IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO.
1, JANUARY/FEBRUARY 2014
Implementation of a High-Efficiency, High-Lifetime,
and Low-Cost Converter for an Autonomous
Photovoltaic Water Pumping System
João Victor Mapurunga Caracas, Student Member, IEEE, Guilherme de Carvalho Farias, Student Member, IEEE,
Luis Felipe Moreira Teixeira, Student Member, IEEE, and Luiz Antonio de Souza Ribeiro, Member, IEEE
Abstract—This paper proposes a new converter for photovoltaic
(PV) water pumping or treatment systems without the use of
chemical storage elements, such as batteries. The converter is
designed to drive a three-phase induction motor directly from
PV energy. The use of a three-phase induction motor presents a
better solution to the commercial dc motor water pumping system.
The development is oriented to achieve a more efficient, reliable,
maintenance-free, and cheaper solution than the standard ones
that use dc motors or low-voltage synchronous motors. The de-
veloped system is based on a current-fed multiresonant converter
also known as resonant two-inductor boost converter (TIBC) and
a full-bridge three-phase voltage source inverter (VSI). The classic
topology of the TIBC has features like high voltage gain and low
input current ripple. In this paper, it is further improved with
the use of a nonisolated recovery snubber along with a hysteresis
controller and the use of a constant duty cycle control to improve
its efficiency. Experimental results show a peak efficiency of 91%
at a rated power of 210 W for the dc/dc converter plus the
three-phase VSI and a peak efficiency of 93.64% just for the
dc/dc converter. The system is expected to have a high lifetime due
to the inexistence of electrolytic capacitors, and the total cost of
the converter is below 0.43 U$/Wp. As a result, the system is a
promising solution to be used in isolated locations and to deliver
water to poor communities.
Index Terms—AC motor drives, dc–ac power conversion, dc–dc
power conversion, photovoltaic (PV) power systems, solar power
generation.
I. I NTRODUCTION
C URRENTLY, over 900 million people in various countries
do not have drinkable water available for consumption.
Of this total, a large amount is isolated, located on rural areas
where the only water supply comes from the rain or distant
rivers [1]. This is also a very common situation in the north
part of Brazil, where this work was developed. In such places,
Manuscript received August 31, 2012; revised February 1, 2013; accepted
April 6, 2013. Date of publication June 26, 2013; date of current version
January 16, 2014. Paper 2012-SECSC-448.R1, presented at the 2012 IEEE
Applied Power Electronics Conference and Exposition, Orlando, FL, USA,
February 5–9, and approved for publication in the IEEE T RANSACTIONS ON
I NDUSTRY A PPLICATIONS by the Sustainable Energy Conversion Systems
Committee of the IEEE Industry Applications Society. This work was sup-
ported in part by the Federal University of Maranhão (UFMA), in part by
Eletrobrás, in part by CP Eletrônica, in part by Elétrica Visão, in part by Texas
Instruments, and in part by IEEE PELS.
The authors are with the Departamento de Engenharia de Eletricidade,
Universidade Federal do Maranhão, São Luís 65080-805, Brazil.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIA.2013.2271214
0093-9994 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
631
the unavailability of electric power rules out the pumping and
water treatment through conventional systems. One of the most
efficient and promising way to solve this problem is the use of
systems supplied by photovoltaic (PV) solar energy. This kind
of energy source is becoming cheaper and has already been
put to work for several years without the need of maintenance.
Such systems are not new and are already used for more than
three decades [1]. Nevertheless, until recently, the majority of
the available commercial converters in Brazil are based on
an intermediate storage system, performed with the use of
lead–acid batteries, and dc motors to drive the water pump
[3]. More sophisticated systems have already been developed
with the use of a low-voltage synchronous motor [4], but these,
although presenting higher efficiency, are too expensive to be
used in poor communities that need these systems.
The batteries allow the motor and pump system to always
operate at its rated power even in temporary conditions of
low solar radiation. This facilitates the coupling of the electric
dynamics of the solar panel and the motor used for pumping [5].
Generally, the batteries used in this type of system have a low
life span, only two years on average [5], which is extremely
low compared to the useful life of 20 years of a PV module.
Also, they make the cost of installation and maintenance of such
systems substantially high. Furthermore, the lack of battery
replacement is responsible for the failure of such systems in
isolated areas.
The majority of commercial systems use low-voltage dc
motors, thus avoiding a boost stage between the PV module and
the motor [7]. Unfortunately, dc motors have lower efficiency
and higher maintenance cost compared to induction motors
and are not suitable for applications in isolated areas, where
there is no specialized personnel for operating and maintaining
these motors. Another problem is that low-voltage dc motors
are not ordinary items in the local markets. Because of the
aforementioned problems, this work adopted the use of a three-
phase induction motor, due to its greater robustness, lower
cost, higher efficiency, availability in local markets, and lower
maintenance cost compared to other types of motors.
The design of a motor drive system powered directly from a
PV source demands creative solutions to face the challenge of
operating under variable power restrictions and still maximize
the energy produced by the module and the amount of water
pumped [8]. These requirements demand the use of a converter
with the following features: high efficiency—due to the low
632 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 1, JANUARY/FEBRUARY 2014
Fig. 1. Simplified block diagram of the proposed system.
energy available; low cost—to enable its deployment where
it is most needed; autonomous operation—no specific training
needed to operate the system; robustness—minimum amount of
maintenance possible; and high life span—comparable to the
usable life of 20 years of a PV panel.
This paper proposes a new dc/dc converter and control suit-
able for PV water pumping and treatment that fulfill most of the
aforementioned features. This paper is organized as follows. In
Section II, the proposed system is described. In Section III, the
dc/dc converter itself is presented and analyzed. In Section IV,
the control strategy is explained, and in Section V, the experi-
mental results are presented.
II. P ROPOSED C ONVERTER
To ensure low cost and accessibility of the proposed system,
it was designed to use a single PV module. The system should
be able to drive low-power water pumps, in the range of 1/3 hp,
more than enough to supply water for a family. Fig. 1 presents
an overview of the proposed system. The energy produced by
the panel is fed to the motor through a converter with two power
stages: a dc/dc two-inductor boost converter (TIBC) stage to
boost the voltage of the panels and a dc/ac three-phase inverter
to convert the dc voltage to three-phase ac voltage. The inverter
is based on a classic topology (three legs, with two switches
per leg) and uses a sinusoidal pulse width modulation (PWM)
(SPWM) strategy with 1/6 optimal third harmonic voltage
injection as proposed in [9]. The use of this PWM strategy is
to improve the output voltage level as compared to sinusoidal
PWM modulation. This is a usual topology, and further analyses
on this topology are not necessary. For the prototype used to
verify the proposed system, a careful selection of the voltage
source inverter (VSI) components is more than enough to
guarantee the efficiency and cost requirements.
The required dc/dc converter for this kind of system needs to
have a large voltage conversion ratio because of the low-voltage
characteristic of the PV panels and small input current ripple so
that it does not cause oscillation over the maximum power point
(MPP) of the PV module [10]–[12], thus ensuring the maximum
utilization of the available energy.
The commonly used isolated voltage-fed converters normally
have a high input current ripple, which forces the converter
to have large input filter capacitors. These are normally elec-
trolytic, which are known to have a very small lifetime and
thus affect the overall life span and mean time before failure
of the converter. Furthermore, the inherent step-down char-
acteristic of the voltage-fed converters, the large transformer
turns ratio needed to boost the output voltage, the high output
diode voltage stress, and the need of an LC output filter
[13] make voltage-fed converters not the best choice for this
application.
When compared to the voltage-fed topologies, current-fed
converters have some advantages. Usually, they have an in-
ductor at the input, so the system can be sized to have input
current ripple as low as needed, thus eliminating the need of the
input capacitor at the panel voltage. Current-fed converters are
normally derived from the boost converter, having an inherent
high step-up voltage ratio, which helps to reduce the needed
transformer turns ratio. The classical topologies of this kind are
the current-fed push–pull converter [14], [15], the current-fed
full-bridge [16], and the dual half-bridge converter [17]. Al-
though the current-fed topologies have all the aforementioned
advantages, they still have problems with high voltage spikes
created due to the leakage inductance of the transformers and
with high voltage stress on the rectifying diodes [18].
One of the solutions to the current-fed PWM converters is
the use of resonant topologies able to utilize the component
parasitic characteristics, such as the leakage inductance and
winding capacitance of transformers, in a productive way to
achieve zero current switching (ZCS) or zero voltage switch-
ing (ZVS) condition to the active switches and rectifying
diodes [19].
In this paper, the use of a modified TIBC for the first-stage
dc/dc converter is proposed, due to its very small number of
components, simplicity, high efficiency, easy transformer flux
balance [20], [21], and common ground gate driving for both
switches. These features make it the ideal choice for achieving
the system’s necessary characteristics. Aside from the high
dc voltage gain of the TIBC, it also compares favorably with
other current-fed converters concerning switch voltage stress,
conduction losses, and transformer utilization [22], [23]. In
addition, the input current is distributed through the two boost
inductors having its current ripple amplitude halved at twice the
PWM frequency. This last feature minimizes the oscillations at
the PV module operation point and makes it easier to achieve
the MPP.
In its classical implementation, the TIBC is a hard-switched
overlapped pulse-modulated converter; this way, at least one of
the switches is always closed, creating a conduction path for the
input inductor current. Nevertheless, the TIBC can be modified
to a multiresonant converter by adding a capacitor at the trans-
former’s secondary winding [24], [25]. A multiresonant tank
is formed by the magnetizing inductance of the transformer,
its leakage inductance, and the added capacitor, as shown in
Fig. 2(a). The intrinsic winding capacitance of the transformer
is included in the resonant capacitor.
CARACAS et al.: IMPLEMENTATION OF HIGH-LIFETIME LOW-COST CONVERTER FOR WATER PUMPING SYSTEM
Fig. 2. Modified TIBC topology: (a) resonant tank, (b) voltage doubler
rectifier, and (c) snubber.
By adding this capacitor and using the parasitic components
of the transformer to create the resonant tank, it is possible
to achieve ZCS condition for the input switches and output
rectifying diodes, and this enables the converter to operate at
high frequencies with greater efficiency.
With the use of a voltage doubler rectifier at the secondary
side of the transformer, as shown in Fig. 2(b), it is possible to
reduce the transformer turns ratio, the necessary ferrite core,
and the voltage stress on the MOSFETs to half of the original
ones. As a result, the transformer is cheaper, the MOSFETs
are cheaper, and the number of diodes in the secondary side is
halved. Also, the output dc bus capacitor can be integrated with
the capacitors of the rectifier, particularly because the second-
stage three-phase VSI has almost dc input current, exempting
the bus capacitor from any ac decoupling current.
Classically, the TIBC have a minimum operation load to
maintain an established output voltage. Below a certain load
level, the energy transferred to the output capacitor is not
completely transferred to the load and causes an increase in the
output voltage. This happens because the inductors are charged
even if there is no output current. As a result, this converter has
a drawback when used in motor drive systems. Since the motor
is a variable load and it has large time constants, it will demand
low power at some operation points, i.e., at low speed and start-
up/stop transients. As a solution, a nondissipative regenerative
snubber circuit is presented.
The regenerative snubber is formed by two diodes and a
capacitor connecting the input side directly to the output side of
the converter, as shown in Fig. 2(c). This makes it a nonisolated
converter, which has no undesirable effect in the PV motor
driver applications. The voltage over the MOSFETs is applied
to a capacitor connected to the circuit ground, and the voltage of
this capacitor is coupled in series with the output of the rectifier.
This modification allows part of the energy to be transferred
from the input directly to the output, through the snubber,
without going through the transformer, reducing its size and
improving even more the efficiency of the converter.
A modification in the control strategy of the proposed con-
verter is also proposed, when compared to the classical TIBC
control. Section IV shows that, for the application of a PV water
633
pumping system, the correct design of converter voltage gain
makes the converter able to operate with a constant gain, using
a fixed duty cycle modulation for the primary switches. To solve
the minimum load condition, the use of a hysteresis controller
for the dc output voltage of the first stage is proposed. When
operating below the minimum required load, an uncontrolled
increase of output voltage will be seen, and once this voltage
reaches the upper threshold, both primary switches are turned
off. This would be impossible in all the previously reported im-
plementations of the TIBC. However, with the added snubber,
even with both switches turned off, there is still a path for the
input inductors’ current. Their energy is directly transferred to
the snubber capacitor Cs . This capacitor must be sized to have
a minimum voltage increase during this hysteresis action. As
this capacitor is in series with the output rectifier, the same
voltage increase will be noted in the output voltage. After
the switches are turned off and the input inductors’ energy
is transferred to the snubber, the output voltage will start to
decrease, reaching the lower hysteresis threshold and restarting
the PWM operation with the fixed duty cycle.
III. O PERATION P RINCIPLE
To simplify the analysis of the proposed converter, the fol-
lowing assumptions need to be true during a switching interval:
The input inductors Li1 and Li2 are sufficiently large so that
their current is almost constant; the capacitors Co1 , Co2 , and Cs
are large enough to maintain a constant voltage; and the output
capacitors Co1 and Co2 are much larger than Cr to clamp the
resonant voltage.
In the hard-switched operation of the TIBC, the two primary
switches Q1 and Q2 operate at an overlapped duty cycle switch-
ing scheme to guarantee a conduction path for the primary
inductor current. When both Q1 and Q2 are turned on, Li1 and
Li2 are charged by the input energy. When Q1 (Q2 ) is opened,
the energy stored in Li1 (Li2 ) is transferred to Co1 (Co2 )
through the transformer and the rectifier diode Do1 (Do2 ).
Once the multiresonant tank is introduced, two different
resonant processes occur: 1) When both switches are closed, the
leakage inductance Lr participates along with capacitance Cr
in the resonance at the primary current switching and current
polarity inversion, allowing ZCS operation for the primary
switches, and 2) during the conduction time interval (between
t4 and t5 in Fig. 3), when at least one of the switches is open,
Lr is associated in series with Li1 or Li2 , not participating on
the transformer’s secondary current resonance, formed only by
Lm and Cr . The key waveforms for a switching period of the
TIBC are presented in Fig. 3. In this figure, VgQ1 and VgQ2
are the gate signals of the switches Q1 and Q2 , respectively;
VdsQ2 is the drain-to-source voltage of MOSFET Q1 ; IQ2 is
the current of MOSFET Q2 ; VT is the voltage at the primary
of the transformer; IT is the current at the primary of the
transformer; ILi1 and ILi2 are the currents of inductors Li1 and
Li2 , respectively; and Iin is the input current of the converter
and also the current supplied by the PV panel.
At time t1 , the rectifying diode Do1 is already conduct-
ing, and the voltage on resonant capacitor Cr is clamped at
+Vout /2. At this instant, the switch Q1 is activated by VgQ1 .
634 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 1, JANUARY/FEBRUARY 2014
Fig. 3. Key waveforms of the TIBC during a switching period.
As the switch is turned on, its voltage drops to zero, and the
snubber diode Ds1 is forced to stop conducting. From t1 to t2 ,
Cr transfers its energy to the leakage inductance Lr , beginning
the primary switch’s resonant process and forcing the current
IQ2 on the switch Q2 to decrease.
At the time t2 , the rectifying diode Do1 stops conducting, and
Cr continues to resonate with the magnetizing inductance Lm .
From t2 to t3 , the primary switch’s resonance (Q2 ) continues to
force its current to decrease until it reverses its polarity. When
IQ2 is negative, the switch can be turned off. This happens at
instant t3 when VgQ2 is forced to zero.
At the time t3 , the voltage VdsQ2 starts to increase, Q2
is completely blocked, and the snubber diode Ds2 begins to
conduct, transferring energy directly to the snubber capacitor
Cs . Between t3 and t4 , Cr and Lm continue to resonate,
decreasing the voltage on the doubler rectifier’s input and on
VCr . At instant t4 , the voltage across Cr reaches −Vout /2, and
the rectifying diode Do2 starts to conduct, clamping VCr in
−Vout /2.
From t4 to t5 , the capacitor Co1 is charged, and the current
of Do2 starts to decrease. At the instant t5 , Q2 is turned on,
initiating the resonant process on Q1 . As Q2 is activated, Ds2
is forced to stop conduction. At the instant t6 , the current in
Do2 reaches zero, and Do2 stops conducting, reinitiating the
resonance between Cr and Lm . From this moment, until the
end of the switching period, the process repeats symmetrically
as explained for the other input switch.
An extended description of multiresonant TIBC without the
snubber is presented and analyzed in [24], resulting in a detailed
mathematical modeling for both resonant processes during its
operation. However, the analysis in [24] is based on several
complex mathematical models, and consequently, the presented
design method shows several dependent variables, which trans-
lates in a design methodology difficult to be implemented.
In this paper, a simplified methodology based on the ef-
fect of each resonant process, the resonant frequencies, and
the switching frequency is applied. Spice simulations and a
prototype are used to show that, despite the simplicity of the
design methodology, the correct operation of the converter
is guaranteed, particularly the soft switching of the primary
switches for the whole operating load range.
Although the resonant process affects the output voltage,
depending on the resonant tank component values and the load,
this can be neglected because of its small influence and complex
effect. Thus, neglecting the resonant effect over the output
voltage, including the voltage doubler rectifier and the snubber
connecting the primary and the secondary side of the converter,
the static voltage gain (Kv ) of the converter is defined as
 
Vout 1 Ns
= Kv = 2 +1 (1)
Vin 1−D Np
where D represents the duty cycle of each switch and must be
higher than 50% to guarantee the necessary overlapping for the
correct operation. Ns /Np represents the transformer turns ratio.
According to Yuan et al. [24], to minimize the influence
of the load on the resonant process on the primary current
commutation interval, the switching frequency (Fsw ) should be
higher than the resonant frequency (Frs ) of Lm and Cr by a
value of at least 1.1. Thus,
1 Fsw
Frs = √ ≤ . (2)
2π Lm Cr 1.1
During the primary current commutation interval, when both
switches Q1 and Q2 are turned on, inductor Lr participates on
the resonance in parallel with Lm and Cr ; thus, the resonant
frequency for this interval is defined as
1
Frp =  . (3)
Lm Lr
2π Lm +Lr Cr
CARACAS et al.: IMPLEMENTATION OF HIGH-LIFETIME LOW-COST CONVERTER FOR WATER PUMPING SYSTEM 635

Considering that Lm represents the magnetizing inductance

and Lr represents the leakage inductance, then Lm is much
larger than Lr ; thus, (3) can be simplified to
1
Frp = √ . (4)
2π Lr Cr
The duration of the commutation interval is equal to the
overlapping time (Tov ) of the pulse driving signals and can be
calculated as
(D − 0, 5)
Tov = . (5)
Fsw
The whole resonant process of the primary switches has
the duration of half a resonant cycle, as shown in Fig. 3. To Fig. 4. PV I–V curves. The dashed line shows the VMPP points.
guarantee the ZCS condition for the entire load range, the energy with a smaller rms current. Therefore, the losses in the
following conditions must be satisfied: input inductors (Li1 and Li2 ), in the MOSFETs (Q1 and Q2 ),
π and in the transformer are smaller. As a result, the efficiency of
Tov = (6)
ωr the converter improves.
Fsw The operation with a fixed duty cycle makes the converter
Frp = . (7) work with a constant voltage gain Kv , almost independent of
2D + 1
the input voltage. With the correct design of Kv , the system
Another important constraint is the energy accumulated in will always be able to transfer energy from the PV module to
Cr at the beginning of the primary switching resonant process. the motor. Assuming that the converter is always operating at
This energy needs to be completely transferred to the leakage the MPP of the solar panel, the output dc/dc converter voltage
inductance Lr during this process. From this condition, the (dc bus voltage) will be
following equation is derived:
VBUS = VMPP Kv (9)
V2
Lr ≤ o2 Cr . (8) with VMPP being the MPP voltage of the solar panel. Fig. 4
2Iin
shows the I–V characteristic curves for a typical solar panel. It
is shown that the voltage at the MPP (VMPP ) has only small
IV. C ONTROL OF THE S YSTEM variations for different radiation levels. The different VMPP
points for various radiation levels are represented by the black
There are three main aspects in the proposed converter’s
dashed line.
control: 1) During normal operation, a fixed duty cycle is used
On the other hand, it is important to analyze the minimum
to control the TIBC MOSFETs, thus generating an unregulated
dc voltage on the inverter dc bus necessary to drive the motor
high bus voltage for the inverter; 2) an MPP tracking (MPPT)
at a specific power level. For this application, the volt/hertz
algorithm is used along with a PI controller to set the speed of
controller was used to maintain approximately constant the
the motor and achieve the energy balance of the system at the
pump’s motor flux. By doing this, the controller maintains
MPP of the PV module; and 3) a hysteresis controller is used
the capability of the motor to generate nominal torque at any
during the no-load conditions and start-up of the system. Each
speed below its rated value. Neglecting the effect of rotor
of these aspects is described in the following sections.
slip in induction machines and considering that the centrifu-
gal water pump has its torque proportional to the square of
A. Fixed Duty Cycle Control the motor speed and that the frequency (f ) in the volt/hertz
control is proportional to the voltage (V ), the motor output
One of the most important control aspects of this system
power (P ) can be expressed as a cubic function of the motor
is the fact that it is possible to use an unregulated dc output
voltage
voltage and a fixed duty cycle for the first-stage dc/dc converter.
As a resonant converter, there are definite time intervals in the P ∝Tf (10)
switching period for the resonance process to occur. By altering
the duty cycle or the switching period to control the output T ∝f 2
(11)
voltage, the converter may no longer operate at ZCS condition. f ∝ V. (12)
Therefore, the fixed duty cycle is used to overcome these design
problems and ensure that the converter is going to operate in Considering that nominal voltage is necessary to achieve
ZCS condition despite the input voltage or output load. nominal power, Fig. 5 shows a comparison between the output
The duty cycle was chosen to guarantee that the amount of voltage of the dc/dc converter (dashed line) and the minimum
transferred energy occurs during most part of the switching dc bus voltage required to operate the motor (solid line—cubic
interval. Therefore, it is possible to transfer the same amount of function) in the full power range. This minimum dc bus voltage
636 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 1, JANUARY/FEBRUARY 2014
Fig. 5. DC bus value and minimum voltage needed for pump operation in
constant volt/hertz.
was calculated by considering that the inverter is operating at
the maximum voltage with a modulation index of 1 (no over-
modulation is allowed). The correct design of Kv guarantees
that the output voltage of the first stage will always be greater
than the minimum voltage necessary to drive the motor.
Using the SPWM strategy with third harmonic voltage in-
jection allows the inverter to generate a maximum line voltage
equal to the bus voltage. For a three-phase pump with Vrms as
nominal line voltage, the gain Kv can be calculated by
√
Vrms × 2
Kv > . (13)
VMPP,max
B. MPPT Control
The MPPT is a strategy used to ensure that the operating
point of the system is kept at the MPP of the PV panel [27].
The widely used hill-climbing algorithm was applied due to its
simple implementation and fast dynamic response.
This MPPT technique is based on the shape of the power
curve of the PV panel. This curve can be divided into two
sides, to the left and to the right of the MPP. By analyzing the
power and voltage variation, one can deduce in which side of
the curve the PV panel is currently operating and adjust the
voltage reference to get closer to the desired point. The voltage
reference is used on a PI controller to increase or reduce the
motor speed and consequently adjust the bus and panel voltage
by changing its operating point.
C. Hysteresis Control
The main drawback of the classical TIBC is its inability to
operate with no load or even in low-load conditions. The TIBC
input inductors are charged even if there is no output current,
and the energy of the inductor is lately transferred to the output
capacitor raising its voltage indefinitely until its breakdown.
Classically, the input MOSFET cannot be turned off because
there is no alternative path for the inductor current. However,
with the addition of the proposed snubber, the TIBC switches
can be turned off. Thus, a hysteresis controller can be set up
based on the dc bus voltage level. Every time a maximum
voltage limit is reached, indicating a low-load condition, this
TABLE I
PANEL AND M OTOR PARAMETERS
TABLE II
C ONVERTER D ESIGN S PECIFICATIONS
mode of operation begins. In this case, the switches are turned
off until the dc bus voltage returns to a normal predefined level.
As a result, the switching losses are reduced during this period
of time.
V. S YSTEM D ESIGN
The hardware was designed to achieve a final system with
low cost, thus an accessible and market-friendly product. The
motor used in the centrifugal pump is a 0.2-hp (mechanical
power) three-phase induction motor. Considering the nominal
power and efficiency of the motor, a 210-W PV panel was
chosen [28]. The main characteristics of the used panel and
motor are shown in Table I.
The main parameters for the converter are summarized in
Table II. The maximum allowed input current ripple was de-
fined to guarantee a small oscillation around the MPP. The bus
voltage was defined using the minimum necessary voltage for
the chosen inverter topology and PWM strategy, as shown in
√
Vbus > Vrms × 2 = 350 V. (14)
The TIBC switching frequency must be a tradeoff between
switching losses and the size of the transformer and inductors.
A frequency of 100 kHz was used based on simulation and
previous experimental knowledge.
The Kv gain necessary for the converter can be calculated
using (13)
√
Vrms × 2
Kv > = 11.69. (15)
VMPP,max
CARACAS et al.: IMPLEMENTATION OF HIGH-LIFETIME LOW-COST CONVERTER FOR WATER PUMPING SYSTEM
TABLE III
C OMPONENTS U SED IN THE P ROTOTYPE
To guarantee the overlapping conduction of the MOSFETs
while still maintaining the largest possible duty cycle for the
constant duty cycle operating control, D was chosen to be 53%
based on the minimum required overlapping and commutation
times of the chosen drivers and MOSFETs. Using (1), the
minimum ratio Ns /Np can be determined by
Ns K × (1 − D) − 1
> = 2.25. (16)
Np 2
The high efficiency, low cost, and robustness of the system
depend directly on the choice of components. To obtain the
expected level of efficiency, each component was carefully
selected. The TIBC and inverter’s MOSFETs were chosen to
minimize the conduction losses. Extra low equivalent series
resistance (ESR) MOSFETs were used. Normally, the ESR is
inversely proportional to the gate charge and output capacitance
of the MOSFET. These features have large impacts on the
switching losses, but this effect is reduced by the fact that the
dc/dc converter is resonant and has reduced switching losses.
All the diodes used in the voltage doubler rectifier and in
the snubber of the TIBC were chosen as superfast diodes.
Therefore, not only the conduction losses but also the ringing
caused by the diode reverse recovery time is reduced. Table III
shows the main components used in the design.
Considering the fact that the system is isolated and only
fed by a single PV module, the energy used by the control
components must be derived from the module’s power. A high
efficiency integrated switching power supply is used to perform
this task. It was observed that a gain of 1% in the system
efficiency is obtained using this as opposed to a linear regulator.
To reduce power consumption and cost, the whole control
system was implemented on a single low-power digital signal
controller, the DSPIC30F3010. Fig. 6 presents an overview of
the control system. Based on the measured PV panel voltage
(VPV ) and current (IPV ), the MPPT estimates a frequency
reference (fOP ) to drive the motor, which indirectly serves to
regulate the PV voltage by modifying the amount of power
transferred to the motor. A volt–hertz controller calculates the
637
Fig. 6. Block diagram of the control system.
output voltage based on the operating frequency. Then, the
output voltage is used to calculate the modulation index (MPV )
for the switching pattern.
VI. S IMULATION AND E XPERIMENTAL R ESULTS
The proposed converter was simulated on Pspice Orcad
v16.5. Fig. 7 shows the schematics used for the first-stage
TIBC. All parasitic series resistances were included in the trans-
former and capacitors. The control of the primary MOSFETs
was simulated using a fixed pulse modulation and a voltage-
controlled source to implement the hysteresis control within the
limits of 380 V ± 10 V.
Fig. 8 shows the overlapped pulses used to control Q1 and
Q2 , the current in both input inductors, and the current in the
PV module. It is observed that each one of the inductors has a
current ripple at the converter switching frequency and out of
phase with each other; however, both currents are supplied by
the PV module, and when they are analyzed together (IPV ), a
reduction in the ripple amplitude to half of the original ones
is seen.
Fig. 9 shows the gate-to-source voltage, the drain-to-source
voltage, and current for one of the primary MOSFETs. It is
observed that there are no voltage spikes or increased voltage
stress over the switches. In addition, the figure shows that both
turn-on and turnoff occurs at almost ZCS.
Fig. 10 shows voltage and current on the output rectifying
diode Do1 . It is shown that not only the primary MOSFETs are
operated under ZCS condition but also the rectifying diodes.
The operation under ZCS condition allows the use of fast
recovery diodes instead of the expensive silicon carbide ones,
thus reducing the total cost of the system.
Fig. 11 shows a picture of the developed prototype. It was
tested using real water pumping system and a PV module. A
complete autonomous operation was obtained. This includes
the system automatic start-up and shutdown according to the
radiation level.
A programmable source was used to emulate the solar radia-
tion during the period of a day. The converter was connected to
this source to evaluate its overall performance. Fig. 12 presents
the obtained curves. The system is supposed to operate on
the MPP curves at all times, meaning that it needs to stay
as close as possible to the green curves. These green curves
638 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 1, JANUARY/FEBRUARY 2014
Fig. 7. Circuit used in the simulation.
Fig. 8. Simulation results of the input of the dc/dc converter. (a) Input
switches’ driving signals, (b) inductors’ input currents ILi1 and ILi2 , and total
input current (IPV ).
are the MPPT voltage, current, and power levels produced
by the programmable source. The system was able to operate
near the ideal voltage (VMPP ) and current (IMPP ) during the
programmed time, thus achieving operation close to the MPP.
The generated power of the programmable source and the corre-
sponding water flow for the same period of time are also shown.
The MPPT routine analysis was also performed using a real
PV module. A real time analysis of the operation point was
done, i.e., the capability of keeping a steady position on the
MPP of the PV panel. Fig. 13 demonstrates a real time I–V
curve, and the operating point of the system is emphasized by
the darker spot. The system starts its operation at the open
circuit voltage (VOC ) and then stabilizes on the MPP. As the
solar radiation varies, the MPPs move. The system was able to
Fig. 9. Verification of the ZCS condition on the input switch Q2 . (a) VgQ2
driving signal and IQ2 current. (b) VdsQ2 drain–source voltage.
stably track these points. The black dashed line shows all the
points during a day.
Fig. 14(a) shows the current and voltage waveforms in one of
the MOSFETs during a switching interval, with the converter
operating at full load. It is observed that ZVS is achieved for
both turn-on and turnoff events. Fig. 14(b) shows the TIBC
output voltage (green line), the MOSFET Q1 drain-to-source
voltage (pink line), and the MOSFET Q1 current (orange line)
for operation with no output load (the ac motor was removed).
The stable operation of the hysteresis controller for the dc
output voltage is noted. In this test, the hysteresis voltage
band was set to 380 V ± 7.5 V. A small part of the figure
was amplified (zoom picture) to analyze the output voltage
variation. It is observed that, when the system reaches 372.5 V, at
CARACAS et al.: IMPLEMENTATION OF HIGH-LIFETIME LOW-COST CONVERTER FOR WATER PUMPING SYSTEM 639

Fig. 10. Verification of the ZCS condition on the rectifying diode Do1 .
(a) Diode Do1 current and forward voltage. (b) Diode Do1 reverse voltage.

Fig. 12. Voltage and current curves during PV emulation test. Red curves
show open circuit voltage and short circuit current. Green curves show ideal
voltage and current at the MPP. Blue curves show the system voltage and
current during the test.

Fig. 11. (a) Developed prototype and (b) test bench composed of PV panel,
water pump, electronic converter, water tank, PV voltage, PV current, and water
flow sensors.

time Th1 , the converter starts to operate, and the output voltage
increases at a constant slope. Once it reaches 387.5, at time Th2 ,
the PWM signals are turned off. However, the output voltage
still increases with a different slope between times Th2 and Fig. 13. Real-time MPP operation at a real PV module.
Th3 . This is caused by the energy transferred from the input
inductors L1 and L2 through the snubber to the capacitor Cs . Because of the high-frequency nature of the output voltage
Once all the energy is transferred, the PWM signals stay OFF supplied to the motor by this converter, it is not trivial to mea-
until the output voltage decreases reaching again 372.5 V. At sure its efficiency, at least without very expensive equipment.
this voltage level, the system is restarted. Considering that, in an induction motor, only the fundamental
640 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 50, NO. 1, JANUARY/FEBRUARY 2014
Fig. 14. Voltage and current curves in the MOSFET for two different opera-
tion conditions. (a) Drain-to-source voltage and current showing ZCS condition
for a MOSFET. (b) Hysteresis control of the output voltage.
Fig. 15. TIBC and VSI measured efficiency considering only fundamental
output power.
voltage and current components are converted into mechanical
power, we based our measurements only in the fundamental
output power. A first-order filter was used to obtain the fun-
damental component of the voltage, and a digital oscilloscope
was used to obtain the output power of the converter. Fig. 15
presents the measured efficiency considering the input power
versus the fundamental power supplied to the motor. A max-
imum efficiency of 93.64% was obtained for the TIBC first-
stage dc/dc converter, and that of 91% was obtained for the
complete system. These curves were obtained during a real
operation with a PV panel, driving the water pump with varying
solar radiation. The efficiencies take into account all the energy
used for the control electronics and drivers and are valid for a
commercial prototype.
Nowadays, there are reports of more efficient converters.
During the design of the proposed converter, some compro-
mises were made to achieve high efficiency and low cost.
Fig. 16. Price distribution of the (a) system and (b) converter.
The semiconductor components were preferably cheaper than
efficient since they are the most expensive components of the
converter. Also, the magnetic components were handmade,
although in the cost calculation, commercial prices were used.
Fig. 16 presents the distribution of costs of the entire system
and of the prototype. From Fig. 16(a), it can be seen that the
converter takes only a small share (approximately 10%) of the
overall system cost. Fig. 16(b) shows the price distribution of
the converter itself. The overall price of the prototype consider-
ing large scale production was U$90.90. This cost can be further
divided into U$52.28 for the VSI and U$38.62 for the TIBC.
The control system and housing were divided equally between
VSI and TIBC.
VII. C ONCLUSION
In this paper, a converter for PV water pumping and treat-
ment systems without the use of storage elements was pre-
sented. The converter was designed to drive a three-phase
induction motor directly from PV solar energy and was con-
ceived to be a commercially viable solution having low cost,
high efficiency, and robustness. This paper presented the system
block diagram, control algorithm, and design. The experimental
results suggest that the proposed solution could be a viable
option after more reliability tests are performed to guarantee
its robustness.
R EFERENCES
[1] M. Chunting, M. B. R. Correa, and J. O. P. Pinto, “The IEEE 2011
international future energy challenge—Request for proposals,” in Proc.
IFEC, 2010, pp. 1–24.
[2] A. Hahn, “Technical maturity and reliability of photovoltaic pumping sys-
tems,” in Proc. 13th Eur. Photovoltaic Solar Energy Conf., Nice, France,
pp. 1783–1786.
CARACAS et al.: IMPLEMENTATION OF HIGH-LIFETIME LOW-COST CONVERTER FOR WATER PUMPING SYSTEM
[3] M. A. Vitorino and M. B. R. Correa, “High performance photovoltaic
pumping system using induction motor,” in Proc. Brazilian Power Elec-
tron. Conf., 2009, pp. 797–804.
[4] G. Teröde, K. Hameyer, and R. Belmans, “Sensorless control of a per-
manent magnet synchronous motor for PV-powered water pump systems
using the extended Kalman filter,” in Proc. 9th Int. Conf. Elect. Mach.
Drives, 1999, pp. 366–370.
[5] H. Harsono, “Photovoltaic water pump system,” Ph.D. dissertation, Dept.
Intell. Mech. Syst. Eng., Faculty Kochi Univ. Technol., Kochi, Japan,
Aug. 2003.
[6] D. Linden, Handbook of Batteries and Fuel Cells. New York, NY, USA:
McGraw-Hill, 1984.
[7] D. Tschanz, H. Lovatt, A. Vezzini, and V. Perrenoud, “A multi-functional
converter for a reduced cost, solar powered, water pump,” in Proc. IEEE
ISIE, 2010, pp. 568–572.
[8] M. A. Vitorino, M. B. R. Correa, C. B. Jacobina, and A. M. N. Lima, “An
effective induction motor control for photovoltaic pumping,” IEEE Trans.
Ind. Electron., vol. 58, no. 4, pp. 1162–1170, Apr. 2011.
[9] S. R. Bowes and A. Midoun, “Suboptimal switching strategies for micro-
processor controlled PWM inverter drives,” Proc. Inst. Elect. Eng.—Elect.
Power Appl., vol. 132, no. 3, pp. 133–148, May 1985.
[10] M. Cacciato, A. Consoli, and V. Crisafulli, “A high voltage gain dc/dc con-
verter for energy harvesting in single module photovoltaic applications,”
in Proc. IEEE ISIE, 2010, pp. 550–555.
[11] P. J. Wolfs, “A current-sourced dc-dc converter derived via the duality
principle from the half-bridge converter,” IEEE Trans. Ind. Electron.,
vol. 40, no. 1, pp. 139–144, Feb. 1993.
[12] P. Wolfs and Q. Li, “An analysis of a resonant half bridge dual converter
operating in continuous and discontinuous modes,” in Proc. IEEE Power
Electron. Spec. Conf., 2002, pp. 1313–1318.
[13] W. Li, L. Fan, Y. Zhao, X. He, D. Xu, and B. Wu, “High step-up and high
efficiency fuel cell power generation system with active clamp flyback-
forward converter,” IEEE Trans. Ind. Electron., vol. 59, no. 1, pp. 599–
610, Jan. 2012.
[14] T.-J. Liang, R.-Y. Chen, J.-F. Chen, and W.-J. Tzeng, “Buck-type current-
fed push-pull converter with ZCS for high voltage applications,” in Proc.
IEEE Region 10 Conf., 2007, pp. 1–4.
[15] P. M. Barbosa and I. Barbi, “A new current-fed, isolated PWM dc-dc
converter,” IEEE Trans. Power Electron., vol. 11, no. 3, pp. 431–438,
May 1996.
[16] R.-Y. Chen, T.-J. Liang, J.-F. Chen, R.-L. Lin, and K.-C. Tseng, “Study
and implementation of a current-fed full-bridge boost dc-dc converter with
zero-current switching for high-voltage applications,” IEEE Trans. Ind.
Appl., vol. 44, no. 4, pp. 1218–1226, Jul./Aug. 2008.
[17] J. Kim, H.-S. Song, and K. Nam, “Asymmetric duty control of a dual-
half-bridge dc/dc converter for single-phase distributed generators,” IEEE
Trans. Power Electron., vol. 26, no. 3, pp. 973–982, Mar. 2011.
[18] B. Liu, C. Liang, and S. Duan, “Design considerations and topology
selection for dc-module-based building integrated photovoltaic system,”
in Proc. 3rd IEEE Conf. ICIEA, Jun. 3–5, 2008, pp. 1066–1070.
[19] J. Biela and J. W. Kolar, “Using transformer parasitics for resonant
converters—A review of the calculation of the stray capacitance of
transformers,” in Conf. Rec. 40th IEEE IAS Annu. Meeting, 2005,
pp. 1868–1875.
[20] L. Yan and B. Lehman, “Isolated two-inductor boost converter with one
magnetic core,” in Proc. IEEE Appl. Power Electron. Conf. Expo., 2003,
pp. 879–885.
[21] Y. Jang, “Two-boost converter,” U.S. Patent 6 239 584, May 29, 2001.
[22] Q. Li and P. Wolfs, “The power loss optimization of a current fed ZVS
two-inductor boost converter with a resonant transition gate drive,” IEEE
Trans. Power Electronics, vol. 21, pp. 1253–1263, Sep. 2006.
[23] W. C. P. De Aragão Filho and I. Barbi, “A comparison between two
current-fed push-pull dc-dc converters—Analysis, design and experimen-
tation,” in Proc. INTELEC, 1996, pp. 313–320.
[24] B. Yuan, X. Yang, X. Zeng, J. Duan, J. Zhai, and D. Li, “Analysis and
design of a high step-up current-fed multiresonant dc-dc converter with
641
low circulating energy and zero-current switching for all active switches,”
IEEE Trans. Ind. Electron., vol. 59, no. 2, pp. 964–978, Feb. 2012.
[25] D. Li, B. Liu, B. Yuan, X. Yang, J. Duan, and J. Zhai, “A high step-up
current fed multi-resonant converter with output voltage doubler,” in Proc.
IEEE Appl. Power Electron. Conf. Expo., 2011, pp. 2020–2026.
[26] Y. Jang and M. M. Jovanovic, “New two-inductor boost converter with
auxiliary transformer,” in Proc. IEEE Appl. Power Electron. Conf. Expo.,
2002, pp. 654–660.
[27] R. Faranda and S. Leva, “Energy comparison of MPPT techniques for
PV systems,” WSEAS Trans. Power Syst., vol. 3, no. 6, pp. 446–455,
Jun. 2008.
[28] KD 200-54 P Series Data Sheet, Kyocera Solar, Inc., San Diego, CA,
USA, 2011.
João Victor Mapurunga Caracas (S’09) received
the B.S. and M.S. degrees in electrical engineering
from the Federal University of Maranhao, São Luís,
Brazil, in 2010 and 2013, respectively, where he is
currently working toward the Ph.D. degree.
He has experience in electrical engineering with
emphasis on power electronics and renewable energy
sources, with focus on grid-connected systems and
control.
Guilherme de Carvalho Farias (S’10) is currently
working toward the B.S. degree in electrical engi-
neering at the Federal University of Maranhão, São
Luís, Brazil.
His main research interests include microinvert-
ers for photovoltaic applications and control of
switched-mode converters. He is currently working
on grid-tied microinverters at the Instituto de Energia
Elétrica.
Luis Felipe Moreira Teixeira (S’10) is currently
working toward the B.S. degree in electrical engi-
neering at the Federal University of Maranhão, São
Luís, Brazil.
His main research interests include the con-
trol of switched-mode converters and grid-tied
microinverters.
Luiz Antonio de Souza Ribeiro (M’98) received the
M.Sc. and Ph.D. degrees from the Federal University
of Paraiba, Campina Grande, Brazil, in 1995 and
1998, respectively.
From 2004 to 2006, he was a Visiting Researcher
at the University of Wisconsin, Madison, WI, USA,
working on sensorless control of permanent-magnet
synchronous machines. Since 2008, he has been an
Associate Professor with the Federal University of
Maranhão, São Luís, Brazil. His main areas of re-
search interest include electric machines and drives,
power electronics, and power systems using renewable energy sources.