000000615403  N2 , (9)
= 1.82773-0.00256
Q5 :QP5 := P1.82773 − 0.00256 · N
N++ 0.00000125039
0.00000125039·NN 2.. (11)
(11)
Figure 88 presents Q
Figure P = 0.88357-0.00105
4 : fitted
the fitted PN curves
PN curvesat  N + 0.000000952153
atdifferent
different flowrates,
flow rates,and
andNthethe
2
, relevant
relevant (10)
results
results of
of each
each curve
curve areare
shownshown in Table
in Table 2. coefficient
2. The The coefficient of determination
of determination 2
R of all R six
2 of all six curves
fitted fitted
curves
is closeistoclose tothe
1, and 1, andQ5the
residual = 1.82773-0.00256
: Presidual
sum sum of(RSS)
of squares  N +is0.00000125039
squares (RSS)
close  Nto2indicates
tois0, close
which .0, whichthat indicates
the(11)
six
that thehave
curves six curves have high
high accuracy accuracy
and and are
the results the reliable.
results are reliable.
Figure 8 presents the fitted PN curves at different flow rates, and the relevant results
of each curve are shown in Table 2. The coefficient of determination R2 of all six fitted
curves
5 is close
Q0 toQ11, and
Q2 the residual sum of squares (RSS) is close to 0, which indicates
that the sixQcurves
3 Q4 haveQ5 high accuracy and the results are reliable.

35 Q0 Q1 Q2
P (kW)

Q3 Q4 Q5

24
P (kW)

13

02
1500 1750 2000 2250 2500 2750 3000
1 N (rpm)

Figure8.
Figure PNcurves.
8. PN curves.
0
1500 1750 2000 2250 2500 2750 3000
N (rpm)

Figure 8. PN curves.
Processes 2022, 10, 2163 8 of 16
Processes 2022, 10, x FOR PEER REVIEW 8 of 16

Table 2. Curve fitting results.

Table 2. Curve fitting results.
Items Q0 Q1 Q2 Q3 Q4 Q5
Items Q0 Q1 Q2 Q3 Q4 Q5
R2 2 0.9999 0.9999 0.99999 0.99996
R 0.9999−4 0.9999−4 0.99999 0.99996 11 −5 0.99994
0.99994
RSS 1.465 × 10 −4 1.051 × 10 −4 4.146 × 10−−55 1.716 × 10− 4 2.455 × 10−5 3.312 × 10−4
RSS 1.465 × 10 1.051 × 10 4.146 × 10 1.716 × 10−4 2.455 × 10 3.312 × 10−4

Figure 9 presents the the flow

flow chart
chartof ofthe
theprediction
predictionmodel.
model.For Forinstance,
instance,the thefirst
firststep
stepis is
to obtain a large amount
amount of of experimental
experimentaldata dataofofshaft
shaftpower
powerP,P,speedspeedN,N,and andflow flowraterateQQ
adjusting the
by adjusting theinput
inputfrequency
frequencyand andchanging
changingthe the valve
valve opening,
opening, thenthen organizing
organizing thethe
data
dataPN
into intocurves
PN curves with appropriate
with appropriate flow The
flow rates. rates.second
The second
step is step is to the
to collect collect
shaft thepower
shaft P
power
and P and
speed N ofspeed N of the centrifugal
the centrifugal pump and pump and the
calculate calculate
powerthe P0 ,power
P1 , P2 ,PP0,3P,P1, 4P , P3,PP54,of
, 2and
and P5 Q
curves of0curves
, Q1 , Q2Q, 0Q, 3Q, 1Q
,Q , Q 3, Q
4 ,2and Q54,at
and Q5 at different
different speeds. Thespeeds.
flowTherateflow rate
of the of the centrif-
centrifugal pump
ugal
is pump from
obtained is obtained from the
the collected collected
power power
and the and thepower.
calculated calculated power. For
For instance, theinstance,
predicted
the predicted
flow 0 m3rate
rate Qest =flow Qest = P0 ≤
/h when when P j≤≤PP
m3P/h0 ; when 0; ≤
when
Pj+1 ,Pthe
j≤ P ≤ Pj+1
flow , the
rate flow
can rate can beby
be calculated
calculated
the internalby the internalmethod;
interpolation interpolation P ≥ P5 , the
whenmethod; whenflowP ≥rate
P5, can
the be
flow rate can by
calculated be external
calcu-
lated by external
interpolation interpolation method.
method.

2.Estimation Centrifugal
pump system

P N

1. Plot PN curves
5 Q0 Q1 Q2
Q3 Q4 Q5

Input frequency Centrifugal P

P (kW)

3
N
Valve opening pump system
Q
2

0
1500 1750 2000 2250 2500 2750 3000
N (rpm)
5 Q0 Q1 Q2
Q3 Q4 Q5 P5
4 P4

P3
P (kW)

3
P2
2 P1
P0
1

0
Case1:P P0
1500 1750 2000 2250 2500 2750 3000
Qest = 0 m3 h N (rpm)

Case2:P0＜Pj P Pj+1 P5
Pt − f j ( N t ) Sorting
Qest = Q j +  (Q j +1 − Q j )
f j +1 ( N t ) − f j ( N t ) P、P0、P1、P2、
P3、P4、P5
Case3:P＞P5

P − f5 ( N )
Qest = 21 + 6
f5 ( N ) − f 4 ( N )

Figure 9.
Figure 9. Flow
Flow chart
chart of
of flow
flow rate
rateestimation
estimationalgorithm.
algorithm.
Processes 2022, 10, x FOR PEER REVIEW 9 of 16

Processes 2022, 10, 2163 9 of 16

4. Influence Factors on the Estimation Accuracy

4.1. PN Curves
4. Influence Factors on the Estimation Accuracy
4.1. The PN curves fitted by limited working points somehow deviated from the actual
PN Curves
PN curves, which would lead to erroneous estimates of the pumps’ flow rate. At the same
The PN curves fitted by limited working points somehow deviated from the actual PN
time, thewhich
curves, difference
would in flow
lead rates of adjacent
to erroneous estimates PNof thecurves
pumps’ will
flow also affect
rate. At thethe
sameprediction
time, re-
sults. In addition,
the difference therates
in flow location of the PN
of adjacent centrifugal
curves will pump operating
also affect point affects
the prediction theInpredic-
results.
tion results.
addition, theIn Figure
location of 10, we assumed
the centrifugal pump that the shaft
operating power
point affectsofthethe motor produces
prediction results. ΔP
fluctuations.
In Figure 10, The error inthat
we assumed its flow prediction
the shaft power of isthe
more than
motor 6 m /h in
produces 3 ∆Parea A of theThe
fluctuations. PN curve
error in the
because its flow prediction
curve is more
in this area than 6 m3 /h
is relatively in area
dense, A of the
which PN curve
means that because
the power the curve
variation of
in this
the area isPN
adjacent relatively
curve is dense, which
slight. Whenmeans
the that
pump the operates
power variation
in areaofB,the theadjacent
error in PNits flow
curve is slight. When the pump operates in area B, the error in its flow
prediction3is between 3 and 6 m /h. When the pump works in area C, the error in its flow
3 prediction is between
3 and 6 m /h. When the pump works in area C, the error in its flow prediction is smaller
prediction3 is smaller than 3 m3/h, where the curve is sparse, and the power of the adjacent
than 3 m /h, where the curve is sparse, and the power of the adjacent PN curve varies
PN curveItvaries
widely. widely.
indicates It indicates
that centrifugal that operating
pumps centrifugal in pumps
the low-speedoperating
regionin are
themore
low-speed
region aretomore
sensitive power sensitive
variations.to Therefore,
power variations.
the locationTherefore, the location
of the centrifugal pump’s ofoperating
the centrifugal
pump’s operating point significantly influences
point significantly influences the prediction results. the prediction results.

5 Q0 Q1 Q2
Q3 Q5 Q6
C
4
B
P (kW)

3 A ΔP
2

1
ΔP

1750 2000 2250 2500 2750 3000

N (rpm)

Figure10.
Figure Variation of
10.Variation ofestimation
estimationforfor
different working
different points.
working points.
4.2. Impact of Shaft Power
4.2. Impact of Shaft Power
In addition to the PN curve, the fluctuation of the motor shaft power can change the
In addition
estimation to the
flow rate. PN curve,
Hence, it also the fluctuation
impacts of theresults.
the prediction motor We shaft powerthe
recorded canmotor
change the
shaft power
estimation fluctuations
flow 50 times
rate. Hence, at 50
it also Hz, 45 the
impacts Hz,prediction
40 Hz, 35 Hz, and 30We
results. Hzrecorded
and adjusted
the motor
the valve to set the flow rate at 0 m 3 /h, 9 m3 /h, and 18 m3 /h, respectively. As shown in
shaft power fluctuations 50 times at 50 Hz, 45 Hz, 40 Hz, 35 Hz, and 30 Hz and adjusted
Figure
the valve11,to
thesetfluctuation
the flowofrate
motor at shaft
0 m3power
/h, 9 mis3more significant
/h, and 18 m3/h, at high input frequencies
respectively. As shown in
than at low input frequencies. However, the prediction results are less affected by power
Figure 11, the fluctuation of motor shaft power is more significant at high input frequen-
fluctuations when the input frequency is high since the PN curves of the centrifugal pump
cies than at low input frequencies. However, the prediction results are less affected by
are sparse in this region (see Figure 8). Meanwhile, motor shaft power fluctuates slightly
power
at the fluctuations when thearea
lower speed operating input frequency
where the PNiscurve
highissince theTherefore,
dense. PN curves theofimpact
the centrifugal
of
pump are sparse in this region (see Figure 8). Meanwhile,
power fluctuations is also insignificant at low input frequencies. motor shaft power fluctuates
slightly at the lower speed operating area where the PN curve is dense. Therefore, the
impact of power fluctuations is also insignificant at low input frequencies.
Fluctuations in motor speed also impact the prediction. However, this impact is
pretty small and much lower than the impact of shaft power fluctuations. Therefore, it is
not discussed in this paper.
Processes 2022, 10, x FOR PEER REVIEW
Processes 2022, 10, 2163 10 of 16
10

0 50 100 150 200 250

0.04 50 Hz 45 Hz 40 Hz 35 Hz 30 Hz

Q = 0 m3/h

ΔP (kW)
0.02

0.00

−0.02

0.0194
Q = 9 m3/h
ΔP (kW)

0.0097

0.0000

−0.0097

0.0198
Q = 18 m3/h
ΔP (kW)

0.0099

0.0000

−0.0099

0 50 100 150 200 250

Number of test samples

Figure 11. Fluctuation of the shaft power.

Figure 11. Fluctuation of the shaft power.
Fluctuations in motor speed also impact the prediction. However, this impact is pretty
small and much
5. Results andlower than the impact of shaft power fluctuations. Therefore, it is not
Discussion
discussed in this paper.
We measured the flow rates at input frequencies of 46 Hz, 41 Hz, 36 Hz, and 3
5.and
Results and Discussion
compared them with the predictive flow rates to verify the accuracy of the PN cu
interpolation
We measured method.
the flowMeanwhile,
rates at inputwe also provided
frequencies of 46 Hz,the predictive
41 Hz, 36 Hz, and flow31 rates
Hz usin
and compared them with the predictive flow rates to verify the accuracy
QP method [14] and the BPNN method [23]. The PN curves interpolation method is of the PN curves
interpolation method. Meanwhile, we also provided the predictive flow rates using the
accurate than other methods regardless of the input frequency of 46 Hz, 41 Hz, 36 H
QP method [14] and the BPNN method [23]. The PN curves interpolation method is more
31 Hz (see Figure 12).
accurate than other methods regardless of the input frequency of 46 Hz, 41 Hz, 36 Hz, or
31 Hz (see Figure 12).
Measured
20
Measured
20
The horizontal andinterpolation
PN curves vertical axes of Figure 13 show the measured and predicted flow
PN curves interpolation
BPNN 16 BPNN
rates, respectively.
16 QP The closer the points in the graph are to the
QP line y = x, the closer the
predicted flow is to the measured flow. Whether the12flow rate is small or large, the point
Q (m3/h)

Q (m3/h)

from the 12PN curves interpolation method46isHzalways the closest distance to the line 41 y Hz
= x,
which indicates
8 that the PN curve interpolation method8 can better predict the flow rate
(see Figure 13).
4 4

0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
Number of testing samples Number of testing samples

15
18 Measured Measured
PN curves interpolation PN curves interpolation
15 BPNN 12 BPNN
QP QP
Q (m3/h)

Q (m3/h)

12
9
36 Hz 31 Hz
9
6
6
3
3

0 0
 5. Results and Discussion
We measured the flow rates at input frequencies of 46 Hz, 41 Hz, 36 Hz, and 31 Hz
and compared them with the predictive flow rates to verify the accuracy of the PN curves
interpolation method. Meanwhile, we also provided the predictive flow rates using the
Processes 2022, 10, 2163 QP method [14] and the BPNN method [23]. The PN curves interpolation method is more 11 of 16
accurate than other methods regardless of the input frequency of 46 Hz, 41 Hz, 36 Hz, or
31 Hz (see Figure 12).
20
20 Measured Measured
PN curves interpolation PN curves interpolation
BPNN 16 BPNN
16 QP QP

Q (m3/h)

Q (m3/h)
12
12 46 Hz 41 Hz

8
8

4 4

0 0
0 3 6 9 12 15 18 0 3 6 9 12 15 18
Number of testing samples Number of testing samples

Processes 2022, 10, x FOR PEER REVIEW18 Measured

15
Measured
11 of 16
PN curves interpolation PN curves interpolation
15 BPNN 12 BPNN
QP QP
Q (m3/h)

Q (m3/h)
12
9
36 Hz
31 Hz
The horizontal and vertical axes of Figure 13 show the measured
9 and predicted flow
6
rates,
6 respectively. The closer the points in the graph are to the line y = x, the closer the
predicted flow is to the measured flow.
3
3
Whether the flow rate is small or large, the point
from0 the PN curves interpolation method 0 is always the closest distance to the line y = x,
0 3 6 9 12 15 18 0 3 6 9 12 15 18
which indicatesNumber ofthat
testing the PN curve interpolation
samples Numbermethod can better predict the flow rate
of testing samples
(see Figure 13).
Figure 12. Comparison of the predictive flow rates of the three methods.
Figure 12. Comparison of the predictive flow rates of the three methods.

PN curves interpolation
20 BPNN
QP
Predicted flow rate Q (m3/h)

15

10

0 2 4 6 8 10 12 14 16 18 20 22

Actual measured Q (m3/h)

13.Comparison
Figure 13.
Figure Comparisonof of
thethe
results of the
results of three estimation
the three methods.
estimation methods.

Figure 14 presents the absolute and relative errors of the three methods, and the
Figure 14 presents the absolute and relative errors of the three methods, and the spe-
specific data are shown in Table 3. The mean absolute error of the interpolation method is
cific
0.1756data arewhich
m3 /h, shown in Table
is much lower3.than
Thethe
mean
otherabsolute error methods.
two prediction of the interpolation
Meanwhile, the method is
0.1756 m /h, which
3
mean relative error ofisthe
much lower than
interpolation the other
method two prediction
is 2.6919%, methods.
while the mean Meanwhile,
relative error the
mean
of BPNN relative
and QP error of the interpolation
prediction method
methods is 16.6261% andis23.3509%,
2.6919%, both
whileof the mean
which are relative
much error
of BPNN
higher thanand
the QP prediction
relative error of methods is 16.6261%
the interpolation method.and 23.3509%,
This indicates both of PN
that the which are much
curve
interpolation
higher prediction
than the relativemethod is the
error of more accurate andmethod.
interpolation stable. This indicates that the PN curve
interpolation prediction method is more accurate and stable.

0 5 10 15 0 5 10 15

4.2 PN curves interpolation BPNN QP PN curves interpolation BPNN QP

144
δ (m3/h)

ε (%)

2.8 46 Hz 46 Hz
96
1.4 48
0.0 0
3.3 60
/h)

41 Hz 41 Hz
)
3
Processes 2022, 10, 2163 12 of 16

(a) (b)
Figure 14. Errors of the three estimation methods: (a) absolute errors and (b) relative errors.

Table 3. Error data of the three estimation methods.

Errors
Items
Maximum Average Minimum
δ (m3 /h) 0.5409 0.1756 0.0021
PN curves interpolation
ε (%) 16.9814 2.6919 0.0181
δ (m3 /h) 2.7565 1.3851 0.0717
BPNN
ε (%) 61.9739 16.6261 2.0429
δ (m3 /h) 3.4191 1.8315 0.0496
QP
ε (%) 137.6936 23.35089 1.3291

The QP curve method has nice applications for flow prediction, but the affinity law,
which is an important basis of the QP curve method, is not applicable when the speed
difference is large [16]. The BPNN flow prediction model is proposed without considering
the applicability of the affinity law. Therefore, the BPNN model has higher accuracy than
the QP model, as we can see from Figure 12, since even a slight deviation in power can cause
significant errors in the flow rate prediction of BPNN. We propose an interpolation method
based on the PN curves. The influence of power deviation is reduced by interpolation
of the two PN curves in the calculation of the flow rate, as shown in Figure 6. Therefore,
the interpolation method based on the PN curves has the highest consistency with the
measurement results than the other two methods.

6. Application of the PN Curves Interpolation Method

A validation test was run in a closed centrifugal pump system to verify the applicability
of the PN curves interpolation method. We controlled the pump’s flow rate through the
valve and frequency converter, and the controller transmitted the speed and power of the
motor to the host software via Ethernet. Then, the flow rate was calculated by the PN curve
interpolation method programmed into the software. Figure 15 presents the PN curve
interpolation method, accurately predicting the flow rates under different input frequencies
and valve openings. To determine whether the centrifugal pump works at low or high flow

1
Processes 2022, 10, 2163 13 of 16

Processes
Processes2022,
2022,10,
10,xxFOR
FORPEER conditions,
PEERREVIEW
REVIEW the points derived from the prediction model are distributed on the line13yof
13 =16
of x,
16
which indicates that the flow rate of the centrifugal pump can be well predicted based on
the PN curves interpolation method (see Figures 16 and 17).

18
18 Measured 18
18 Measured
Measured Measured
Predicted
Predicted Predicted
Predicted
15
15 15
15

12
12 12
12

/h)
/h)

(m33/h)
(m33/h)

47.5
47.5Hz
Hz 42.5
42.5Hz
Hz

QQ(m
QQ(m

99 99

66 66

33 33

00 00
00 22 44 66 88 10
10 12
12 14
14 16
16 18
18 00 22 44 66 88 10
10 12
12 14
14 16
16 18
18
Number
Numberof
oftest
testsamples
samples Number
Numberof
oftest
testsamples
samples

18
18 Measured 18
18 Measured
Measured Measured
Predicted
Predicted Predicted
Predicted
15
15 15
15

12
12 12
12
/h)

/h)
(m33/h)

(m33/h)

37.5
37.5Hz
Hz 32.5
32.5Hz
Hz
QQ(m

QQ(m

99 99

66 66

33 33

00 00
00 22 44 66 88 10
10 12
12 14
14 16
16 18
18 00 22 44 66 88 10
10 12
12 14
14 16
16 18
18
Number
Numberof
oftest
testsamples
samples Number
Numberof
oftest
testsamples
samples

Figure
Figure15. Results
15.Results of
Resultsof flow
offlow
flowestimation.
estimation.

18
18
/h)
(m33/h)

15
15
rateQQ(m

12
12
flowrate
Predictedflow

99
Predicted

66

33

00
00 22 44 66 88 10
10 12
12 14
14 16
16 18
18 20
20
3
Actual
Actualflow
flowrate
rate Q
Q (m
(m3/h)
/h)
Figure
Figure16.
16.Comparison
Comparisonof
Comparison ofpredicted
predictedand
andactual
actualflow
flowrates.
rates.
Processes 2022, 10, x FOR PEER REVIEW 14 of 16
Processes 2022, 10, 2163 14 of 16

(a) (b)
0 5 10 15 20 0 5 10 15 20

0.33 47.5 Hz 8.4 47.5 Hz

δ (m3/h)

ε (%)
0.22 5.6

0.11 2.8

0.00 0.0

0.57 42.5 Hz 22.8 42.5 Hz

δ (m3/h)

ε (%)
0.38 15.2

0.19 7.6

0.00 0.0

0.72 37.5 Hz 29.1 37.5 Hz

δ (m3/h)

ε (%)
0.48 19.4

0.24 9.7

0.00 0.0

0.72 32.5 Hz 19.8 32.5 Hz

δ (m3/h)

ε (%)

0.48 13.2

0.24 6.6

0.00 0.0
0 5 10 15 20 0 5 10 15 20

Actual flow rate Q (m3/h) Actual flow rate Q (m3/h)

Figure 17. 17.

Figure Errors of of
Errors thetheestimated
estimatedflow
flow rate: (a)absolute
rate: (a) absoluteerrors
errors and
and (b)(b) relative
relative errors.
errors.

7. Conclusions
7. Conclusions
This paper proposed a flow rate prediction model based on the PN curve interpolation
This paper
method. proposed
The factors a flow
affecting the rate prediction
accuracy model based
of the estimation on were
method the PN curve Three
analyzed. interpola-
tiondifferent
method. The factors affecting the accuracy of the estimation method were analyzed.
estimation models were compared, and the feasibility of this method was verified
Three different
through estimation
experiments. Themodels were
following compared,
conclusions wereand the feasibility of this method was
drawn:
verified through
(1) The experiments.
PN curves The
containing following
speed, power, conclusions were drawn:
and flow information were presented, and
(1) TheanPN interpolated flow prediction
curves containing speed,model
power,based
andon these
flow curves waswere
information established. The and
presented,
factors affecting the accuracy of the estimation method were analyzed,
an interpolated flow prediction model based on these curves was established. The which mainly
include the flow rate and accuracy of the PN curve, the location of the centrifugal
factors affecting the accuracy of the estimation method were analyzed, which mainly
pump operational point, and the fluctuation of the shaft power. Particularly in this
include the flow rate and accuracy of the PN curve, the location of the centrifugal
aspect of the location of the centrifugal pump operational point, centrifugal pumps
pump operational
operating at a lowpoint,
speedsandwerethe fluctuation
more ofpower
sensitive to the shaft power.Furthermore,
variations. Particularlythein this
aspect of the location of the centrifugal pump operational point, centrifugal
fluctuation of the motor shaft power can cause variations in the estimation flow rate. pumps
(2)operating at a lowestimation
Three different speeds were modelsmore
weresensitive
compared. to power variations.
The average absoluteFurthermore,
and relative the
fluctuation
errors of ofthethe motor shaft
interpolated power
flow can cause
prediction variations
method m3estimation
in the
are 0.1756 flow rate.
/h and 2.6919%,
3 3
(2) Three different
compared estimation
to 1.3851 m /h models
and 6.6261%wereforcompared.
BPNN andThe average
1.8315 m /h andabsolute and relative
23.35089% for
the QP
errors method.
of the This means
interpolated flowthat the PN curve
prediction interpolation
method method
are 0.1756 m3/hhas high
and accuracycom-
2.6919%,
in estimating
pared to 1.3851 them3/hflow
andrates.
6.6261% for BPNN and 1.8315 m3/h and 23.35089% for the
(3) The PN curve interpolation method was tested in an industrial application, and its
QP method. This means that the PN curve interpolation method has high accuracy
average absolute errors at 47.5 Hz, 42.5 Hz, 37.5 Hz, and 32.5 Hz are 0.1442 m3 /h,
in estimating the flow rates.
0.2047 m3 /h, 0.2197 m3 /h, and 0.1979 m3 /h, and its average relative errors are
(3) The2.0816%,
PN curve interpolation method was tested in an industrial application, and its
3.2875%, 3.6981%, and 2.9419%. This highly accurate prediction capability
average absolute
fully meets errors at 47.5
the monitoring and Hz, 42.5requirements
control Hz, 37.5 Hz, andand 32.5 Hz
improves theare 0.1442ofm3/h,
accuracy
0.2047 m3/h, 0.2197
centrifugal pump m 3/h, and 0.1979 m3/h, and its average relative errors are 2.0816%,
flow prediction at small flow rates.
3.2875%, 3.6981%, and 2.9419%. This highly accurate prediction capability fully meets
the monitoring and control requirements and improves the accuracy of centrifugal
pump flow prediction at small flow rates.

Author Contributions: Conceptualization, Y.W.; methodology, Y.W.; software, M.F.; validation,

D.W. and Y.G.; formal analysis, G.X.; investigation, D.W.; resources, J.M.; data curation, Y.G.; writ-
Processes 2022, 10, 2163 15 of 16

Author Contributions: Conceptualization, Y.W.; methodology, Y.W.; software, M.F.; validation, D.W.
and Y.G.; formal analysis, G.X.; investigation, D.W.; resources, J.M.; data curation, Y.G.; writing—
original draft preparation, Y.G.; writing—review and editing, D.W.; visualization, M.F.; supervision,
J.M.; project administration, D.W.; funding acquisition, D.W. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was funded by the Natural Science Foundation of Zhejiang Province, grant
number LGG21E090002, the China Postdoctoral Science Foundation, grant number 2021M691383,
and the Science and Technology Program of Zhejiang Province, grant number 2021C01052.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. De Almeida, A.; Fong, J.; Brunner, C.; Werle, R.; Van Werkhoven, M. New technology trends and policy needs in energy efficient
motor systems—A major opportunity for energy and carbon savings. Renew. Sustain. Energy Rev. 2019, 115, 109384. [CrossRef]
2. Torregrossa, D.; Capitanescu, F. Optimization models to save energy and enlarge the operational life of water pumping systems.
J. Clean. Prod. 2018, 213, 89–98. [CrossRef]
3. Wang, Z.; Qian, Z.; Lu, J.; Wu, P. Effects of flow rate and rotational speed on pressure fluctuations in a double-suction centrifugal
pump. Energy 2018, 170, 212–227. [CrossRef]
4. Goman, V.; Oshurbekov, S.; Kazakbaev, V.; Prakht, V.; Dmitrievskii, V. Energy Efficiency Analysis of Fixed-Speed Pump Drives
with Various Types of Motors. Appl. Sci. 2019, 9, 5295. [CrossRef]
5. Oshurbekov, S.; Kazakbaev, V.; Prakht, V.; Dmitrievskii, V.; Gevorkov, L. Energy Consumption Comparison of a Single Variable-
Speed Pump and a System of Two Pumps: Variable-Speed and Fixed-Speed. Appl. Sci. 2020, 10, 8820. [CrossRef]
6. Capurso, T.; Bergamini, L.; Torresi, M. A new generation of centrifugal pumps for high conversion efficiency. Energy Convers.
Manag. 2022, 256, 115341. [CrossRef]
7. Shankar, V.K.A.; Umashankar, S.; Paramasivam, S.; Hanigovszki, N. A comprehensive review on energy efficiency enhancement
initiatives in centrifugal pumping system. Appl. Energy 2016, 181, 495–513. [CrossRef]
8. Viholainen, J.; Tamminen, J.; Ahonen, T.; Ahola, J.; Vakkilainen, E.; Soukka, R. Energy-Efficient Control Strategy for Variable
Speed-Driven Parallel Pumping Systems. Energy Effic. 2013, 6, 495–509. [CrossRef]
9. Gu, Y.; Zhang, J.; Yu, S.; Mou, C.; Li, Z.; He, C.; Wu, D.; Mou, J.; Ren, Y. Unsteady Numerical Simulation Method of Hydrofoil
Surface Cavitation. Int. J. Mech. Sci. 2022, 228, 107490. [CrossRef]
10. Luna, T.; Ribau, J.; Figueiredo, D.; Alves, R. Improving Energy Efficiency in Water Supply Systems with Pump Scheduling
Optimization. J. Clean. Prod. 2019, 213, 342–356. [CrossRef]
11. Lindstedt, M.; Karvinen, R. Optimal Control of Pump Rotational Speed in Filling and Emptying a Reservoir: Minimum Energy
Consumption with Fixed Time. Energy Effic. 2016, 9, 1461–1474. [CrossRef]
12. Luo, Y.; Zhixiang, X.; Sun, H.; Yuan, S.; Yuan, J. Research on the Induction Motor Current Signature for Centrifugal Pump at
Cavitation Condition. Adv. Mech. Eng. 2015, 7, 708902. [CrossRef]
13. Benlaloui, I.; Drid, S.; Chrifi-Alaoui, L.; Ouriagli, M. Implementation of a New MRAS Speed Sensorless Vector Control of
Induction Machine. IEEE Trans. Energy Convers. 2015, 30, 588–595. [CrossRef]
14. Hammo, S.; Viholainen, J. Providing Flow Measurement in Parallel Pumping Systems from Variable Speed Drives. World Pumps
2006, 2006, 30–33. [CrossRef]
15. Ahonen, T.; Tamminen, J.; Ahola, J.; Viholainen, J.; Aranto, N.; Kestilä, J. Estimation of pump operational state with model-based
methods. Energy Convers. Manag. 2010, 51, 1319–1325. [CrossRef]
16. Ahonen, T.; Tamminen, J.; Ahola, J.; Kestila, J. Frequency-Converter-Based Hybrid Estimation Method for the Centrifugal Pump
Operational State. IEEE Trans. Ind. Electron. 2011, 59, 4803–4809. [CrossRef]
17. Tamminen, J.; Viholainen, J.; Ahonen, T.; Ahola, J.; Hammo, S.; Vakkilainen, E. Comparison of Model-Based Flow Rate Estimation
Methods in Frequency-Converter-Driven Pumps and Fans. Energy Effic. 2014, 7, 493–505. [CrossRef]
18. Ahonen, T.; Tamminen, J.; Ahola, J.; Niinimaki, L.; Tolvanen, J. Sensorless estimation of the pumping process characteristics by a
frequency converter. In Proceedings of the 15th European Conference on Power Electronics and Applications (EPE), Lille, France,
3–5 September 2013; pp. 1–10. [CrossRef]
19. López-Lineros, M.; Estévez, J.; Giráldez, J.V.; Madueño, A. A New Quality Control Procedure Based on Non-Linear Autoregressive
Neural Network for Validating Raw River Stage Data. J. Hydrol. 2014, 510, 103–109. [CrossRef]
20. Claveria, O.; Torra, S. Forecasting Tourism Demand to Catalonia: Neural Networks vs. Time Series Models. Econ. Model. 2014, 36,
220–228. [CrossRef]
21. Wu, Q.; Shen, Q.; Wang, X.; Yang, Y. Estimation of Centrifugal Pump Operational State with Dual Neural Network Architecture
Based Model. Neurocomputing 2016, 216, 102–108. [CrossRef]
Processes 2022, 10, 2163 16 of 16

22. Han, W.; Nan, L.; Su, M.; Chen, Y.; Li, R.; Zhang, X. Research on the Prediction Method of Centrifugal Pump Performance Based
on a Double Hidden Layer BP Neural Network. Energies 2019, 12, 2709. [CrossRef]
23. Kong, X.; Ma, S.; Ma, T.; Li, Y.; Cong, X. Mass Flow Rate Prediction of Direct-Expansion Solar-Assisted Heat Pump Using R290
Based on ANN Model. Sol. Energy 2021, 215, 375–387. [CrossRef]
24. Wu, D.; Huang, H.; Qiu, S.; Liu, Y.; Wu, Y.; Ren, Y.; Mou, J. Application of Bayesian Regularization Back Propagation Neural
Network in Sensorless Measurement of Pump Operational State. Energy Rep. 2022, 8, 3041–3050. [CrossRef]
25. Luo, H.; Zhou, P.; Shu, L.; Mou, J.; Zheng, H.; Jiang, C.; Wang, Y. Energy Performance Curves Prediction of Centrifugal Pumps
Based on Constrained PSO-SVR Model. Energies 2022, 15, 3309. [CrossRef]