Chapter 2

Literature Review

In this Chapter, literature review related to pump running as turbine is presented. Historical
development and present status of PAT technology are discussed. Theoretical, experimental and
numerical investigations carried out worldwide by various researchers for pump selection,
cavitation study, performance enhancement of PAT etc. are summarized. Based on the literature
review, research gaps are identified and objectives of the present study are framed.

2.1 Historical Development of PAT

When the pumps were first used in turbine mode is unclear. In 1931, when Thoma and Kittredge
(1931) were trying to evaluate the complete characteristics of pumps, they accidentally found that
13
pumps could be operated very efficiently in the turbine mode. The turbine mode operation
became an important research question for many manufacturers as pumps were prone to
abnormal operating conditions. Later in 1941, Knapp (1941) published the complete pump
characteristics for a few pump designs based on experimental investigations. In the 1950s and
1960s, the concept of pumped storage power plants, in the range of 50 to 100 MW, was evolved
mainly in developed countries to manage the peak power requirements. In later years, chemical
industries became another area for the application of PATs for energy recovery. Even in water
supply networks identical applications of this technology were found. This background gave
some momentum to a rich phase of research and then onwards, standard manufactured pumps
were studied in turbine mode. In later years, many more techniques were developed by many
researchers (Rawal and Kshirsagar, 2007).

The technology for the use of PAT for electrical power generation was not available earlier.
However, advances in electrical machinery control technologies which allow the driving
regulation with variable velocity, rotation sense and torque have created the possibility of the
utilization of pumps working in inverse mode for power generation (Fernandez et al., 2004).
Agostenelli and Shafer (2013) tested many pumps in turbine mode over the years and concluded
that when a pump operates in a turbine mode, its mechanical operation is smooth and quiet; its
peak efficiency is same as in pump mode; head and flow at the best efficiency point (BEP) are
higher than that in pump mode and the power output is higher than that the pump input power at
its best efficiency. Various pumps which can be used as turbines for the power range of 1 kW to 1
MW are shown in Figure 2.1 (Chapallaz et al., 1992). It can be seen that multi stage radial flow
pumps are suitable for high head and low discharge sites; whereas, axial flow pumps are
appropriate in low head and high discharge range. Similar chart was also presented by Orchard
and Klos (2009) (range: 5 to 750 kW).

The field applications of multistage, single impeller centrifugal and axial pumps can be compared
with Pelton, Francis and Kaplan turbines respectively (Derakhshan and Nourbakhsh, 2008a). In
line and double suction pumps can also be used as turbines, but in turbine mode they are
comparatively less efficient. Self-priming pumps are not suitable for PAT application because
they have non-return valve to prevent the reverse flow. Dry-motor submersible pumps, in which
14
fin cooling arrangement is provided for motor, are not suitable for power production because they
overheat unless installed below the water level. Wet-motor submersible bore hole pumps usually
contain a non-return valve and may also support thrust bearing; hence, are normally not suitable
for turbine mode applications (Sharma, 1998). Pump works as turbine most efficiently in the head
range of 13 to 75 m. Also, as head increases the cost per kW decreases, as is the case in general
with all turbines (http://www.hawaiirenewable.com/wp-content/uploads/2009/12/5.....).

Figure 2.1. Different pumps suitable as turbines (Chapallaz et al., 1992).

2.2 Criteria for Selection of Pump Running in Turbine Mode

Selection of a proper pump to be used as turbine is a big problem in installation of PAT for a
particular site. PAT can become a cost-effective alternative to traditional turbines if the turbine
mode curves can be determined (Teuteberg, 2010). Normally, the pump manufacturers do not
provide characteristic curves of their pumps working as turbines. This makes it difficult to select
an appropriate pump to run as a turbine for a specific operating condition. One of the main

15
objectives of all PAT researchers all over the world has been to build a method that would make
accurate predictions of the turbine operation of pumps.

A large number of theoretical and experimental studies have been found in the literature for the
prediction of performance of PAT in which relations between BEPs in pump and turbine modes
were derived based on either efficiency or specific speed in pump mode. The specific speed is
one of the main parameters of turbo-machines that characterizes the type of the runner (i.e. radial,
mixed or axial flow), blade shape, spiral casing and other design features. The BEP parameters in
turbine mode are different from that in pump mode. Hence, the relation between these parameters
is derived by the investigators in terms of head and discharge correction factors (h and q), which
are defined as under:
h = Ht/Hp (2.1)
q = Qt/Qp (2.2)

Stepanoff (1957), Childs (1962), Hancock (1963), McClaskey & Lundquist (1976), Sharma
(1985), Luneburg & Nelson (1985), Schmiedl (1988) and Alatorre-Frenk (1994) developed
relations based on efficiency in pump mode; whereas, Kittredge (1961), Diederich (1966),
Grover (1980), Lewinski & Keslitz (1987), Buse (1981) and Hergt (adapted from Lewinski &
Keslitz, 1987) derived the relations based on the specific speed of the pump.

The historical development in the performance prediction methods of PAT (up to late 90s) is
presented in Table 2.1, which shows the relations for h and q derived by different researchers.
The deviations between performance predicted by these methods and experimental results have
been found to be around 20% or even more (Chapallaz et al., 1992; Williams, 1995). Therefore,
these methods are confined to preliminary selection of pumps to be used as turbines which is
important to obtain a rough estimation of turbine mode characteristics from the pump mode
characteristics (Nautiyal et al., 2011). Few of these methods may not include the performance
prediction at no-load, part load and overload operating conditions. Further, these methods have
not been tested on different pump shapes to get more insights about their features like reliability,
accuracy and robustness (Singh, 2005).

16
 Table 2.1 Historical development in the performance prediction methods of PAT.
Reference Criteria Head correction Discharge correction Remarks
Factor (h) factor (q)
Stepanoff (1957) BEP 1 1 Accurate for
p Ns: 40-60
Childs (1962) BEP 1 1 -

Hancock (1963) BEP 1 1 -

Grover (1980) Specific 2.693 2.379 0.0264Nst Applied for
speed 0.0229Nst Ns: 10-50
Hergt (1987) Specific 6 1.6 -
1.3 1.3
speed Nst 3 Nst 5

Sharma (1985) BEP 1 1 Accurate for

1.2 0.8
Ns: 40-60
Schmiedl (1988) BEP 2.5 2.4 -
1.4 + 1.5 +
2
Alatorre-Frenk BEP 1 0.85 5 + 0.385 -
5
0.85 + 0.385
(1994) 2 9.5 + 0.205
Sharma (1998) BEP 2 1.1 1.1 Ng
[ ] 1.2 0.8
f
= 240
p
N

Few other researchers viz. Knapp (1941), Wong (1987), Cohrs (1997) and Amelio and Barbarelli
(2004) have also given the PAT prediction methods based on the pump design, its geometry and
assumptions of some complex hydraulic phenomena like losses and slip effects to bring out more
accurate method. These methods are definitely comprehensive but they are difficult to implement
and simply beyond the reach of planners, since these methods need very detailed information,
which is sometimes patented or available only with the manufacturers (Singh and Nestmann,
2010). In last few years, many researchers have used different theoretical, experimental and

17
numerical techniques to predict the PAT performance from pump characteristics which are
described in this section.

2.2.1 Theoretical studies for selection criteria

Different methods for forecasting PAT performance have been proposed in the literature; some
based on pump-mode performance and others are on the geometry of the machine. By means of
the economic methodology, the accuracy of the former methods was evaluated by Alatorre-Frenk
(1994). A new method, based on pump-mode performance and type of casing was developed as
illustrated in Table 2.1. It was found that, under typical conditions, the inaccuracy of the better
prediction methods reduces the benefit-to-cost ratio of the scheme by less than 1% which may be
considered negligible in view of the high cost of turbine-mode laboratory tests.

Williams (1994) compared eight different PAT performance prediction methods based on turbine
tests on 35 pumps in the specific speed range of 12.7 to 183.3 and studied the effects of poor
turbine prediction on the operation of PAT. The difference between the predicted BEP and the
actual BEP for the PAT was studied using the prediction coefficient (C). The value of C was
computed from the ellipse formed on head versus discharge (H-Q) chart centered on the
guaranteed BEP as per the British Standard and the criterion for an acceptable prediction was C
1. The mean value of C and the number of pumps for which the value of C was greater than 1
(i.e. the error is outside the acceptable limits) for different methods is listed in Table 2.2. None of
the eight methods had given an accurate prediction for all the pumps but Sharmas method was
found to be better than the other methods; hence, it was recommended as a first estimate for
prediction of the turbine performance. It was suggested that, whenever possible once a pump has
been selected for a micro-hydro scheme it shall be tested in turbine mode in order to be certain of
its turbine characteristics before installation.

18
 Table 2.2 Comparison of turbine prediction methods (Williams, 1994).
Prediction method Mean value of C No. of pumps Pumps out
for which C > 1 of range (%)
Stepanoff (1957) 0.847 12 34
Childs (1962) 0.921 14 40
Hancock (1963) 0.906 10 32
Grover (1980) 1.333 22 81
Schmiedl (1985) 1.173 13 40
Sharma (1985) 0.733 7 20
Hergt (1987) 0.865 11 32
Alatorre-Frenk (1994) 0.852 10 29

Saini (1998) developed nomogram based on the specific speed in pump mode which may be used
to determine the correction factors/conversion coefficients for head and discharge. Using these
factors, the values of the BEP parameters in turbine mode can be directly achieved from the
nomogram which reduces the calculations work involved in the selection.

Sharma (1998) discussed the suitability of different pumps which can be used as turbines. The
analysis of performance curves in pump and turbine mode plotted by Grant and Bain (1985)
revealed that the location of turbine BEP is at a higher flow and head than the pump BEP. The
study showed that, ratio of turbine capacity and head at BEP to pump capacity and head at BEP
varies with specific speed in the range of 1.1 to 2.1. The equations for h and q were presented to
convert pump data into turbine performance data, as illustrated in Table 2.1, where Nm and Ng are
motor and generator speeds in rpm; and t are efficiencies in pump and turbine mode; f is
frequency and p is number of poles. It was pointed out that, the proposed equations were only
approximate and the actual values of Qt and Ht may vary by + 20% of the predicted value for the
BEP. Hence, after initial selection, it was recommended to test the pump under turbine condition
to find out power at the available head. The advancements in pump manufacturing technology
and various ways of reducing the cost of PAT based hydropower plants were also discussed.

19
Fernandez et al. (2004) emphasized that most of the studies carried out for performance
prediction were based on the hypothesis of the similarity between maximum efficiencies working
in both modes, which is not easy to maintain, and other studies were based on algebraic relations
as a function of efficiency. All of them have been achieved considering equal rotational speeds in
both working modes. In this paper, the characteristics of pump acting in inverse mode at different
rotational speeds were presented. The experiments were performed on a hydraulic set-up
developed according to ISO 3555 (1977) in the pump and turbine modes; and constant speed and
constant head characteristics were drawn. The fundamental equation of Eulers head was
modified by applying hydraulic characteristics of pump and turbine like hydraulic efficiency, slip
factor etc. and an equation for performance prediction of pump in turbine mode was derived. The
velocity triangles at outlet of pump and inlet of turbine are shown in Figure 2.2.

Figure 2.2. Velocity triangles at (a) impeller outlet in pump mode and (b) impeller inlet in turbine
mode (Fernandez et al., 2004).

Joshi et al. (2005) established the relationship between pump and turbine specific speeds (Eq.
2.3) for selection of PAT particularly for low head sites i.e. high specific speed pumps. The
constant speed and constant head characteristics for pump and PAT were plotted in terms of
normalized head, power and efficiency as a function of normalized discharge for an axial flow
pump using the data from Stepanoff (1957). To select the pump which can be used as turbine, the
value of specific speed in turbine mode can be worked out and based on that the pump specific
speed can be read from Nst-Nsp plot as shown in Figure 2.3. The approach has the advantage of
simplicity and generality but it was based on experimental data from only three pumps and the
effect of efficiency was not considered; hence, the method was recommended for the
approximate analysis.
20
 Nst (rpmbhp/ft54 )t
= (rpm (2.3)
Nsp usgpm/ft34 )p

Figure 2.3. Relationship between pump and turbine specific speeds for equal rotational speed
(Joshi et al., 2005a).

Isbasoiu et al. (2007) mentioned that selecting a PAT for a particular site must be made according
to specific flow and head requirements. Normally, the flow rate is determined by the minimum
flow rate and the head is determined by the vertical height between the intake and the turbine
outlet, after subtracting the head loss in the penstock. The chosen pump should have head and
flow at BEP as close as possible to the site conditions. Although the PAT efficiency will be
approximately same as in pump mode; at BEP, the head and discharge values are quite different
in pump and turbine modes. In order to be sure of the performance of PAT, it was recommended
to test it at various ranges of heads and flows. In addition, for the selection of the feed pump, it
was pointed out that the feed pump must be capable of producing a more head and flow than that
required at the BEP of PAT and it must have approximately four times the power rating than that
of PAT.

Pandey and Saini (2008) developed an Artificial Neural Network (ANN) based model using their
own experimental data to predict the performance of PAT. NNtool-box (MATLAB) software was
used for the performance analysis of PAT and curves were developed for the power output and
efficiency of PAT. The model was used to train, validate and test the data at different head and
flow rate ranging from 7-21 m and 24.5-50 lps respectively. Head, flow rate and power input
21
were taken as input parameters and power output and efficiency of PAT were considered as the
target data for the model. A comparison between the ANN output and actual experimental data
was made to validate the ANN model. It was emphasized that the proposed method may be
helpful for evaluating the performance of PAT at different heads and flow rates and hence ANN
model may be considered a good tool for selection of pump in turbine mode.

Singh and Nestmann (1994) presented an optimization routine for prediction (predicting turbine
mode characteristics of a pump) and selection (selecting the most appropriate pump for turbine-
mode operation) of radial flow centrifugal pumps in low specific speed range. The approach was
based on experimental results of nine model of PATs (Nsp: 2080 rpm) and the fundamentals of
applied turbo-machinery like the specific speedspecific diameter plot (called the Cordier/ Balje
plot), which was first introduced by Cordier (1953) and then pursued intensively by Balje (1981).
The method was experimentally evaluated for three pumps with specific speeds of 18.2 rpm, 19.7
rpm and 44.7 rpm, and the error in turbine performance prediction was found to be within 4%
in the full load operating range. It was reported that, the application of this model is quite simple
and can be incorporated in any computer program which can be made accessible to small
manufacturers and system planners particularly for energy recovery or micro hydro projects
(Singh and Nestmann, 2009).

Yang et al. (2012a) developed a theoretical method of predicting performance of PAT on the basis
of former research results, through theoretical analysis and empirical correlation which are given
below. The effects of variations of pump specific speed and pump maximum efficiency on h and
q were studied and observed that two pumps with same specific speeds may have different h and
q. In the next step, a centrifugal pump was simulated in direct and reverse modes using
commercial 3D Navier-Stokes computational fluid dynamics (CFD) code available in ANSYS-
CFX which has utilized a finite-element based finite-volume method for discretization of the
transport equations. The comparison of proposed method with other two methods viz. Stepanoff
(1957) and Sharma (1998), as shown in Figure 2.4, revealed that BEP characteristics predicted by
the proposed method and CFD were more accurate than the other two methods. The slight
difference between experimental and numerical results was found which may be attributed to the

22
negligence of leakage loss through balancing holes, mechanical loss caused by mechanical seal
and bearings and the surface roughness value set on the machines surface.

1.2
h= 1.1 (2.4)

1.2
= 0.55
(2.5)

Figure 2.4. (a) Head ratio and (b) discharge ratio of tested PATs with various pump maximum
efficiencies (Yang et al., 2012a).

Carravetta et al. (2014) developed a method called variable operating strategy (VOS) for the
optimum design of PAT working under different operating conditions. The characteristic curves
of seventeen different PATs rotating at different speeds were considered for the analysis. To
create hydraulic variability, hydraulic and electric regulations were used. The hydraulic
regulation system was consisted of series-parallel circuit with a PAT and two regulating valves.
Whereas, for the electric regulation the PAT generator was connected with an inverter to change
the rotational speed. It was mentioned that, based on the performance curve of a single PAT, all
the information needed for the application of VOS can be obtained by the application of affinity
law of turbomachines as mentioned in Eq. (2.6). The study revealed that, the performance curves
predicted by affinity law and Suter parameters led to15% error in the evaluation of the head drop
compared to experimental results.

1/2 1/2 3/4 1/2 5/4

=
( ) = ( ) ( ) = ( ) ( ) (2.6)

23
2.2.2 Experimental studies for selection criteria

Derakhshan and Nourbakhsh (2008b) predicted the BEP of centrifugal pump running as turbine
using theoretical analysis based on area ratio method developed by Williams (1992) and
Anderson (1993). The maximum efficiency of PAT was calculated as the ratio of net power
output from the turbine and the hydraulic power supplied at the inlet. The net power output was
worked out by subtracting various losses in the turbine (e.g. volute power losses, leakage losses,
kinetic energy losses at the outlet, hydraulic losses and mechanical losses) from the gross power.
A complete mini hydropower test rig established in the laboratory of University of Tehran, as
shown in Figure 2.5, was used for the experimental verification of theoretical results. At BEP, the
values of discharge number, head number, power number and efficiency predicted by theoretical
methods were found to be 1.1%, 4.7%, 5.25% and 2.1% lower than that of corresponding
experimental values. These deviations may be due to assumptions made in the evaluation of the
volute and the impeller losses. The equation of maximum efficiency of PAT was derived as:

P (Qt Ht );Pvt ;Plt ;Pet ;Pit ;Pmt

t = QntH = (2.7)
t t Qt Ht

Figure 2.5. Mini hydropower test rig established in University of Tehran (Derakhshan and
Nourbakhsh, 2008b).

24
Derakhshan and Nourbakhsh (2008a) derived relations to predict BEP of PATs based on the
experimental work done on four centrifugal pumps with different specific speeds of 14.6, 23,
37.6 and 55.6 (m, m3/s). It was found that, PAT works at higher head and flow rate than the pump
at the same rotational speed. However, the efficiencies were almost the same in both modes.
Pumps with higher specific speeds were subjected to the lower values of h and q. A new method
was derived to predict the BEP of PAT based on pumps hydraulic specifications, especially the
specific speed, which characterized the type of the runner and consequently its hydraulic
behavior. It was shown that, between two pumps (having same specific speeds) the pump with
bigger impeller works more efficiently. Also, the more efficient pump operates as turbine at larger
values of h and q. The predicted h and q by this method (Eq.s 2.8 and 2.9) were in good
coincidence with the experimental data. A procedure was presented to choose a proper centrifugal
PAT for a small hydro-site with Nst < 150 (m, kW).

hnew = h. (0.25/D)1/4 (2.8)

qnew = q. (0.25/D)1/6 (2.9)

Nautiyal et al. (2011) developed relations using experimental data of the tested pump (Ns = 18 m,
m3/s) and pumps of some previous researchers to obtain turbine mode characteristics from pump
mode characteristics. The experimental results were presented in the form of non-dimensional
parameters viz. head coefficient (), discharge coefficient () and power coefficient () which are
expressed below. Various parameters affecting the performance prediction of PAT were
enumerated as mentioned in Table 2.3. As compared to other methods, the deviation between
experimental results and results obtained by proposed relations were low which made the
performance prediction of PAT simpler and closer to accuracy. However, some uncertainties were
still found in prediction of turbine mode characteristics using pump operation data. The relation
between best efficiency and specific speed of pump () was developed by regression analysis and
equations of h and q were developed in terms of which are expressed below.

25
 gH
= n2 D2 (2.10)

Q
= nD3 (2.11)

P
= n3 D5 (2.12)

p ;0.212
= (2.13)
ln(NS )

q = 30.303 3.424 (2.14)

h = 41.667 5.042 (2.15)

Table 2.3 Factors affecting the predicted performance of pump (Nautiyal et al., 2011).
Parameters
Mechanical Considerations
Pump Liquid
System considerations
Factors
High horse power
High suction pressure
Operating speed
Operating temperature
Running clearances
Slurries
Abrasives
High viscosity
Dissolved gases
Net positive suction head
Suction and discharge piping arrangement
Shape of Head-Discharge curve
Run off conditions
Vibration and noise limits

26
Engeda (1987) and Florez and Jimenez (2008) performed a comparison of the dimensionless
coefficients (h and q) that relate the operating points of maximum performance in pump and
turbine modes and emphasized the necessity of more reliable and actual data for validation of
prediction methods. Garcia et al. (2012) compared and analyzed several prediction methods and
algorithms suggested by Stepanoff (1957), Sharma (1985), Alatorre-Frenk (1994) and
Derakhshan and Nourbakhsh (2008a). Figure 2.6 shows the relationship between the
dimensionless parameters of head and flow rate as pump and as turbine at the point of maximum
efficiency. The head and flow rate were found higher in PAT; however, the efficiency was found
almost equal in both modes.

Figure 2.6. Comparison of proposed method with other methods for (a) h and (b) q (Garcia et al.,
2012).

Agarwal (2012) reviewed the research carried out by some researchers and reported that
conversion factors for PAT can be decided on the basis of theoretical and numerical studies but
its performance can not be predicted accurately. The need for further research to develop a
general model for calculating the conversion factors was emphasized.

Many researchers have developed the charts for the selection of turbines (including PAT) in
mini/micro hydro range. The selection chart proposed by Engeda et al. (1988) (range: 1 to 1000
kW) based on head and discharge is shown in Figure 2.7. Williams (1995) also published similar
chart and reported that pumps can be used as turbine over the range normally covered by multi-
jet Pelton turbines, cross flow turbines and small Francis turbines. However, for high head-low
flow applications, a Pelton turbine is likely to be more efficient than a PAT and no more

27
expensive. The biggest advantage of PAT is for medium head sites, where practical and cost
advantages are in favor of pump instead of use of the other types of turbines. PAT requires a fixed
flow rate hence; it is suitable for sites where there is sufficient supply of water is available
throughout the year. Also, long term water storage is not generally an option for micro hydro
schemes because of the high cost of construction of a reservoir. The relations between maximum
efficiency in pump and turbine modes proposed by different researchers are given in Table 2.4.

Figure 2.7. Application range of different turbines (Engeda et al., 1988).

Table 2.4 Relation between maximum efficiency in pump and turbine modes.
Reference Relation between pump
and PAT BEP
Hancock (1963) PAT = p 2%
Childs (1962), Gantar (1988), Williams (1995), PAT p
Isbasoiu et al. (2007), Derakhshan and Nourbakhsh (2008a),
Adams and Parker (2011), Morros et al. (2011)
Chapallaz et al. (1992) PAT = p 3 to 5%
Nautiyal et al. (2011) PAT = p 8.53%

28
2.3 Cavitation Analysis of PAT

Cavitation is the phenomenon of formation of vapor pockets or cavities in the interior or on the
boundaries of a moving liquid and their subsequent condensation. It occurs in the turbomachinery
when the pressure drops, on account of the acceleration of water in the rotor, below the vapor
pressure of the water at prevailing temperature. The design, operation and refurbishment of
hydraulic turbines, pumps and PATs are strongly related to cavitation phenomena, which may
occur in either rotating or stationary parts of the machine (Avellan, 2004). It may be detected by
carrying out the analysis of structural vibrations, acoustic emissions and measurement of
hydrodynamic pressures in the machines (Escaler et al., 2006a). Few investigators have studied
the performance of PAT under cavitating conditions which is summarized in this section.

Gantar (1988) experimentally studied the performance of propeller pump in turbine mode under
cavitating conditions and reported that cavitation characteristics in turbine mode operation are
more favorable than in pump mode Therefore, the turbine mode operation requires smaller
submergence of the impeller than the pump operation.

Alatorre-Frenk (1994) performed number of experiments to study the effects of cavitation

number on the efficiency of PAT. It was found that for > 0.3, for most of the cases, T was in
the range of 79 0.5% with small dispersion due to the oscillations of the system. For = 0.25,
distinct increase in the efficiency of about 1.5% was observed, which was attributed to the
reduction in the skin friction because of the formation of a thin vapor layer on the surface of the
rotor. The critical value of cavitation number was found as 0.25 on account of generation of
instabilities in the test rig below this value. The Tcrit was found to be lying outside the expected
range of 0.16 - 0.21, which may be due to the difference in size and geometry between the PAT
and the conventional turbines as well as due to larger amount of dissolved and entrained air. The
results showed that the cavitation is slightly worse in PATs than in conventional turbines of
similar specific speed. In order to evaluate the crit values and to formulate a general cavitation
theory, it was advocated to carry out both destructive and non-destructive cavitation tests on more
PATs particularly for the large capacity pumps considering the high cost of these tests.
29
Singh (2005) discussed two criteria of analyzing the cavitation performance in PATs viz. one
based on the requirement criterion and the other based on the availability criterion. The
requirement criterion was machine dependent; whereas, the availability criterion depends on the
combination of system conditions like draft tube design, turbine settings and machine conditions
related to suction blade geometry. To understand the cavitation characteristics of PAT, a non-
dimensional Combined Suction Head Number (CSHN) was developed based on the combination
of availability criterion and the impeller properties. The CSHN analysis along with the suction
specific speed was used to predict the pressure near the suction eye of the PAT at different
operating speeds, draft tube designs and turbine settings. The minimum value of the CSHN,
representing the minimum pressure, was considered as the limiting design factor for the
cavitation. From the analysis, the minimum CSHN was found to occur after the BEP (i.e. in the
overload region) for all the radial flow PATs and before the BEP (i.e. in the part load region) for
mixed flow PAT. The proposed analysis along with Dixons criterion (Dixon, 1978), which is
based on the availability conditions, was recommended as the cost effective and useful
methodology to study the critical cavitation in PATs.

It was found that cavitation characteristics in turbine mode operation are more favorable than in
pump mode. Also, its effects may be more critical than that in conventional turbines having
similar specific speed. In order to evaluate the crit values and to develop the generalized
cavitation theory for PAT, it was recommended to carry out more detailed investigations,
destructive as well as non-destructive, in the wide range of specific speed.

2.4 Force Analysis of PAT

When centrifugal pump is operated as centripetal turbine, the point of maximum efficiency shifts
towards a higher head and flow rate than that in pump mode. Hence at BEP, the pump is
subjected to higher head and discharge in turbine mode as compared to that in pump mode
(Fernandez et al., 2004; Prasad et al., 2006; Rawal and Kshirsagar, 2007; Barrio et al., 2010).
Consequently, the flow instabilities and the radial unbalance generated in turbine mode are
30
expected to be different compared to that in pump mode. Many researchers have studied the
forces acting on pump in both modes.

Gantar (1988) carried out experimental investigations on propeller pump running as turbine and
the increase of axial thrust was found to be directly proportional to the head coefficient (). In the
region of over optimum flows, the axial thrust was observed to be 2 to 2.5 times greater than that
in pump mode operation. For the better performance of the machine, it was recommended to
control the load on the axial bearing and eventually, if required, use of large size bearing was
suggested. In addition, in turbine mode the torsion moment was found to be approximately 2 to 3
times greater than that in the pump mode. The variations of head, axial force and blade torque
coefficients with flow coefficient in pump and turbine modes are shown in Figure 2.8. The
performance of PAT under runaway condition was studied and found that the runaway speed was
twice the rotational speed at optimum position of the impeller blades and it was decreasing
rapidly with the closure of impeller blades.

Figure 2.8. Variations of (a) head (b) axial force and (c) blade torque coefficients against flow
coefficient (Gantar, 1988).
31
Barrio et al. (2010) used the numerical model to estimate the radial load on the impeller as a
function of flow rate in pump and turbine modes. The unsteady flow computations were applied
along the one blade passage and the resulting radial load was calculated by integration of the
instantaneous pressure and shear stress distribution on all the impeller surfaces in each of the
time steps. In pump mode, the radial load was found minimum near the design conditions while
in turbine mode, it was observed that the magnitude of the radial load was increasing with the
flow rate. Below turbine rated conditions, the magnitude of the total radial load (steady and
unsteady components) was found to be lower than the maximum total load in the pump mode. In
contrast, significant rise in load was observed above rated conditions. It was concluded that, the
mechanical design of the machine and shaft bearings must be carefully undertaken for the smooth
functioning of pump in turbine mode. The average magnitude of the radial load on the impeller in
pump and turbine modes at different flow rates is shown in Figure 2.9.

Figure 2.9. Average radial load on the impeller as a function of flow rate in (a) pump mode and
(b) turbine mode (Barrio et al., 2010).

Fernandez et al. (2010) estimated the radial force acting on the impeller of a centrifugal pump in
turbine mode by performing steady and unsteady numerical simulations. It was found that the
average radial force varied almost linearly with the flow rate in turbine mode and the force vector
rotated in the same direction as the impeller during the blade passage period. The locus of the
vector followed a quasi-elliptical trace with an increasing magnitude as flow rate was increased.
32
Additionally, the maximum force amplitude was obtained when the trailing edge of one of the
blades (pressure side) was located 3 downstream of the tip of the tongue. Analysis concluded
that, although previous investigations by the authors revealed that the average radial force in
reverse mode is usually lower than that in pump mode (Fernandez et al., 2004), the results of the
present work showed that the unsteady effects due to blade-tongue interaction can be of great
importance, especially at high flow rates, and thus should not be ignored if a pump operates as a
centripetal turbine. Figure 2.10 shows the areas of radial thrust concentration in pump and turbine
modes.

Figure 2.10. Locations of radial thrust concentration rate in (a) pump mode and (b) turbine mode
(Fernandez et al., 2010).

Morros et al. (2011) presented the numerical simulation of a radial flow centrifugal pump
operating as turbine and validated the results with corresponding experimental data. Due to the
peripheral restriction caused by the volute tongue, a non-axisymmetric flow distribution was
observed around the impeller entrance. It led to significant unsteady fluctuations over the blade
loadings and imbalanced axial thrusts over the machine bearings. It was shown that both axial
and radial forces are controlled, exhibiting values very similar to those found in the pumping
mode. It was emphasized that special care must be taken with the unsteady oscillations of these
forces, mainly when working beyond the nominal intervals, as it may produce severe dynamic

33
loads subjected to high levels of unsteadiness. These fluctuations, derived from a strong blade-
tongue interaction, can be as high as 25% of the steady component, constituting a serious risk of
fatigue failure of the mechanical components. Finally, a moderate use of PAT, limited to a few
hours and mainly working at nominal conditions, was recommended to satisfy all the
requirements of small hydraulic resources.

2.5 Loss Distribution in PAT

When pump is operating in turbine mode the direction of flow is reversed; hence, the pattern of
loss distribution may not be same as in the pump mode. To improve the performance of PAT, one
of the important factors is to identify the causes for losses that may occur in turbine mode. Few
investigators have studied the losses taking place in PAT based on theoretical and numerical
investigations.

Chapallaz et al. (1992) reported that the relationship between the pump and turbine mode is not
same for all types and sizes of pumps but depends on the flow pattern through the machine,
expressed by the specific speed, and the losses incurred, expressed by the efficiency of the
machine. As the fluid passes through the pump impeller, it is subjected to friction losses, shock
losses and leakage losses in both the modes. Due to these losses, the ideal energy transfer from
the rotating impeller to the fluid, as expressed by the Euler equation, is not achieved in pump
mode. The magnitude of these losses in the turbine mode is not exactly the same as in the pump
mode as shown in Figure 2.11. When PAT is operated at optimum flow conditions, an increased
pressure ought to act on the PAT which led to friction and shock losses on the fluid. In addition,
small quantities of fluid by-pass the impeller of the PAT and do not contribute to the energy
transfer. To maintain optimum head and flow conditions, the flow approaching the PAT must be
increased to compensate for the by-pass flow.

34
 Figure 2.11. Hydraulic losses in (a) pump mode and (b) turbine mode (Chapallaz et al., 1992).

Rawal and Kshirsagar (2007) mentioned that the remarkable achievement of the CFD model has
been the better understanding of the losses through the different stages of the flow. The numerical
simulation of PAT was carried out to determine the losses in the different components. The
regions of losses were identified as stationary casing, rotating impeller and the portion of the
draft tube up to the location of the measurement point. The hydraulic efficiency of PAT was
estimated by considering hydraulic losses in different flow passages. Table 2.5 shows the
distribution of losses in different domains at different discharge as obtained from CFD. In pump
mode, the losses in the impeller were relatively small compared to the losses in the casing;
however in case of PAT, the loss distribution was found to be reversed i.e. in casing the losses were
less compared to the impeller. It was revealed that, the draft tube losses mainly depend on the
swirl angle and cannot be controlled. The power estimated by CFD analysis was found to be
higher than the net power measured experimentally because the power losses in friction, bearing
and seals were not considered in numerical simulation. Also, the leakages through the stuffing box
and the recirculation behind the impeller were neglected.

35
 Table 2.5 Loss distribution in various domains of PAT (Rawal and Kshirsagar, 2007).
Parameters Discharge QQBEP
0.6 0.8 1.0 1.2 1.4
Losses in casing (m) 0.59 0.14 0.06 0.03 0.04
Net total head across the impeller (m) 3.61 7.83 12.49 16.90 22.32
Losses in impeller (m) 1.27 1.08 1.34 2.03 3.74
Losses in draft tube (m) 0.97 0.42 0.54 1.10 2.01
Total input head in turbine (m) 6.43 9.46 14.13 20.04 28.10

Derakhshan and Nourbakhsh (2008b) analytically determined the various losses occurring in the
pump running in turbine mode viz. volute power losses, leakage losses, kinetic energy losses at
the outlet, hydraulic losses and mechanical losses. The net power output from PAT was worked
out by subtracting these losses from the gross power.

Yang et al. (2012b) discussed the hydraulic loss distribution in the different zones of PAT viz.
volute, gap between volute and impeller, impeller and outlet pipe. The details of losses in three
PATs (specific speeds of 16.48, 34.44 and 46.15) are given in Table 2.6 and its variation with
discharge is shown in Figure 2.12. It was found that the hydraulic loss within the impeller makes
up the majority of the total hydraulic losses and its contribution grows with increasing specific
speed. The PATs highest efficiency was reached when the hydraulic loss within the impeller was
smallest. Analysis concluded that, the efficiency improvement of PAT should mainly be focused
on the optimization of the impeller. A detailed study of the hydraulic loss distribution illustrated
that the hydraulic losses in PAT decreases with the decrease of the blade wrap angle and the most
apparent decrease was observed in the impeller. The two main reasons for this decrease were
reported as reduction in the length of the impeller flow passage and velocity gradient within the
impeller.

36
 Table 2.6 Details of hydraulic loss distribution in PATs at their BEPs (Yang et al., 2012b).
hvolute hgap himpeller hpipe
Ns (m) Ratio (m) Ratio (m) Ratio (m) Ratio
16.48 2.74 25.75 1.25 11.7 5.52 51.82 1.14 10.72
34.44 1.10 12.88 1.36 15.86 5.95 69.47 0.15 1.79
46.15 0.71 13.68 0.46 8.79 3.88 74.21 0.17 3.32

Figure 2.12. Hydraulic loss distribution in the investigated PATs (Yang et al., 2012b).

2.6 Performance Improvement of PAT

The cost per kilowatt of the energy produced by small hydropower plants is usually higher than
that of large hydropower plants. The use of PATs in a wide range of small hydro sites has been
gaining importance worldwide in the recent years, but the subject of hydraulic optimization still
remained an open research problem. Conventional hydro turbines usually show higher efficiency
than a pump which is designed for the same operating conditions. On the other hand, the best
efficiency of a pump running in turbine mode is almost the same as that in the pump mode
(Chapallaz et al., 1992; Williams, 1995; Derakhshan and Nourbakhsh, 2008a). Therefore, the

37
application of reverse pumps in larger capacities is not economical. By increasing the overall
efficiency of reverse pumps, they can be economically applied in the higher power capacities
(Suarda et al., 2006; Derakhshan et al., 2009). Many investigators have attempted different
techniques to improve the performance of PAT. Lobanoff & Ross (1992) reported that, the
optimum efficiency of PAT can be generally achieved when the shock losses at the pump impeller
tips become nearly zero. Therefore, the modification must be made to the impeller tips of the
pump.

Singh et al. (2004) studied the effects of presence and absence of the rib situated in the suction
eye of the casing (which is provided to break the pre-swirl generated at the inlet of the pump in
pump mode) on swirling flow at PAT exit. It was mentioned that, if the rib is large it may lead to
localized acceleration of the flow, increased hydraulic resistance and breaking of the swirl at PAT
outlet. These parameters will change the exit velocity triangles and significantly influence the
overall characteristics namely pressure, torque and efficiency characteristics. From the
experimental analysis, it was concluded that the absence of casing eye rib, affects the loss
patterns of the rib-zone and the internal impeller which resulted in 1.3% rise in maximum
efficiency with a maximum operating efficiency over 82%. Also, the internal swirl study at the
draft tube entrance revealed that the absence of rib caused an increase of the rotational
momentum of the fluid only in the part-load and overload operating regions. In the rib-zone, the
numerical results supported the experimental findings but in the impeller zone the characteristics
were differed which can be improved by modifying the calculation methodology and review of
impeller conditions.

Singh (2005) emphasized that the issue of hydraulic optimization is the next stage of research
activity in PATs and should be treated at par with the topic of prediction model. Various
possibilities of modifying the pump geometry to improve the performance in turbine mode were
demonstrated viz. by rounding of inlet edges of impeller, by modifying the inlet casing rings, by
enlarging the suction eye, by removing the casing eye rib. It was revealed that, among the
different modifications proposed the modification at the periphery of the impeller blades by
rounding the impeller edges was the most beneficial. The modification was applied on eight
different centrifugal pumps and found that it has consistently improved the overall performance
38
of all the tested PATs with an efficiency rise of 1-2.5% in the BEP and overload regions. This
modification was first carried out by Lueneberg and Nelson (1985) and Cohrs (1997) on
individual pumps and both reported an efficiency improvement in the range of 1.52%.
Derakhshan et al. (2009) used a computer model to study the effects of impeller blade rounding
on a low specific speed pump.

Derakhshan et al. (2009) redesigned the shape of the blades using a gradient based optimization
method involving incomplete sensitivities for radial turbo machinery developed by Derakhshan et
al. (2008) to obtain higher efficiency. Optimization program was coupled to FINETURBO V.7,
to solve 3D incompressible NavierStokes equations and AUTOGRID 5 mesh generator
developed by Numeca software (http://www.numeca.be/index.php?id=numeca). In the next step,
the blades leading edges were rounded with a radius equal to half of the blade thickness such
that the overall diameter remains unaltered. And, the rounding of the outer and the inner sharp
edges of the hub and shroud were done. After each modification, a new impeller was
manufactured and tested in the test rig. The rounding of the impeller has reduced the net flow
separation loss at the blades and the impeller. However, it can also cause a rearrangement of inlet
velocity triangle which may change the shock loss component. These modifications led to rise in
maximum efficiency in the range of around 3-5% at rated discharge. The experimental results
confirmed the numerical efficiency improvement at all the measured points. Study revealed that,
the efficiency of the pump in reverse operation can be improved just by impeller modification.

Sedlar et al. (2009) studied the effects of diameter on radial-flow multistage pump working as
turbine. Two versions of the impeller were used: the first one with the full design diameter of
0.25 m and the second one with the diameter reduced (including the hub and shroud discs) to
0.22 m. The flow analysis was carried out at seven flow rates for several rotational speeds in both
modes; nevertheless the results for only 1000 rpm were presented. In case of turbine mode, the
flow rate was varied from 8.3 to 23.6 lps and the values of peak hydraulic eciency were found
to be 90.5% and 90.1% at optimal flow rates of 14.9 lps and 13.8 lps with full scale and smaller
impellers respectively. The comparison of flow patterns inside two impellers showed that, the
flow characteristics were similar in the stator but large global separations were observed inside
the impeller passages, which may be the reason for the poor performance with smaller impeller.
39
The effects of two impellers on head and hydraulic efficiency are shown in Figure 2.13. The
hydraulic design with the reduced impeller diameter showed that the optimal ow rate slightly
shifts towards the left but the hydraulic eciency decreases with this modication.

Figure 2.13. Comparison of results with full-scale and reduced size impeller (a) head Vs.
discharge and (b) efficiency Vs. discharge (Sedlar et al., 2009).

Singh and Nestmann (2011) studied the effects of impeller rounding on 9 PATs covering a
specific speed range of 20 to 94.4 rpm (8 radial and 1 mixed flow type). The details of
instruments used in experimental setup are given in Table 2.7. The radius of rounding was kept as
half the blade thickness at blade inlets and half the shroud thickness at shroud inlets respectively
and was kept constant for all the tested pumps. Due to these modifications, the system loss
coefficient, which is the main control variable, has consistently decreased whereas the exit
relative flow angle marginally increased for all the PATs. The decrease in system loss coefficient
led to drop of the net head but did not affect the net Euler momentum or shaft power. And, the
increase of exit relative flow angle caused increase in net Euler momentum as well as net head
across the PAT. The modifications resulted in reduced wake formation due to bullet shaped
rounding at the inlet of the blades and decreased flow separation at the entrance at shrouds as can
be seen in the front and side views of the impeller respectively in Figure 2.14. As a result, the
modification improved the efficiency for all PATs (within + 2% band) in the part-load, BEP and
overload regions. It was recommended to standardize the impeller rounding effects over wide
range of pumps, including axial pumps, and the extension of study for reversible pump-turbines.

40
 Table 2.7 Details of instruments used in experimental setup (Singh and Nestmann, 2011).
Variable Device Measurement principle Range Accuracy
Inlet head Pressure transducr Inductive + wheatstone bridge 0-2 bar 0.5% of full
(positive) scale
Exit head Pressure transducr Inductive + wheatstone bridge 0-1 bar 1% of full
(negative) scale
Discharge Magnetic flow meter Fradays magnetic law 0-200 0.1% of full
lps scale
Torque Torque sensor Wheatstone bridge 100 0.1% of full
Nm scale
Speed Speed sensor Optical counts 1 rpm

Figure 2.14. Effects of impeller rounding at (a) inlet of blade and (b) shrouds (Singh and
Nestmann, 2011).

41
Sheng et al. (2012a) applied splitter blade technique to improve the PAT performance which is
one of the techniques used in flow field optimization and performance enhancement of rotating
machinery (Keck and Sick, 2008; Yang and Kong, 2010). The negligence of fluid flow in a pump
front and back chambers was having only a minor effect on the performance of pump in pump
mode but same was not true in turbine mode (Derakhshan and Nourbakhsh, 2008b), hence the
flow domain was split into five components namely volute, impeller passage, front and back
chambers and outlet pipe as shown in Figure 2.15 (Yang et al., 2012b). Also, to get relatively
stable inlet and outlet flow, four times of pipe diameter was extended both at PAT inlet and outlet
sections. Results showed that, with the increase of splitter blades, its required pressure head was
dropped and efficiency was increased over the full operating range. Unsteady pressure field
analysis and comparison showed that the unsteady pressure field within PAT was improved when
splitter blades were added to the impeller flow passage. However, the variation in shaft power
with and without splitter blades was minor. Numerical predicted efficiency, pressure head and
shaft power were found to be higher than the experimental results which may be attributed to the
negligence of leakage loss through balancing holes and mechanical loss caused by mechanical
seal and bearings.

Figure 2.15. Different flow domains of PAT (Sheng et al., 2012a).

Yang et al. (2012b) numerically investigated three PATs covering low, medium and high specific
speeds with different blade wrap angles to study the influence of blade wrap angle on PAT. A
detailed hydraulic loss distribution and theoretical analysis were performed to investigate the
42
reasons for performance change caused by the blade wrap angle. The results showed that there is
an optimal blade wrap angle for a PAT to achieve the highest efficiency and it decreases with an
increase in specific speed. Also, the PATs flow versus head (Q-H) and flow versus shaft power
(Q-P) curves were dropped and the flow rate at the BEP was increased with the decrease of the
blade wrap angle. The hydraulic losses within the impeller decreased due to the shortened
impeller blade passage and reduced velocity gradient within the impeller flow channel. With the
decrease of the blade wrap angle, the slip factor of impeller was decreased; therefore, its
theoretical head was also decreased. The decrease in hydraulic losses within the impeller and
decrease in the theoretical head were found to be responsible for the poor performance of PAT.
The numerically predicted performance curves were found in agreement with the experimental
results as shown in Figure 2.16.

Figure 2.16. Comparison of CFD and experimental results (Yang et al., 2012b).

Sheng et al. (2012b) carried out an experimental research on single stage centrifugal pump with
original impeller (255 mm) and with two trimmed impellers (235 and 215 mm) in pump and
turbine modes. Figure 2.17 shows the performance curves of three tested impellers in both modes
and Table 2.8 lists their BEP parameters. It was found that, PATs efciency was decreased with
the trimming of impeller diameter. As impeller was trimmed from 255 mm to 215 mm, its
efciency at BEP was dropped by 4.11%, QH curve became increasingly steeper and QP curve
moved almost parallel down. Due to trimming of impeller, its geometric parameters viz. impeller
diameter, blade wrap angle, impeller width and blade angle at inlet were changed. The effects of
these parameters were studied using CFD code ANSYS-CFX which utilizes a nite-element
43
based nite-volume method to discretize the transport equations. The decrease of diameter has
shifted the ow rate at the BEP towards the left of its performance curve; whereas, the decrease
of blade wrap angle and increase of width and blade angles moved its ow rate at the BEP
towards the right. The overall performance change of PAT due to impeller trimming was the
combined effects of these four geometric parameters.

Figure 2.17. Performance curves of the three tested impellers (Sheng et al., 2012b).

Table 2.8 BEPs of impellers in pump and turbine modes (Sheng et al., 2012b).
Machine Impeller dia. (mm) Q (m3/h) H (m) Ps (kW) (%) Nsp
PAT 255 95.23 39.17 6.15 62.95 15.58
235 86.14 32.40 4.56 59.98 17.09
215 93.63 37.52 5.63 58.84 15.96
Pump 255 52.12 18.07 4.19 61.15 20.59
235 51.71 17.17 3.83 63.08 21.31
215 49.80 13.75 3.13 59.55 24.71

44
Yang et al. (2014) studied the effects of blade thickness on three different specific speed PATs (Ns
= 57, 119, 168) with numerical approach. Steady state simulations were carried out in Ansys-
CFX software with standard k- turbulence model. The advection scheme was set to high
resolution and standard wall function was used for the boundary layer treatment. All the wall
surface roughness within the control volume was set to 50 m. As boundary conditions static
pressure was specified at inlet and mass flow rate at outlet. With increase in blade thickness,
efficiency decreased but head and power increased. The increase in head may be attributed to the
blockage of impeller inlet area on account of increased blade thickness. It was also resulted in
rise in velocity in the impeller and hence increase in hydraulic losses. It was recommended to use
the thinner blades with sufficient strength to obtain higher efficiency.

Among the various techniques attempted by different researchers for performance improvement
of PAT, the impeller blade rounding was found to be most promising technique. However, it is
more advantageous with large capacity PATs in view of their higher efficiency compared to small
PATs. It is also required to standardize the impeller blade rounding effects over the wide range of
PATs.

2.7 Other Pumps as Turbines

Pumps operated in the reverse mode are usually of the single stage end suction centrifugal pump
with low or medium delivery head and the description of flow inside such pump-turbines can be
found in literature (Williams, 1994; Saini, 1998; Fernandez et al., 2004; Suarda et al., 2006;
Derakhshan and Nourbakhsh, 2008a; Barrio et al., 2010; Singh and Nestmann, 2010; Nautiyal et
al., 2011). It is also possible to use the multistage pumps in the turbine regime when sufficiently
high head is available. However, there is lack of information about applications of the multistage
pumps in the reverse mode, especially about the flow phenomena occurring inside these pumps
when operating as turbines (Sedlar et al., 2009).

45
Giddens et al. (1982) experimentally demonstrated the feasibility of operating the conventional
submersible pump and motor in the reverse mode as a complete, self-contained, water turbine and
induction generator unit. The pump was directly coupled to a single-phase induction motor and
when operated in reverse, with the necessary controls, generated an autonomous 220 V, 50 Hz,
single-phase supply with an overall system efficiency nearly the same as in the normal pumping
mode. The pumping and generating performance of the machine was compared and the principle
of the self-excited induction generator was discussed. Possibility of operating submersible PAT in
inclined position by housing it in a localized enlargement in the penstock, as indicated in Figure
2.18, was also discussed.

Gantar (1988) described that the propeller pumps which are usually used in irrigation, drainage
and in energy object can also be run as energy recovery turbines. The only problem in using these
PATs is the suction bell mouth, working as convergent section with small losses in pump mode,
which operate as diffuser subjected to higher losses in turbine mode. To utilize the kinetic energy
available at the PAT exit, the bell mouth was replaced by carefully designed diffuser with a bend.
The hill charts for pump and turbine mode operations were plotted and the region of good
efficiency in turbine mode was found to be larger than in the pump mode, which was due to
flexibility of regulating the impeller blades in propeller turbine. The hill chart for turbine mode
operation is shown in Figure 2.19. The maximum efficiency in turbine mode was found to be
approximately equal to that in pump mode; however, BEP parameters were different in two
modes. Analysis concluded that, the total cost per installed kW is higher for axial flow PATs in
comparison with radial flow PATs; nevertheless, in many cases they can be an efficient way of
utilizing the existing energy potentials.

Alatorre-Frenk (1994) reported that, in the domain of pumps where the impulse principle does
not work, very low specific speeds can only be achieved by using multistage pumps. They can be
used as turbines, but their cost advantage over impulse turbines may be offset by the high cost of
multistage pumps than the single-stage pumps and the low cost of small Pelton, Turgo and cross-
flow turbines, which can be fabricated in local workshops. Moreover, multistage pumps have
higher maintenance costs especially for bearings and seals than Pelton turbines (Torbin et al.,
1984).
46
 Figure 2.18. Submersible pump as turbine (Giddens et al., 1982).

Figure 2.19. Hill chart for propeller pump in turbine mode (Gantar, 1988).

47
The double suction centrifugal pumps are very commonly used when cavitation problems are
likely to arise (Kyung et al., 2002). They are usually suitable in applications where a high ow
rate is needed and in situations where the axial forces could establish a limitation for
conventional pumps. Gonzalez et al. (2009) studied the performance of double suction centrifugal
pump in turbine mode using a CFD technique based on the sliding mesh and the real movement
of the impeller. The computational domain is shown in Figure 2.20. The inlet tongue, designed
for the pumping mode, was considered as an obstacle at the outlet in turbine mode operation. In
spite of the difficulties induced by the geometrical complexity at the exit of the machine, due to
the double arrangement with a ow splitting effect, relatively high efficiency of 80% was
predicted in the flow coefficient range of 0.3 to 0.4. From the observed ow patterns in turbine
mode, it was concluded that the proposed geometry would perform quite well for the nominal and
lower ow rates. However, the performance was found to be poor at higher flow rates. Also, the
main drawback of the machine was found to be the possible increase in the radial forces or the
breaking of the axial balance with increasing ow rate.

Sedlar et al. (2009) studied the flow phenomena occurring inside the radial-flow multistage pump
operating in the turbine regime using CFD approach. The numerical analysis was concentrated on
the flow in the middle stage of the pump which consisted of the impeller with six blades and the
stator with eight channels. The fully unsteady Reynolds-averaged NavierStokes (RANS)
equations together with the Menters shear stress transport (SST) turbulence model were solved
on block-structured grid consisted of approximately 1.4 million nodes. As boundary conditions,
mass flow rate and the flow direction were prescribed at the end of the vane less diffuser and the
average static pressure was set at the exit of the straight piping. Two complete stages were
analyzed to avoid the influence of boundary conditions on the results and between the rotor and
stator six interfaces, three rotor-stator interfaces and three stator-rotor ones, were applied. It was
reported that, the hydraulic efficiency of the multistage pump in the reverse mode can be quite
high even without any expensive corrections of the manufactured parts. At design condition, the
flow was without large separations; but at off-design conditions, highly curved part of stator
channel as well as the region of guide vane-rotor blades interaction were found to be critical. The
streamlines in the impeller plane in turbine mode are shown in Figure 2.21.

48
 Figure 2.20. Computational domain for double suction centrifugal PAT (Gonzalez et al., 2009).

Figure 2.21. Streamlines in the impeller plane in turbine mode (Sedlar et al., 2009).

Bozorgi et al. (2013) simulated an industrial axial flow pump in the reverse mode using RANS
equations coupled with Spalart-Allmaras turbulence model in Numeca software
(http://www.numeca.be/index.php?id=numeca). To validate the numerical results, a test setup for
measuring the head, ow rate and power was built and the pump was tested in the reverse mode.
The experimental setup comprised of the axial PAT, generator, torque meter, electronic load
controller, feed pump, electric motor, venture, buttery valve, barometers, several pipes and
connections; which all were installed on the reservoir. The results showed that the pump can
work in a wide range of operation with negligible changes in the efciency as shown in Table 2.9.
A rst-order uncertainty analysis was done using constant odds combination method, based on a
95% condence level as explained by Moffat (1982). The uncertainties for the head, ow rate and
power measurements were 3.5%, 5.4%, 4.1%, respectively. A good coincidence between the
numerical and experimental results was observed. Study revealed that, the tested axial pump can
49
properly work as turbine and is a good alternative for generating electricity from low-head pico
hydro potentials, especially in the developing countries, where proper turbines are not easily
accessible.

Table 2.9 The range of good operation of axial PAT (Bozorgi et al., 2013).
Approach (%)
CFD 0.407-0.485 1.815-3.623 0.433-1.044 58-62
Experiment 0.409-0.485 1.961-3.779 0.454-1.059 58-61

2.8 Numerical Investigations on PAT

The flow inside turbomachinery is highly complex mainly due to 3D flow structures involving
turbulence, secondary flows, cavitation and unsteadiness. Traditional approaches for performance
prediction were based on theoretical and experimental studies and usually conclude with the
proposal of semi-empirical correlations to predict the performance characteristics in reverse
mode from those obtained in pump mode. This has been followed for over the years but none of
the method is 100% reliable to predict accurate turbine performance in the entire range of
specific speed and for the different types of pumps manufactured worldwide. In the recent years,
CFD started to play a key role for the prediction of the flow through pumps and turbines having
successfully contributed to the enhancement of their design (Croba and Kueny, 1996).

The applications of CFD in the design and analysis of hydro turbines and pumps started about 30
years ago. The first step coincided with the introduction of the finite element method (FEM) into
CFD. In late 90s, numerical codes started to evolve from a pure inviscid physical model
assumption into realistic viscous turbulent models. Due to high computer power involved, these
models were only applied to study 2D or Quasi-3D Euler solutions. Over the years, the
complexity continuously increased in stages: via 3D Euler solutions, to steady RANS simulations
of single blade passage using finite volume method (FVM), extending to steady simulations of
whole machines, until today unsteady RANS equations are solved with advanced turbulence
50
models (Keck and Sick, 2008). In addition, a growing availability of computer power and a
progress in accuracy of numerical methods, brought turbomachinery CFD methods from pure
research work into the competitive industrial markets (Pascoa et al., 2009). The list of softwares
available in the market for numerical analysis of turbomachines is given in Table 2.10
(http://www.webinfobits.com/page/f3a9a2a9408b).

Table 2.10 Softwares for numerical analysis of turbomachines

(http://www.blog.kubiak.co.uk/post/81).
Name of software Country
OpenFOAM UK
ANSYS CFX US
ANSYS FLUENT US
COMSOL Multiphysics Sweden
STAR-CD UK
FLOW3D US
AxSTREAM US

CFD is a promising technique that has been used to accurately predict pump and turbine
performance directly from geometric data and fluid properties. It has been widely used in the
flow field visualization and optimization of rotating machinery (Sick and Wilson, 2005; Keck et
al., 2007; Keck and Sick, 2008). It has the advantage of not requiring large scale hydraulic test
facilities which is particularly advantageous for the high flow rates required for high specific
speeds (Joshi et al., 2005a). Yet, it is to be noted that, for useful and correct interpretation of the
CFD computed flow fields; a deep knowledge of the physics underlying the flow inside
turbomachinery is necessary. Moreover, due to the empirical nature of turbulence models applied
in these computations, it is of paramount importance to compare the computed results with
experimental results in order to achieve good pump designs (Tan et al., 2010).

Tamm et al. (2000) carried out an analysis of a reverse operating pump using CFD tools, but had
no experimental data to verify their results. They recommended further calibration of a CFD
results with sound experimental results.
51
Natanasabapathi and Kshirsagar (2004) carried out numerical investigations on PAT using Pro-E
solid modeler as modeling software and CFX for the simulation. 3D computational model of PAT
is shown in Figure 2.22. Across the interface between stationary and rotating components,
multiple reference frame (MRF) using frozen rotor interface were defined. Initially, the
simulations were carried out with unstructured mesh in which, head drop across the turbine was
matched with the experimental values but the deviations were observed in the efficiency at
discharges away from BEP. Also, the unstructured grid led to some unrealistic results like gain of
total pressure, where in reality loss was expected. Then the analysis was carried out with further
refinement of the mesh but the results were not encouraging as the error across the frozen rotor
interface remained same. Finally, two rings of structured grid were introduced in between the
casing and the runner which were connected by a frozen rotor interface. The results were
encouraging and the predicted performance showed a good trend with the experimental test
results in terms of quality and quantity.

Figure 2.22. 3D computational model of PAT (Natanasabapathi and Kshirsagar, 2004).

Rawal and Kshirsagar (2007) carried out experimental and numerical investigations on a mixed
flow pump in pump and turbine modes. The entire geometry consisted of casing, impeller and
draft tube was modeled and an unstructured tetrahedral mesh was applied as it has the advantages
of increasing the grid density at the locations subjected to high gradients of flow variables. In
pump mode, the best efficiency flow rate was 0.100 m3/s at a head and speed of 8.3 m and 1450

52
rpm respectively. The numerical model of PAT showed very good characteristics with a
maximum operating efficiency of 83.10% at a flow rate of 0.127 m3/s at 12.48 m head while
operating at the same speed. The value of peak efficiency was different at different operating
speeds but occurred at a constant Q/ND3 of 0.4. The maximum efficiency at 800 rpm, 900 rpm
and 1000 rpm was found to be 81.4%, 82.7% and 83.3% respectively. The similarity between
experimental and numerical results was satisfactory at flow near BEP flow and above. However,
higher deviations were observed at lower flow rates; to minimize the same, use of improved mesh
quality, numerical schemes and turbulence models was suggested. In addition, it was
recommended to focus on the design of the impeller and the impeller/casing interface for further
optimization of PAT.

Shukla (2008) carried out CFD analysis of flow field inside the different components of
centrifugal PAT using Fluents MRF model. Flow field inside the volute casing, impeller and
draft tube was analyzed at BEP and other values of discharge. CFD results demonstrated
satisfactory agreement with manufacturers data in a portion of operating range of pump, but
beyond this range CFD results showed large deviations. BEP of PAT was obtained at higher
values of head and discharge than that of pump. The cavitation analysis showed that the chances
of cavitation are minimum at the BEP discharge and increases with the increase of discharge
beyond BEP. Correction factors for head and discharge were determined based on the results
obtained by CFD technique and compared with the available relation of other studies and found
in good agreement. In casing and draft tube, hydraulic losses were observed to be minimum at
BEP as compared to other values of discharge.

Derakhshan and Nourbakhsh (2008b) simulated a centrifugal pump (Ns: 23.5 (m, m3/s)) in direct
and reverse modes. To verify numerical results, simulated pump was tested as a turbine in the test
rig. CFD results were found in good coincidence with experimental data for pump mode not only
at BEP but also in part-load and over-load zones. But numerical results were not in acceptable
coincidence with experimental data for turbine mode. Turbine head and power values achieved
by CFD were found to be lower than that of experimental data at same discharges. The only
difference between direct and reverse modes was in flow and in rotating directions. The
difference between real geometry and CFD model was in interaction between volute and
53
impeller, since flow zone in the space between impeller hub/shroud and casing was not included
in the model. The effect of geometric simplification was found to be greater in turbine mode,
because its effect on downstream flow was more than that on upstream flow.

Nautiyal et al. (2010) reviewed the work carried out on the pump running in turbine mode using
CFD as a numerical simulation tool and mentioned that CFD is a recent attempt for predicting the
performance of PAT. The difficulties associated in grid generation in complex region and its
effects on solution convergence were emphasized. It was reported that, the computational
analysis is very useful in identifying the losses in turbomachine components like draft tube,
impeller and casing. The CFD and experimental results did not match accurately in turbine mode,
but it was pointed out that the difference can be minimized through improvement in
computational analysis by using finer mesh, numerical methods and turbulence models. Also, the
future work in the field of computational analysis can further improve the prediction of pumps in
reverse operation.

Silva et al. (2010) presented a full CFD pump computation in pump and turbine modes using an
industrial pump as a benchmark test case. In order to obtain a good quality mesh, grid composed
of 12 geometrical blocks was used after performing the detailed aspect-ratio and equi-angle
adjustments on the mesh domain. The characteristics of three commonly applied approaches for
CFD analysis viz. frozen-rotor, mixing-plane and pure unsteady computation were discussed.
Considering the existence of a small gap between rotor and tongue the flow was simulated by
applying a frozen-rotor approach based on pseudo-unsteady approach in the blade passage, in
front of the tongue and the overall turbomachine flow field. It was pointed out that Spalart-
Allmaras model, which is a low-Re turbulence closure, can be used to compute wall bounded
flows; however it requires high node density in the near-wall region. Numerical results were
compared with the experimental results and the discrepancies were found to be within the
expected region for a frozen rotor computation. It was concluded that, the use of PAT allows to
obtain a turbine, for diverse ranges of operation, at a fraction of the economic cost of designing a
turbine from scratch.

54
Fernandez et al. (2010) carried out 3D flow simulation of a centrifugal pump operating as a
centripetal turbine with the commercial code Fluent. The unsteady flow calculations together
with a sliding mesh technique was applied which made it possible to account for the effect of
bladetongue interactions on the local flow. The comparison between the numerical predictions
and the previously collected experimental data is shown in Figure 2.23. A maximum relative error
of 9% was obtained for the total head and about 5% for the average static pressure around the
periphery of the impeller. The computational model overestimated the magnitude of the
efficiency and the generated power. This may be attributed to the fact that, the lateral space
between the impeller and the casing walls was not included in the model due to which volumetric
and disc friction losses were cancelled during the simulations. The numerical calculations also
showed the existence of a tangential velocity component at the exit of the impeller for all the test
flow rates. This component induced a fluid rotation in the same direction as the impeller for low
flow rates and in the opposite direction for high flow rates.

Figure 2.23. Performance characteristics of PAT at 1750 rpm (Fernandez et al., 2010).
(lines: experimental data; icons: numerical predictions)

Barrio et al. (2011) carried out internal flow analysis of a centrifugal pump having seven twisted
backward blades in pump and turbine modes using the commercial code Fluent. The simulations
were carried out by solving the full unsteady RANS equations with the finite volume method for
the flow rates between 20-160% after testing the dependence of the numerical predictions on grid

55
and time step size. Turbulent effects were incorporated by means of the renormalization group
(RNG) k model together with standard wall functions to calculate the flow variables near solid
boundaries. The numerical results were compared with the experimental measurements which
showed typical differences in the range 3-5%. In turbine mode, internal recirculation in the
impeller was found for low and high flow rates. The patterns of the flow at the inlet of the
impeller showed large regions of backflow in all the impeller passageways. Backflow zones were
also observed at 140% flow rate only in the four passageways nearest to the tongue. Figure 2.24
show contours of static pressure and velocity vectors near the tongue region in turbine mode at
different discharge.

Figure 2.24. Contours of static pressure and velocity vectors near the tongue region in turbine
mode at (a) 100% and (b) 140% discharge (Barrio et al., 2011).

Morros et al. (2011) presented a numerical simulation of a pump operating as a turbine and
validated the results through experimental data. The fundamental equations viz. mass
conservation, momentum conservation and two equation turbulence models (standard k-) were
discussed and used. The second order implicit scheme was introduced for the time dependent
terms. The study was focused on the analysis of the basic characteristics of the flow providing
insight into the unsteady features of the energy transfer. A non-axisymmetric distortion was
observed in the flow at the entry of the impeller due to the peripheral restriction caused by the
volute tongue. It was demonstrated that PAT can provide acceptable efficiencies, lower than in
the design configuration, but high enough to offset the expensive manufacturing costs of
specifically designed hydraulic turbines with respect to the reduced prices of commercial pumps.

56
Both absolute and relative reference frames have been considered in order to present the impact
of the tongue disturbances on the circumferential blade-to-blade gradients within the impeller
passages.

Li (2015) carried out steady flow simulations to study effects of viscosity on low specific speed
centrifugal pump in pump and turbine modes. The flow was analyzed by examining the velocity
magnitude, flow angles, angle of attack, velocity vectors in the volute, swirling patterns of flow
in the draft tube and deviation angle at impeller outlet. The external performance parameters and
internal variables viz. impeller theoretical head correction factors in both modes were related to
impeller Reynolds number very well by using a three-parameter-power function. The effects of
viscosity on the head and hydraulic efficiency in both modes are shown in Figure 2.25. It was
found that the flow rate, total head and hydraulic efficiency of the turbine increased with increase
in viscosity at BEP. Whereas, the output shaft power was mainly affected by the density of liquid
than the viscosity. Hence, more power was generated in turbine mode with water than with any
viscous oils. The flow separation and incidence losses in the impeller along with increased wall
shear stresses were responsible for the poor hydraulic efficiency in turbine regime compared to
pump mode. It was concluded that the liquid viscosity is having more adverse effect in turbine
mode that that in pump mode.

Figure 2.25. (a) Head Vs. flow rate and (b) hydraulic efficiency Vs. flow rate at different
viscosities in pump and turbine modes (Li, 2015).
57
Several numerical studies on PAT have been reported in the literature, which has proven that
URANS equations can provide a reasonable approximation of the general performance of the
machine from an engineering point of view with errors typically less than 10% with the
experimental data. The reason for such error is difficulties in creation of numerical model; as well
as, numerical simulation predicts only the hydraulic performance; whereas, experimental results
include mechanical and volumetric losses along with hydraulic losses. The discrepancies between
the experimental data and numerical simulation results can be further minimized through
improvement in the CFD simulation by use of finer mesh along with advanced numerical
schemes and turbulence models. More experience will be needed to realize accurate convergence
of CFD results with experimental data. The details of different computational parameters used by
various researchers for CFD analysis of PAT are summarized in Table 2.11.

Table 2.11 Comparison of parameters used by various researchers for CFD analysis of PAT.
Authors Turbulence model Pressure- Boundary conditions
Velocity inlet outlet
coupling
Rawal and Kshirsagar (2007) Standard k- - total pressure mass flow rate
Derakhshan et al. (2009) Standard k- - mass flow rate static pressure
Gonzalez et al. (2009) Standard k- SIMPLEC total pressure static pressure
Barrio et al. (2010) Standard k- SIMPLE total pressure static pressure
Fernandez et al. (2010) Standard k- SIMPLE velocity inlet static pressure
Silva et al. (2010) Spalart - Allmaras PISO total pressure static pressure
Barrio et al. (2011) RNG k SIMPLE total pressure static pressure
Morros et al. (2011) Standard k- SIMPLEC static pressure total pressure
Shi et al. (2012) RNG k SIMPLEC velocity inlet static pressure
Yang et al. (2014) Standard k- - static pressure mass flow rate
Li (2015) Standard k- SIMPLE velocity inlet static pressure

58
2.9 Case Studies on PAT

The performance of any hydropower system in the field may differ from that of the performance
studied under laboratory conditions; due to various reasons like generation of artificial head, use
of silt and debris free water, scaling effects etc. Hence, it is utmost important to ascertain the
behavior of PAT in the field. In this section, few case studies related to PAT based micro
hydropower plants installed worldwide are presented.

Maher et al. (2003) revealed that, there exist a hydropower potential of at least 3 MW in Kenya
(for schemes up to 5 kW) which was sufficient to supply basic electricity to 1,50,000 households
in the rural areas of Kenya. In view of this, in December 2001, a community-owned 2.2 kW pico
hydro scheme was commissioned at Thima in Kenya for electrification of 110 low-income rural
households. The funding agency (European Commission) met the costs of the penstock pipe and
generating equipments; whereas a substantial proportion of the total cost including building
materials, distribution cables and house wiring components was met by the consumers. A
standard mono-block centrifugal pump was used as a turbine and its motor was used as a
generator. The pump impeller was turned down on a lathe to achieve a better operating efficiency.
The main specifications of the project are given in Table 2.12. It was revealed that, in addition to
funding further grid extension, the rural electrification levy should be used to encourage the
development of pico hydro and solar home systems particularly for low-income households.

Table 2.12 Specifications of pico hydro demonstration schemes (Maher et al., 2003).
Parameter Scheme A: Kathamba Scheme B: Thima
Power output (electrical) 1.1 kW 2.2 kW
Number of houses connected 65 110
Type of turbine Pelton (200 mm pcd) PAT (monoblock centrifugal)
Head (net) 28 m 18 m
Flow (design) 8.4 lps 28 lps
Efficiency (turbine & generator) 48% 45%
Distance of furthest house 550 m 800 m
59
In order to promote small industries in rural communities of Tanzania, United Nations Industrial
Development Organization (UNIDO), Environmental Tectonics Corporation (ETC) Netherlands
and Tanzania Traditional Energy Development Organization (TaTEDO) installed a PAT based
micro hydropower project of 10 kW capacity in Kinko village of Lushoto District in Tanzania in
July 2006 (UNIDO Project Report, 2010). The produced electricity was supplied to 100 houses in
the village to assist in rural development and poverty reduction through improvement of existing
and introduction of new livelihood strategies. The maximum PAT efficiency and an overall water
to wire efficiency were found to be 80% and 64% respectively. Singh et al. (2006) evaluated the
field performance of PAT unit for this plant and compared the predicted and field hydraulic
characteristics with uncertainty bands for field trials, as shown in Figure 2.26. The maximum
deviation for the net head, discharge and power output were found to be within 2-4% only.

A 3-kW micro-hydropower project was installed in Mae Wei village of Tha Song Yang District in
Thailand (http://palangthai.blogspot.in/2008/02/3-kw-pump-as-turbine-microhydro-at-mae.html)
which uses a centrifugal pump running backwards as a turbine. It was installed over 10 days in
February 2008 by a team of villagers from Mae Wei, Engineering Studies Program (ESP)
students from the Mae La refugee camp, American students from the Institute for Village Studies
and Spring Street High School, Border Green Energy Team (BGET) members and Palang Thai.
The gross head available at the site was 35 m and water was supplied to the PAT using 4
diameter PVC pipe. The pump's induction motor was used as generator to generate 3-phase 240-
volt (delta) electricity which was converted into single-phase 240 volt using a capacitor
arrangement. A micro-hydro controller was used to dump excess electricity to a resistive ballast
load. The generated electricity was utilized to light 16 classrooms, a dormitory and teacher's
homes, 10 computers and video equipment.

The PAT has been installed worldwide in many countries for power generation, in water supply
pipe lines and in reverse osmosis systems. The details of some typical PAT installations in the
world are given in Table 2.13.

60
 Figure 2.26. Comparison of predicted and field hydraulic characteristics (Singh et al., 2006).

Table 2.13 Worldwide installations of PAT.

Location Capacity of Plant Year of
Installation
Laos, Xiagnabouli province, South East Asia 2 kW 2008
(Arriaga, 2010)
Thima, Kenya (Maher and Smith, 2003) 2.2 kW 2001
Mae Wei village, Thailand* (2008) 3 kW 2008
Barnacre, North-west of England (Williams, 1996) 3.5 kW 1996
West Java, Indonesia (MHP Project Report, 1995) 4.5 kW 1992
Kinko, Lushoto, Tanzania (Singh et al., 2006) 10 kW 2006
Ambootia, Darjeeling, India (Singh, 2005) 50 kW 2004
British Columbia, Canada$ 200 kW -
#
Breech, Germany 300 kW* (8 units) 2006
* http://palangthai.blogspot.in/2008/02/3-kw-pump-as-turbine-microhydro-at-mae.html
$ http://www.smallhydropower.com/pumps.htm

# http://www.ksb.com/linkableblob/ksb-en/111086-356644/data/Pumpe_als_Turbineen-data.pdf

61
2.10 Other Applications of PAT

The need for saving energy in water supply systems has become one of the main concerns of
system managers and it will become more important in a near future. New strategies must be
developed and implemented in the major energy consumption systems like those for water
supply. In drinking pipe systems, the pressure reducing valves (PRVs) are commonly used as an
energy dissipating device by localized head loss to maintain the uniform pressure. The use of
PATs in place of PRVs seems to be an alternative technical and environmental available solution
to control the pressure as well as to produce energy. PAT could be a convenient choice, but a deep
study of the machine in different operating conditions is necessary in order to prevent the water
system from ruptures (Fecarotta et al., 2011).

Gulich (2007) presented the design of centrifugal pump and its performance in turbine mode
using mathematical equations. The use of PAT was recommended for energy recuperation in
processes where a large amount of fluid energy is dissipated in valves or other throttle devices.
The forces exerting in different components during turbine mode were discussed. Theoretical and
actual characteristics of a reverse running centrifugal pumps as turbines were illustrated.

Garcia et al. (2012) mentioned that in many case the main pipes in water distribution systems
have an excess of static pressure which is usually dissipated by means of intermediate reservoirs,
pressure-reducing valves or any other device that produces the required energy loss. This
hydraulic energy can be used directly to drive a mechanical system or to generate electric power
up to a capacity of 100 kW where use of PAT was recommended. Two cases pertaining to PAT
based water system distribution of Murcia and Elche (in Spain) were considered to compare the
prediction methods given by Williams (1995), Fernandez et al. (2004) and Derakhshan and
Nourbakhsh (2008a) and Audisio (http://www.scribd.com/doc/49218247/BOMBAS-
UTILIZADAS-COMO-TURBINAS). The results obtained in two cases under study were found
to be very similar and the resulting dimensionless coefficients between points of maximum
efficiency were in agreement in both modes with those experimentally obtained by other authors.
It was emphasized that the large-scale deployment of PAT is economically viable and acceptable

62
pay-backs can be achieved, which represents an investment opportunity for the companies
responsible for water network supply.

Fecarotta et al. (2011) studied the performance of multi-stage pump in a water supply system
numerically and experimentally and found that the response of PAT could be quite different from
the response of a PRV. In case of PRV, the pressure downstream the valve can be regulated by the
correct setting of the valve opening, but for a PAT (which is normally not equipped with the guide
vanes) a discharge variation causes a head-loss variation and the pressure waves pass through the
runner inducing high efforts and tensions. The difference in dynamic behavior of PAT and PRV
for a fast closure of a pipe downstream valve is shown in Figure 2.27 (Ramos et al., 2004). It was
emphasized that, under some abnormal conditions PAT may be subjected to water hammer
situations, which may lead to ruptures of the pipelines and loss in efficiency; hence more detailed
analysis was recommended under such conditions. The most severe hydro transients occur during
extreme operating conditions, such as full-load rejection and turbine stoppages, particularly in
hydraulic systems with high head or long pipeline length (Ramos and Almeida, 2002). Ramos
and Almeida (2001) studied the behavior of a turbine equipped with a guide vane under runaway
conditions and the effects of this behavior on a penstock of a small hydropower plant with a long
penstock.

Carravetta et al. (2012) reported that, water distribution networks (WDNs) allow to obtain
significant amount of energy generation by exploiting the head drop due to the network pressure
control strategy for leak reductions. Based on variable operating strategy (VOS), the design
methodology was proposed to predict PAT behavior and to find the optimal solution in the
variable-operating conditions of hydro-plant scheme. The proposed design procedure was applied
on two different machines viz. A (centrifugal, single stage, NsT = 44.0 [rpm kW1/2 m4/3]) and B
(semi-axial, 4 stages, NsT = 35.9 [rpm kW1/2 m4/3]). VOS design led to maximum efficiency of
the power plant and a continuous pressure regulation for a PAT installed in a series-parallel
configuration as shown in Figure 2.28. Moreover, VOS along with CFD analysis, provided very
accurate design solutions. Analysis concluded that, in the absence of a complete set of PAT
performance curves provided by the manufacturer, CFD can be considered as valid alternative to
experiments.
63
 Figure 2.27. Dynamic behavior of (a) PRV and (b) PAT: for a fast closure of pipe downstream
valve (Ramos et al., 2004).

Figure 2.28. PAT installation scheme in water distribution network (Carravetta et al., 2012).

PATs can also be used as an energy recovery devices (ERDs) in reverse osmosis (RO) systems to
recover hydraulic energy from the brine stream for various applications viz. to raise the pressure
in the feed stream (hydraulic pressure boosters and turbochargers); to pump a side stream of feed
water (pressure exchangers); and to reduce the load on the motor driving the high pressure feed
pump. All these approaches have achieved commercial success with hundreds of examples of
each type on seawater reverse osmosis (SWRO) systems. However, very few ERDs have been
installed so far on brackish water reverse osmosis (BWRO) systems, though BWRO systems
appear to greatly outnumber SWRO systems and have a vastly higher aggregate capacity
(Oklejas et al., 2007).

Raja and Piazza (1981) discussed the applications of reverse running centrifugal pumps as
hydraulic power recovery turbines (HPRTs) to recover the energy from the waste brine. The use

64
of multistage split casing centrifugal pump was recommended as HPRT, in comparison with the
variable inlet vane turbines and impulse turbines, in view of its simple and cost-effective control
systems. The schematic diagram of PAT based SWRO system is shown in Figure 2.29. About
70% of the total pumped fluid was passed through the HPRT as waste brine, by closing the
bypass valve, in the pressure range of 800-1000 psi. The system may recover up to 80% of the
wasted energy, which is used to take up the partial load of the motor to reduce the input power
requirement of the motor. For the cost analysis, different sizes of similar HPRT designs were
considered with SS-316 material of construction, an overrunning clutch and a bypass valve
excluding the installation cost. It was concluded that depending on the plant size, power cost per
kWh and payout periods, the use of HPRTs can considerably reduce the power consumption per
unit volume of product water thereby increasing the overall plant efficiency. Also, as the size of
plant increases, the HPRTs become more economical and offer considerable reduction in payout
periods.

Figure 2.29. Schematic diagram of PAT based SWRO system (Raja and Piazza, 1981).

65
PATs also found applications in many fields other than hydropower plants, water piping systems
and RO systems viz. in oil and chemical industries (for the decompression processes in gas/liquid
separation), in washing/clearing chemical media (for hydrocarbon synthesis by means of hydro-
cracking technique), in pressure pipelines in heating, fuel supply and reclamation systems
(https://www.google.co.in/url?sa.....). Gantar (1988) described the possibility of using the high
specific speed pumps in turbine mode on smaller rivers as well as in low head applications like in
the irrigation system with gravitational pipe net, in cooling water system of the thermo-energy
and industrial plants etc. However, the information pertaining to such applications is limited and
more detailed investigations are required for the implementation of PATs for such applications.

2.11 Cost Analysis of PAT

The efficiency of PATs is lower than the conventional hydro turbines but their applications are
recommended in view of their lower initial and maintenance cost. The initial cost of the machine
affects the cost of the hydropower plants only in initial phase of the project; however, the lower
efficiency of the machine affects the plant on daily basis. Hence, to justify the use of PAT in
mini/micro hydropower plants, it is required to carry out the detailed cost analysis. Few
researches have carried out cost analysis of micro hydropower plants by considering both the
options as prime movers; i.e. conventional hydro turbine as well as an equivalent PAT.

Alatorre-Frenk (1994) carried out the cost analysis of micro hydropower plant based on PAT and
conventional hydro turbine. The cost of the scheme was considered as the sum of the initial
investments in turbomachinery, penstock, storage reservoir and miscellaneous items. These costs
were considered as a function of the size of the scheme, which were further depending on the
rated flow, and the lifetime was assumed to be same for both the cases. To normalize the flow, the
rated flow was divided by the annual average flow. The yearly operation and maintenance costs
were considered as a fixed proportion of the initial investment for all the cases. The
characteristics of three most frequently used financial parameters for evaluation and comparison
of project investments viz. the net present value (NPV), the benefit/cost ratio (BCR) and the
66
 internal rate of return (IRR) were discussed and the use of BCR was recommended for the
analysis. The parallel and intermittent operations of PAT were compared with the conventional
turbines. It was found that, in spite of the lack of flow control devices in PATs, usually the large
reduction in cost makes them more economical than conventional machines. Also, from water-
hammer considerations, it was recommended to use an external device to damp the pressure
fluctuations instead of using a very thick-walled penstock which led to additional cost saving.

Maher et al. (2003) carried out a cost comparison of different off-grid electricity generation
options viz. PAT based pico hydro plant, solar home system and battery system in a rural area of
Kenya. The annual life cycle cost (ALCC) and cost per kWh were worked out for different
options by considering the installation cost, life of the system and annual maintenance and
operation costs. The details of cost analysis are given in Table 2.14. The total cost of 2.2 kW pico
hydropower project was worked out as US$ 7865; which included the costs of civil works (3%);
PVC penstock (8%); turbine, generator, controller and protection (22%); distribution system
cable, house wiring and energy saving bulbs (43%); house-wiring labor (5%) and costs for
design, delivery and project management (19%). From the analysis, a PAT based pico hydro plant
was found to be more cost effective, at less than half the installed cost per household, than an
equivalent solar power system. The cost per kWh was worked out to be less than 15% of the
cheapest solar home system, which had put it within the reach of most low-income households.

Table 2.14 Cost comparison of off-grid options of household electrification in Kenya (Maher et
al., 2003).
Annual Energy per
Installed Life
Life of maintenance household Average
cost per cycle
Type of system system and over cost US$
household cost
(years) operation lifetime per kWh
(US$) (US$)
costs (US$) (kWh)
Amorphous silicon, 12 Wp 200 10 35 480 271 1.77
Crystalline, 20 Wp 330 20 35 960 879 1.09
Pico hydro, 16 W average 81 20 12 321 2102 0.15
Auto battery only, 50 Ah 70 2 33 33 31.2 4.36
67
Chuenchooklin (2006) presented the cost analysis for pico hydropower plant for a farming village
at the upstream of Wangthong Watershed in Thailand where pump was installed as turbine. The
actual gross power produced was 1.116 kW with the net head of 10.74 m and flow rate of 21 lps
at maximum system efficiency of 81.5%. The construction cost of the project was approximately
US$ 4000 (45% for pipe systems, 37% for control and electricity systems and 18% for pump and
turbine systems). Based on the overall electricity consumption of 8760 kWh per year and
electricity charges of US 0.75 cent per kWh, the economic recovery period was estimated as 6
years. The results showed that the produced electricity was enough for the indoor electrical
appliances such as electric light and some house-ware appliances. It was recommended to install
PAT based pico/micro/mini hydropower plants in larger farming villages where the higher head
and larger flow rate may be available depending on the topography characteristics.

Arriaga (2010) carried out the cost analysis of 2 kW capacity project in the Xiagnabouli province
in the Lao Peoples Democratic Republic for isolated communities (40-500 people). Three
options were considered for the analysis viz. power generation from hydro resources using PAT
or Vietnamese turbine and solar energy using PV panels; and the prices were taken from the
market of Vientiane. The total cost was comprised of costs of energy generation equipments
(EGEs), civil works (CW) and energy distribution. The EGE included PAT, induction generator
controller (IGC) and related electrical equipments. The CW costs involved the weir, canal,
forebay, penstock and PAT foundation material, excluding the costs for labor and equipment
transportation. Table 2.15 summarizes the EGE and CW costs for the three options for power
generation. The CW costs were maintained equal in both hydropower cases for ease of
comparison. From the analysis, the lowest installation cost was found in case of PAT while the
PV approach was subjected to the highest investment cost. Study revealed that, the proposed PAT
based hydropower project can provide a long-term reproducible system for communities where
pico-hydro propeller turbines were not sufficient and proper turbines were significantly more
expensive.

68
 Table 2.15 Costs of energy generation equipment and civil work components (Arriaga, 2010).
Equipment Capacity EGE CW + EGE +CW EGE EGE + CW
(US$) Transm. line (US$) (US$/kW) (US$/kW)
(US$)
Proposed 2.0 kW 2,545 3,665 6,210 1,275 3,105
PAT
V. turbine 2.0 kW 4,800 3,665 8,465 2,400 4,235
PV panels 2.0 kWp 9,000 - 9,000 4,500 4,500

Motwani (2012) carried out an ALCC analysis for 3 kW capacity micro hydropower plant, by
considering PAT and an equivalent Francis turbine as prime mover, based on initial cost of the
project (Co), capital recovery factor (CRF) and annual expenses (Ac). For the analysis, only
initial cost of machine was considered to evaluate the initial cost of the project assuming that the
costs of civil works, building and miscellaneous items would be same for both the cases. The
CRF was estimated considering 12% annual discount rate (d); and the equipment life (L) of PAT
and Francis turbines were taken as 10 and 25 years respectively based on the market survey. As
annual expenses, operation cost (@ 5% of initial cost) and maintenance cost (@ 10% of initial
cost) were considered; assuming that the costs of manpower and miscellaneous items would be
same for both the cases. Based on the analysis, the ALCC and the cost of electricity generated per
kWh were found to be 85% and 80% less for PAT compared to Francis turbine, which has
justified the use of PAT in place of Francis turbine for their case study. The ALCC and CRF were
worked out using following equations.

ALCC = (Co CRF) + Ac (2.16)

d(1:d)L
CRF = (1:d)L ;1 (2.17)

69
2.12 Market Status of PAT

The concept of running a centrifugal pump in reverse mode as turbine has been recognized by
pump manufacturers for many years and within the water supply industry this concept has been
exploited to a limited degree as a means of generating power in locations where it is considered
too expensive to purchase a hydro turbine. It has been noticed by water suppliers, operators of
small hydropower plants and pump manufacturers that running pumps as turbines is an efficient
method of generating energy as well as recovering energy and contributing to energy savings.
Worldwide, many pump manufacturers have carried out research on PAT and supplied different
types of pumps for power generation in hydropower plants, water supply system etc.

KSB Aktiengesellschaft is one of a group of pump manufacturers that is active in investing

resources in PATs and the company has recorded considerable success with its solutions in
several parts of the world. KSB has already supplied the pumps running as turbines for various
applications like small hydropower systems (<10 MW), major water transport systems, reverse
osmosis and industrial systems where the technology can be employed as an alternative to
throttling devices. KSB has been active in supplying volute casing and ring-section pumps for
PATs duties over several years, mostly into the small hydropower market (Orchard and Klos,
2009).

Andritz Hydro (http://man.g.andritz.com/c/com2011.....) is a global supplier of electro-

mechanical systems and services (water to wire) for hydropower plants and has more than 170
years of accumulated experience in turbine design. Their single-stage and multi-stage centrifugal
pumps are used as mini-turbines for different applications e.g. as recovery turbines in pulp and
paper mills, in small hydropower plants and to supply energy to mountain refuges and forest
lodges. They offer two different series pumps viz. ACT and FPT suitable for turbine mode
operation. The ACT Series is characterized by an open impeller and wear resistant design which
can handle not only drinking water but also residual and waste water, as well as pulp suspensions.
The ACT series PATs are suitable for head, discharge and power up to 80 m, 0.8 m/s and 250 kW

70
respectively; whereas corresponding values for FPT series PATs are 80 m, 6 m/s and 2 MW
respectively.

Kirloskar Brothers Ltd. (KBL) (Natanasabapathi and Kshirsagar, 2004) is a major player in the
manufacturing and supply of fluid handling products. The companys core products are pumps,
turbines and valves and are used in irrigation schemes, power plants, process industries and
domestic applications. KBL has its product range in power generation sector and is supplying
mini and micro hydro turbines. It was mentioned that, interest on pumps operating in turbine
mode is growing for micro hydropower projects, particularly in the range of 10 kW to 100 kW,
and the pumps could be used most efficiently and economically for this power range.

Suzler Pumps (Adams and Parker, 2011) has provided hundreds of HPRTs in various
configurations and has established methods to calculate HPRT performance from pump
performance. The typical HPRT test setup is shown in Figure 2.30. Based on the experience they
have drawn following guidelines for operation of PATs as HPRTs:
At BEP, efficiency in turbine mode could be nearly same as in pump mode or it can be
slightly higher depending on the size of the machine.
Pump in turbine mode gives better efficiency at overload conditions.
In turbine mode, the BEP is obtained at higher head and discharge i.e. the capacity is
higher in the turbine mode than in pump mode.
In most instances, the shaft power at BEP is somewhat higher than that in the pump.
Runaway speed of a radial machine can be between 140% and 200% of the nominal speed
depending on the specific speed and the rated conditions.
For PAT application in processes with entrained gas or vapor (natural gas treating,
fertilizer plants, hydro treaters, etc.), it was recommended to use cavitation-resistant
materials for runner and hardening of wear parts to reduce the contact damage.

71
 Figure 2.30. Typical HPRT test setup in a closed loop (Adams and Parker, 2011).

The pump manufacturers normally do not provide the characteristic curves of their pump
working in turbine mode, which are necessary to select the correct PAT for the hydropower plants
(Fecarotta et al., 2011). Worldwide, the focus of the pump companies as well as the scientists has
been to develop accurate prediction methods for the turbine mode operation of different designs
of centrifugal pumps. Despite there being considerable work by various scientists as reported by
Williams (1992), Derakhshan and Nourbakhsh (2008a), Singh and Nestmann (2010) the accuracy
of these methods has remained a question mark for all pumps with different specific speeds and
capacities. Therefore, it is highly desirable to encourage pump manufacturers to test at least some
of their pumps also in turbine mode, which may further widen their markets and contribute to
better utilization of available SHP potential (https://www.google.co.in/url?sa.....). Also, they may

72
produce two impellers for their centrifugal pumps: one for pump mode and another for turbine
made (Derakhshan et al., 2009).

2.13 Limitations and Recommendations

From the literature survey, it was found that the most critical step in the PAT development
technology is selection of the most appropriate machine for given application and the accurate
prediction of the turbine mode performance characteristics (Singh, 2005). Many researchers have
developed relations for the selection of pump to be used as turbine based on theoretical,
experimental and numerical investigations and the results are encouraging.

The hydropower plants are subjected to two kinds of variations in their operating conditions; viz.
short term, due to changes in load demand, and seasonal, due to changes in the available head and
flow. In conventional hydro turbines, the rotating speed is maintained constant against variable
power demand by regulating the discharge by changing the guide vane positions. But in case of
PAT, guide vanes are not there hence it may be considered as the turbine with full guide vane
opening and thus the speed varies according to the varying power output. This may lead to
instabilities in PAT at part load and results in poor part load efficiency, which is one of the major
issues impeding the PAT technology. The performance curves for different turbines are shown in
Figure 2.31 (https://www.google.co.in/url?sa...). It can be seen that, conventional turbines have
wide operating range between 20-90% of the discharge, whereas the PAT works with higher
efficiency in discharge range of only around 80-100%; hence, their applications are
recommended at maximum attainable efficiency for fixed load applications, close to full load.

To take care of seasonal flow variations, the PATs can be designed for the minimum annual flow
rate or the better option could be to run several PATs in parallel to achieve good performance at
part load. If two or more PATs are operated in parallel, they can be switched on and off according
to the available flow (Alatorre-Frenk, 1994). The part-load problem was effectively solved by
synchronizing 3 PATs on a single shaft in the tea garden in Darjeeling, India (Singh, 2005).
73
Parallel operation has proven to be more cost-effective than a single conventional hydraulic
turbine of comparable capacity (Spangler, 1988): the limit for this advantage is 5 PATs in parallel
according to Fraser & Associates (Fraser et al., 1981), 7 according to Nicholas (1988) and
Hochreutiner (1991). In addition, PATs can also be installed in parallel with conventional
turbines, the fine-tuning being given by the later. Running several PATs in parallel requires
minimum control; however, this type of arrangement may diminish the low cost advantage of the
PAT over the use of a single turbine in some typical cases (Orchard and Klos, 2009).

Figure 2.31. Performance characteristics of different turbines (Steller et al.,

https://www.google.co.in/url?sa.....).

Another technique to accommodate flow variations with a fixed geometry PATs is to store water
in a reservoir and to release it intermittently. As compared with parallel operation, intermittent
operation is more efficient and simpler to operate, because it uses all the available water and its
operation is automatic. Furthermore, it uses a larger machine that will usually be more efficient
and cheaper than several small machines. However, the intermittent operation does not have the
two advantages of parallel operation: firstly, the possibility of part-load operation during
maintenance, and secondly, the possibility of installing some of the PATs at a later stage, to
reduce the initial cost (Alatorre-Frenk, 1994).

74
Some researchers suggested that the upper limit of application of PATs is established by their
market availability off the shelf, i.e. around 100 kW as mentioned by Schmiedl (1988) or 250
kW by Engeda and Rautenberg (1988). However, other authors affirmed that, even if PATs are to
be manufactured for a typical applications, they continue to be more economical than
conventional turbines up to a much higher threshold: Lawrence (1979) mentioned 1.5 MW;
Garay (1990) 2 MW; Grant & Barn (1984) talk about several MW, with various 500 kW
machines in parallel; and Hochreutner (1991) described a Swiss hydropower station with 7
numbers of 931 kW PATs in parallel.

Hydroelectric power plants are often subjected to run at off-design operation in order to satisfy
the fluctuating load demands which would naturally require higher performances at part loads.
On the basis of literature review, it was found that many efforts have been put by different
researchers for improving the performance of conventional hydro turbines under part load
conditions viz. by introducing the axial jet from the center of runner of Francis turbine (Resiga et
al., 2006), by providing the vane in the bend of the draft tube (Patel and Satanee, 2006), by
placing the guide vanes in hydro turbines (Yassi and Hashemloo, 2010; Fei et al., 2011) etc.
Similar techniques can be applied in PATs to improve its part load performance which may widen
its scope of applications. Steller et al. (https://www.google.co.in/url?sa.....) summarized the main
research directions in the field of PAT technology viz. identification of home manufactured
impeller pumps suitable for energy generation, determination of performance characteristics for
specified series of pumps and establishing general relationships between BEP parameters in
pump and turbine modes and low-cost modifications of the flow part geometry for enhancement
of performance of PAT.

The current practice in the design of draft tube for PAT is to follow the standards available for
conventional hydro turbines, as standards for the design of draft tube for PAT are not available.
However, in certain cases, positive pressure has been observed at the exit of draft tube of PAT;
which has recommended the use of constant area pipe at PAT exit in place of diverging shape
draft tube. More detailed experiments were recommended for the optimization of draft tube in
view of economy involved with constant area draft tube in comparison with diverging draft tube
(Motwani, 2012). Another issue of interest would be the economic and technical comparison
75
between hydraulic ram pumps and pump-PAT sets, aimed at proposing an application boundary
between both technologies (Alatorre-Frenk, 1994).

On the basis of literature review, it was found that many efforts have been put by the researchers
for selection of pump to be used as turbine as well as for the improvement of BEP performance.
However, very less work has been done for the optimization of geometric and operational
parameters of PAT as well as for the part load performance enhancement. Also, not much
information is available on the cavitation characteristics of PAT.

2.14 Objectives of the Present Study

From the literature review, it was found that the efficiency of PAT is usually lower than that of the
conventional hydro turbines. Many researchers have presented theoretical and empirical relations
to predict the PAT characteristics from the pump mode performance but the deviations between
these methods and experimental results were found around 20% or even more. The efficiency of
pump running in turbine mode was found mostly less than that in the pump mode. Several
numerical studies on PAT have been reported in the literature, which has proven that URANS
equations can predict the machine performance with less than 10% error compared to
experimental data. The scope of applications of PAT can be widened by improving maximum
and/or part load efficiency by optimizing its geometric/operational parameters viz. impeller
diameter and rotational speed or with some low cost modifications viz. blade rounding, provision
of flow controlling mechanism etc. Moreover, very limited information has been found on
cavitation characteristic of PAT. The future work for the improvement of PAT performance will
be focused on low cost modifications in the impeller, improvement in the selection criteria, part
load performance enhancement, cavitation analysis, studies on water hammer, fluid structure
interaction etc.

76
The objectives of the present study are as under:
To optimize the geometric and operational parameters of PAT viz. impeller trimming,
rotational speed and impeller blade rounding for performance enhancement
To develop an empirical correlation for prediction of PAT efficiency based on non-
dimensional impeller diameter and rotational speed
To improve the part load performance of PAT with flow controlling mechanism
To identify the cavitation free operating conditions for PAT for improved life of the
system

In the next Chapter, selection and design of various components and details of experimental setup
are presented.

77
78