19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, June 7-11, 2004

PUMPING SYSTEM SIZING AND MODELLING TOOL

André Mermoud
CUEPE (University Center for the Study of Energy Problems) – University of Geneva
Battelle Bât A, 7, route de Drize, 1227 Carouge (Geneva), Switzerland
Tel +41 22 379 06 50 – Fax +41 22 379 06 39 – e-mail andre.mermoud@cuepe.unige.ch

ABSTRACT: Solar pumping system sizing and optimization is a rather complex task relying on numerous
parameters. Most pump manufacturers do indeed propose their own "standard" system configuration, or simple tools
for a fast sizing, but only applicable on one standard irradiation day. The aim of this project is to implement a general
and complete pumping system treatment in the existing software PVsyst. This should offer two complementary
approaches, a pre-sizing and a detailed simulation tool. Any significant part of the pumping system will be treated in
detail, including the user's needs definition, the storage, the well dynamic behaviour, the PV array yield and
characteristics (over significant meteo time series, i.e. yearly simulation), several regulation and power converter
strategies, and using pumps of any technology (defined from the manufacturer usual datasheet). Results will include
several indicators like water supply availability, power losses (for example PV-pump mismatch), well drawdown
limitations, over-capacity (full storage), economic evaluation, etc.
Keywords: PV Pumping, Modelling, Simulation, Software

1 INTRODUCTION repeated with different array and pump size

With the increasing acuity of water supply problems,
especially in developing countries, solar pumping
systems will probably take a great importance during the
next years. However, Solar Pumping System sizing and
optimization is a rather complex task. Most pump
manufacturers do indeed propose their own "standard"
system configurations, or graphical tools for evaluating
the production of their pump during a normalized day.
There are several studies aiming to extend these basic
data to more general cases. But to our knowledge there is
no general tool available on the market able to simulate
such systems with sufficient generality and accuracy, in
order to allow to compare the performance of different
system configurations in a given situation.
The aim of this project is to implement a general
pumping system study in the existing software PVsyst
[7]. The program will offer two complementary
approachs: a pre-sizing part and a detailed simulation
part.
2. PRE-SIZING TOOL
While a pumping system is a rather complex system,
involving many variables and operating at very different
conditions over the day and the year, component sizing is
a rather difficult task. The customer usually specifies its
water needs in volume, as well as the head (level
difference) at which it should be pumped. He may also
define a required autonomy duration and a time fraction
during which he accepts that the water needs are not met
by the system (empty tank).
Starting from these requirements, and using rough,
but quick yearly simulations, the program is able to
determine the PV array, pump size and tank volume that
are necessary to meet these water demand requirements.
It also gives a very rough estimation of the costs.
This very simplified simulation runs over one year in
daily values. It uses a well established meteo model to
determine the irradiance on the PV plane, and very
general performance parameter (such as PV array
efficiency, pump efficiency, etc), pre-defined for several
pump and module technologies. This simulation is
arrangements, until matching the user's requirements.
The results show the monthly yield of water (with
respect to the specified needs), the missing water, and the
excess PV energy (when the tank is full).
Of course, this early layout proposition should be
asserted by a detailed simulation, using real
commercially available components, and taking all
system features into account.
3. THE SIMULATION PROCESS
Any pumping system shows highly non-linear
characteristics. We can mention the pump starting
threshold, PVarray – pump electrical mismatch,
converter efficiencies, well characteristics, operating
limits, etc. This is why using "thumb rules" or extensions
of manufacturer partial data for accurate performance
evaluations are not suitable.
Therefore a detailed study should involve a step-by-
step simulation over a realistic meteo period. PVsyst
offers the framework of such a simulation in hourly
values, according to various meteo data and detailed
modelling of components.
3.1. Features of the Pumping system
The parameter definition and the simulation process
have to take the following aspects of the system into
account:
• Definition of the user's water needs, which may be in
seasonal or monthly values,
• Characterization of the storage tank, if any,
• Dynamic behaviour of the well, with eventual
drawdown limits,
• Photovoltaic energy yield (including meteo data,
plane orientation, PV array characteristics, shadings,
etc),
• Motor-pump device,
• Regulation and power converters strategy,
• Eventual battery storage or auxiliary back-up
generator.
3.3. Photovoltaic Energy Yield
The simulation will take advantage of the existing
 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, June 7-11, 2004

complete treatment of the available PV-power in PVsyst: 4.2 Borehole modelling

wide meteo input choice (including extensive database, If we consider the borehole as an impervious tube,
and import from several external sources), plane when pumping the water level will drop as the flowrate Q
orientation (with tracking or shed arrangement [m3/h] divided by the hole area Aw [m²].
possibilities), shadings, PV-array characterization (one- On the other hand, the re-filling of the well from the
diode model, special features for thin film modules, wide surrounding porous medium is a diffusive process. One
PV-module database), evaluation of all losses (IAM, can admit as a reasonable hypothesis that the refilling
module quality, mismatch, temperature, wiring flowrate is proportional to the stress, i.e. the drawdown
resistance, etc). dynamic head.
3.4. Motor-Pump device behaviour Under these hypotheses the real level in the well (or
HD evolution) will obey the following equation:
A general pump model was developed for this
project, which should be applicable to any available dHD / dt = - 1/τ * HD + Q (t) / Aw
commercial pump, and is therefore based on the One can easily see that for steady-state conditions
specifications usually available from the manufacturer's (dHD / dt = 0), this equation leads to a drawdown height
datasheets. This model is described elsewhere [6]. HD linear with the flowrate. Indeed, compared to a
reference case, we have for any flowrate:
4 TOTAL HEAD and WELL MODELLING HD = Q * HDref / Qref
Head is usually expressed in units of level difference Under this hypothesis, the ratio HDref/Qref is a
[meter]. Physically, passing to pressure units involves characteristics of the well, which we will call the
multiplying the height by the water density (= 1000 "specific drawdown" (expressed in [meter / m3/h]).
kg/m3) and gravitation constant ( g = 9.81 m/sec² ). For This parameter is mainly related to the geologic
getting [bar], we have to divide by 100'000 [Pa/Bar]. properties of the surrounding ground (permeability,
storage capacity), and the construction technique of the
4.1 Total Head
borehole. It may be measured rather easily, using a
The pump has to provide a total head resulting of portable engine-pump and measuring the water depth and
several contributions. flowrate in stabilized conditions.
For example in case of a pump immersed in a well: if
4.3 Borehole parameter in PVsyst
we take reference to the ground level, we have (fig 1):
As a matter of fact, a pumping test is often performed
HT = HG + HS + HD + HF
for measuring the borehole performance, which yields
where: essentially 3 parameters: the static level (HS), a reference
HG = height of the outlet pipe above the ground flowrate available from the well Qref, and the
(assuming that outlet pressure is negligible). corresponding dynamic level (HDref). Ref [1] reports
HS = static head due to the depth of the water level in several results of such tests in Africa, of which we give
the well, in absence of any pumping. some examples.
HD = dynamic head: in a borehole well, the effective
water level is dynamically lowered by the water flow HS HDref Qref HD/Q
extraction. [m] [m] [m3/h] [m/m3/h]
HF = friction losses in the piping circuit, which depend Angola
on the flowrate, Rotunda 20 25 7.2 3.5
Chamaco 12 20 6.9 2.9
Lupale 20 24 5 4.8
Morocco
Abdi 13 22 21.6 1.0
HG Storage Tank
Ourika 17 2 10.8 0.2
Iferd 10 50 36 1.4

Ground level
Table 1. Characteristics of some borehole in Africa [1].
We can observe from these examples that the
HS
Dynamic contribution is not to be neglected !
The recovery time τ (corresponding to a 1/e re-
Static water level filling) is easily calculated from the steady state
conditions:
Hmax τ = Aw * HDref / Qref
HD
For example, in the case of a borehole of diameter
0.15 m in Rotunda, this is about 4 minutes. Therefore this
Water level when pumping dynamic model describes the short term behaviour of the
well.
Pump inlet level Long term variations are likely due to modifications
of the phreatic water level along the seasons. They may
Pump be introduced in PVsyst by specifying a monthly profile
of the static head HS. Long-term exhausting effects
caused by an excessive water drain involve complex (and
Fig 1. Schematic of a borehole well not sustainable) phenomena which are not modelled here
in PVsyst.
 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, June 7-11, 2004
Finally, the simulation (as well as the real system
regulation) should take the maximum head Hmax, i.e. the
inlet level of the pump, into account for stopping the
pump, avoiding dry-running.
4.4 Storage
In most installations, the storage tank is placed rather
high, in order to allow water utilization by free outflow.
Storage tank should be sized according to the desired
autonomy (number of days without sufficient sun, under
the specified consumption).
Usually the pipe outlet pours water at the top of the
storage tank. The residual pressure (giving rise to kinetic
energy in the jet), is neglected.
A slight efficiency improvement could be expected
by filling the tank through the bottom, which would
avoid loosing the height between the outlet and the water
level in the tank; this of course requires the use of a non-
return valve in the pipe. PVsyst will also treat this case
for evaluating the benefits.
5 POWER CONVERTER and REGULATION
5.1 Direct coupling
Direct coupling of DC-pumps with the PV array is
still widely used, especially in small pumping systems,
owing to its simplicity. Many studies already reported the
Array-Pump coupling and its unavoidable mismatch in
the Current/Voltage characteristics [2].
Such a configuration implies in fact a very careful
optimization. Fig 2 shows a typical pump behaviour,
superimposed on a I/V array characteristic. At any time,
the operating point is the intersection of the two
characteristics. If the pump curve is too high (array
current undersized), the pumping threshold will be high,
penalizing the low irradiances (low season, bad days and
morning/evening). If it is too low, the full potential
power of the array is not used during bright hours. The
optimal sizing is therefore depending either on the
irradiance distribution, and on the periods at which the
water needs are the more important.
PV array and Pump Characteristics
12
10
8
Current [A]
6
Head = 40m
.
Pump characteristics . over the maximum voltage required by the pump at the
Head = 20m
4 maximum desired flowrate.
2 5.4 MPPT–DC Converter
DC-DC converter
0
0 10 20 30 40
Voltage [V]
Fig 2. PV array and pump I/V characteristics
(Ex: diaphragm pump Watermax BU)
On the other hand, the pump characteristic is strongly
dependent on the head, displacing the curve parallel to
itself. Therefore, the sizing will also be dependent on the
conditions of use, impeding using simple "thumb rules"
valid for any system at any place.
5.2. Booster devices
Moreover most displacement pumps require a
significant peak of current (at low voltage) when starting,
in order to overcome the internal friction forces. We see
on the diagram that this array configuration is not able to
provide the peak unless by waiting very high insolation,
increasing the irradiance threshold.
Help is usually provided by an electronic device
named "Booster", which stores the PV energy in a
capacity and gives it back as a peak of current.
This non-linear behaviour should also be reproduced
in detail in the PVsyst simulation.
5.3. DC–DC converter
Use of a DC-DC converter shows a much favourable
figure. This cheap electronic device absorbs the power of
the PV array at a fixed voltage, and behaves as a current
generator for feeding the DC-motor of the pump.
At the input side, the voltage may be chosen close to
the maximum power point, and stays quite near for any
irradiance (fig 2). On most commercial DC-DC devices,
the input voltage may be adjusted by hardware.
Performances are only related to the array and meteo,
they don't depend on the pump configuration.
At the output side, the power is transmitted to the
motor at the optimal current/voltage point corresponding
to the available power.
Nowadays, the converter efficiency is usually of the
order of 95% in the high powers region. The converter
efficiency drops toward low powers as other similar
devices (for example inverters). It is treated as such in
PVsyst: an efficiency profile may be specified; but by
default, a standard efficiency profile is constructed using
the maximum and "euro" average efficiency, defined in a
similar way as for inverters.
By the way, efficiency drop often arises at powers
which are below the hydraulic threshold of the pump;
therefore it doesn't affect the normal running.
DC-DC converter also plays the role of a "Booster".
The starting high current is usually required under very
low voltage, therefore low power. The current threshold
becomes a power threshold, which is easily provided
already at low irradiances (for example on fig. 2, a peak
starting current of 10A under 5V (50W) would
correspond to an irradiance threshold of less than 100
1000 W/m², 55°C
W/m²).
800 W/m², 50°C
600 W/m², 45°C
We have to point out here a design constraint: most
400 W/m², 40°C of the DC-DC converters operate on "step-down"
200 W/m², 35°C principle. This means that they cannot deliver a voltage
greater than the input voltage. Therefore the PV-array
MPP voltage (under bright sun conditions) should be
These are equivalent to DC-DC converters, with an
50
60 70 80 90 input voltage continuously adjusted to the maximum
power of the array.
PVsyst includes a tool for estimating the gain
obtained using MPPT with respect to a DC-DC converter
operating at fixed voltage, for any meteo location or PV
array configuration. Surprisingly, over the whole year
this shows that the optimum fixed voltage performance is
usually very near from the MPPT (only 1-2% lower in
most cases, see fig 3).
MPPT devices for driving DC motors are still rather
rare on the market. Therefore it should be determined in
each case whether the overcost of such a refinement is
really worth while, by respect to a standard DC-DC
 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, June 7-11, 2004
converter (of lower cost). Keeping in mind that this last
solution implies a careful voltage initial adjustment.
Array average effic. by respect to MPP operation
120%
Average MPP operation (67 V)
By respect to MPP operation
100%
80%
60%
Yearly array Yield at fixed operating voltage
(maxi 99% at 65.5V)
40%
20%
0%
40 45 50 55
Fixed operating voltage
fig 3. Relative annual PV-array yield with a fixed DC
input (tool available in PVsyst).
5.5 DC – AC inverter
Pumps driven by an AC motor require of course a
power converter in any case. Manufacturers of AC
pumps especially designed for solar use usually propose
their own suitable converter. In these cases, they don't
specify the intermediate values (voltage, current,
frequency) between the output of the converter and the
input of the pump: the simulation process should
consider the set converter-motor-pump as a whole: the
pump model electric input is only based on the power,
without information on the voltage/current.
A very promising design is also based on the use of
standard frequency converters (FCs), coupled to standard
submersible centrifugal pumps (refs. [3],[4]). This allows
for choosing any commercial pump, not dedicated to
solar use. As the market for these materials is much
greater than the solar market, this results in cost
reductions (the cost of a standard FC is typically half the
cost of a PV inverter), improvement in reliability and
service, extended power range availability, etc.
Again, DC-AC inverters are often "step-down"
devices. This implies that the array MPP voltage should
be greater than the peak AC values of the sinus (i.e about
nominal voltage * 1.4); otherwise the signal will be
distorted, resulting in motor inefficiencies.
With standard 230V pumps, the PV array should
provide around 320V. Therefore use of standard
materials is suited for rather high powers.
5.6 Array reconfiguration
Direct coupling mismatch may be improved by
performing a PV-array reconfiguration: if we consider
four identical groups of PV-modules, at low irradiance
all groups are connected in parallel, providing the high
currents necessary to the pump starting. From a given
irradiance level, the groups are connected in series,
doubling the voltage and reducing the current of the PV
array. This requires an electronic switch of rather simple
technology [5].
5.7 Pumps cascading
If the system is equipped with several pumps, the
regulation should switch them ON according to the
available PV power, in order that each pump runs near its
optimum efficiency.
5.8 Use of batteries
Although not often used because of its price and
maintenance requirements, batteries may help to regulate
the power availability when using DC-pumps. The
starting peak current is no more a problem, and such a
system will operate at constant voltage, i.e. constant
flowrate for a given head. This of course simplifies the
design, and allows to use the PV installation in
conjunction with other domestic, artisanal or industrial
appliances.
5.9 Other strategies
All of these system configurations will be treated in
detail in the PVsyst simulations. This implies that the
user's interface allows to comfortably define the involved
60
65 70 75 80
parameters in any case, and that pertinent output results
are calculated and made available.
Other strategies may be investigated, and will also be
treated in PVsyst in a second step.
6. CONCLUSIONS
Proper design of solar pumping systems require a lot
of expertise. Design rules, optimal options and best
practices are disseminated in many specialist documents.
This program aims to gather this information and to offer
an integrated environment presenting the main available
design possibilities, and allowing to closely compare
their performances for each specific foreseen project.
In pumping systems, design values and their
interdependence, as well as their implications on the
overall system performances, are often not intuitive. The
software will check the coherence of the parameters
specified by the user and should produce warnings for
incompatibilities. This "expert" part of the software will
probably be the more difficult part of the package, and
will be progressively improved according to the
experience gained when using the software (and
hopefully with the remarks and suggestions of the users).
We hope that this new module of PVsyst will be
available within some months.
REFERENCES
[1] L. Navarte, E. Lorenzo, E. Caamaño: PV pumping
analytical design and Characteristics of Boreholes.
Solar Energy 68, no 1, pp 49-56, 2000.
[2] J.A. Roger: Theory of the direct coupling between
DC Motors and PV solar Arrays.
Solar Energy 23, pp 193-198, 1979.
[3] M. Alonso-Abella, E. Lorenzo, F. Chenlo: PV
Water Pumping Systems based on Standard
Frequency Converters. Progress in Photovoltaic
Research and Applications 2003, 11:1-13.
[4] M. Alonso-Abella, F. Chenlo, J. Blanco, D. Manso:
Use of Standard Frequency Converters in PV
pumping systems. 2nd World Conf. and Exhibition
on PVSEC, Vienna, 1998.
[5] Ziyad Salameh, A. K. Mulpur, F. Dagher: Two-
stage Electrical Array Reconfiguration controller
for PV-powered Water Pump.
Solar Energy 44, no 1, pp 51-56, 1990.
[6] André Mermoud: Pump behaviour modelling for
use in a general PV simulation software. 19th
European PVSEC, Paris, France, June 7–11, 2004.
[7] PVsyst software. CUEPE, University of Geneva,
www.pvsyst.com