You are on page 1of 9

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO.

2, FEBRUARY 2011

403

Cascaded DCDC Converter Photovoltaic


Systems: Power Optimization Issues
Antoneta Iuliana Bratcu, Member, IEEE, Iulian Munteanu, Member, IEEE,
Seddik Bacha, Member, IEEE, Damien Picault, and Bertrand Raison, Member, IEEE

AbstractThis paper investigates the issues of ensuring global


power optimization for cascaded dcdc converter architectures
of photovoltaic (PV) generators irrespective of the irradiance
conditions. The global optimum of such connections of PV modules is generally equivalent with performing the maximum power
point tracking (MPPT) on all the modules. The most important
disturbance occurs when the irradiance levels of modules happen
to be sensibly different from a module to anotherin this case,
voltage-limitation requirements may be broken. The proposed
supervisory algorithm then attempts to establish the best suboptimal power regime. Validation has been achieved by MATLAB/
Simulink numerical simulation in the case of a single-phase gridconnected PV system, where individual MPPTs have been
implemented by an extremum-seeking control, a robust and lessknowledge-demanding perturb-and-observe method.
Index TermsExtremum-seeking control (ESC), maximum
power point tracking (MPPT), photovoltaic (PV) power systems,
supervisory algorithms.

I. I NTRODUCTION

OWADAYS, the world pays growing attention to renewable energy sources, clean and practically inexhaustible,
and interdisciplinary research is continuously developed in
order to sustain the improvement of existing conversion technologies and the development of new ones [1], [2]. Photovoltaic
(PV) installations are an already familiar landscape, either as
small (less than 5 kW) residential stand-alone or even grid
connected, or as larger (hundreds of kilowatts) building integrated or not [3], as well as parts of hybrid power systems,
also containing other renewable energy sources [4], [5]. Interest
is focused on rendering the PV systems more adequately to
the wide use in terms of power, efficiency, grid compliancy
and communication capacity for those grid-connected ones,

Manuscript received June 30, 2009; revised October 28, 2009 and
December 11, 2009; accepted January 27, 2010. Date of publication
February 18, 2010; date of current version January 12, 2011.
A. I. Bratcu and I. Munteanu are with the Grenoble Electrical Engineering
Laboratory (G2ELab), Grenoble National Institute of Technology, 38402,
Grenoble France, and also with the Department of Electronics and Telecommunications, Dunarea de Jos University of Galati, 800008 Galati, Romania
(e-mail: antoneta.bratcu@ugal.ro; iulian.munteanu@ugal.ro).
S. Bacha is with the Grenoble Electrical Engineering Laboratory (G2ELab),
Grenoble National Institute of Technology, 38402 Grenoble, France, and
also with Joseph Fourier University, 38041 Grenoble, France (e-mail:
bacha@g2elab.grenoble-inp.fr).
D. Picault is with Grenoble Electrical Engineering Laboratory (G2ELab),
Grenoble National Institute of Technology, 38402 Grenoble, France (e-mail:
picault@g2elab.grenoble-inp.fr).
B. Raison is with Joseph Fourier University, 38041 Grenoble, France (e-mail:
raison@g2elab.grenoble-inp.fr).
Digital Object Identifier 10.1109/TIE.2010.2043041

reliability and service time, safety and security, etc. [6][9]. A


PV system usually undergoes an optimization procedure guided
by various and often contradictory criteria, in order for it to
result as the best tradeoff within a specific application [10].
Exploitation of PV systems has proved the necessity of
methods of automatic control and information processing for
optimizing dynamic performance, reactivity to the variability
of the primary energy source, i.e., the light, and robustness to
as various as possible kinds of disturbances. In the specific case
of grid applications, one of the most important aspects is the
maximum power point tracking (MPPT), aiming at maximizing
the extracted energy irrespective of the irradiance conditions
[11]. An important research effort has been devoted to finding
simple, efficient, and minimal-knowledge-demanding methods
of MPPT [12][14]. Among them, the so-called perturb-andobserve (P&O) class of methods, based upon injecting highfrequency small-amplitude (usually harmonic) perturbations in
the system in order to detect the sign of the power gradient, has
proved as one of the most successful [15], [16].
One can say that the MPPT is practically solved in the case
of a PV system whose cells receive the same irradiance level.
Such a system has a unimodal power characteristic, which may
not be the case of large spatially distributed systems, with
high probability of undertaking partial shading. Indeed, specific
configurations of PV modules may have global power characteristics that exhibit multiple maxima [17]. Recent works reflect
the interest of designing MPPT methods able to track multiple
peaks under rapidly changing irradiance conditions [18], [19].
One such particular topology is the object of this paper.
Several PV generators, with each coupled to its dcdc boost
converter, are cascaded on the same dc bus and interfaced to
the grid by means of a dcac inverter. The problem is to find a
robust control strategy of extracting the maximum power available from this architecture, given that the PV generators may
undertake supplementary constraintsexpressed mainly as
output-voltage-limitation (OVL) requirementswhen exposed
to strongly variable irradiance conditions. This problem was
first addressed in [20]here, more extended discussion of the
results and comparison with other configurations are provided.
This paper is organized as follows. In Section II, the
structure, operation, and steady-state analysis of the considered
cascaded dcdc converter PV topology are presented, with emphasis on advantages versus some other more usual cases. Some
simulation results suggest the problems that can occur when operating all the PV generators at MPPT but under quite different
levels of irradiance. Hence, a more complex control structure,
called a supervisor, is necessary in order to achieve the global

0278-0046/$26.00 2011 IEEE

404

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 2, FEBRUARY 2011

Fig. 1. Block diagram of the considered cascaded PV architecture.

power optimization in the cascaded case, according to how it


is detailed in Section III. Numerical simulation results of the
supervised PV system, including the comparison with a parallel
dcdc converter configuration, are presented in Section IV.
This paper ends with the conclusions and future work.
II. C ASCADED DCDC C ONVERTER PV A RCHITECTURE
A. Structure, Operation, and Steady-State Analysis
The analyzed PV system consists of a series topology of
per-panel dcdc converters. A high-voltage string connected
to a single dcac inverter results, having the advantages of
a converter-per-panel approach, obviously cheaper but also
more efficient than individual dcac grid-connected inverters
[21]. The buck and boost converters are the most efficient
topologies for a given cost, as shown in [21]. Here, boost dcdc
converters have been employed (Fig. 1).
The considered structure has some intuitive advantages over
the parallel-connection case. Thus, for a given dc-bus voltage, the series connection allows individual boost converters
with relatively small step-up ratio being used (usually three
to four), which ensures good efficiency along with low cost.
Unlikely, the parallel case would require higher step-up ratios
(more than four), which leads either to poorer efficiency or to
higher cost due to eventually using two-stage dcdc conversion.
Constraints regarding current limitation may also be involved.
Both of these drawbacks may however be avoided if steppingup the PV output voltage by series connections of many panels
before the dcdc conversion stage. However, series connections
of PV panels are disadvantageous from the viewpoint of the
possibility of harvesting the maximum power because they are
limited by the power given by the worst irradiated panel. Thus,
the total power provided risks to diminish in the case of, e.g.,
partial shading on some panels. A numerical-simulation-based
comparison between the cascaded case and a parallel case of
the same power can be found in Section IV.
Fig. 1 shows n dcdc converter modules supposed identical
connected in series to a single dcac inverter in charge with
transferring the power to the grid. Each converter module i is

based on the PV generator PVi , consisting of either a single


panel or a connection of panels, supposed to receive all the
same irradiance, such that the power characteristic remains
unimodal. To this, the boost chopper i is associated. Such a
PV generator can independently control and thus optimize the
power flow from its source of irradiance Irri : The pulsewidth
modulation (PWM) signal uchi {0, 1} of chopper i can be
controlled in order to impose the operating point of PVi , i.e.,
(IPVi , VPVi ), independently from the operation of the other
generators.
In grid applications, each chopper performs MPPT for its PV
generator, while the grid inverter regulates the dc-bus voltage.
At steady state, the same current Idc passes through all the
choppers,with the sum of their output voltages being the dc-bus
n

voltage
i=1 Voi = Vdc . Let Vdc denote the desired dc-bus

voltage value and (IPVi , VPVi ), i = 1, 2, . . . , n, be the n operating points. The steady-state values of all variables in Fig. 1
can be deduced [20]. Thus, the chopper output voltages are
 n


Voi = Vdc
IPVi
VPVi
IPVi
VPVi
(1)


i=1



wi

showing that the dc-bus voltage distribution on the converter


modules depends on the weights of the individual PV powers
in the global power provided wi , i = 1, 2, . . . , n.
The PV power varies directly with the irradiance level [22].
Hence, as long as the irradiance levels are almost the same for
all PV generators, one can impose practically the same operating pointin particular, the maximum power oneensuring
that the dc-bus voltage be almost equally supported by the n
dcdc converters and the total power be maximized. However,
when the irradiance levels are sensibly different for some time,
then the weights wi become unbalanced, and some dcdc
converters will undergo overvoltages, as shown next.
B. Simultaneous MPPT for All PV Generators
As widely known, the power curve of a PV module versus the
module voltage PPV (VPV ) is unimodal [9], [22]. Let the locus
of MPPs corresponding to various irradiance levels be called the
maximum power regime (MPR) curve. MPPT operation means
to track the MPR irrespective of the irradiance level. In this
paper, the MPPT on each PV generator is implemented by an
extremum-seeking control (ESC), a P&O method that is able to
find the extremum of some difficult-to-model and not precisely
known unimodal dynamics by feeding the plant with sinusoidal
probing signals [23].
Fig. 2 shows how the ESC can be applied to PV modules.
When feeding a PV module with a -frequency sinusoidal voltage variation having a sufficiently small amplitude a, a power
variation is induced, which is generally nonsinusoidal due to the
characteristics nonlinearity and has a -first-order harmonic.
This latter can be extractede.g., by means of a high-pass
filter called washout filterand it is in phase with the voltage
variation if the operating point is on the ascending part of the
voltagepower curve and with a phase lag of for the descending part. The product of the two signals has two components:

BRATCU et al.: CASCADED DCDC CONVERTER PHOTOVOLTAIC SYSTEMS: POWER OPTIMIZATION ISSUES

Fig. 2.

405

ESC principle applied for the PV modules.

a small-amplitude 2-frequency component, which can be neglected, and a continuous component, which will toggle its sign
as the operating point moves from a side to the other of the PPV
curves maximum. This latter component is then passed through
an integrator of gain k in order to provide the voltage step VPV
that is necessary to move the operating point to the optimal
position with a convergence speed depending proportionally on
k, a, and 1/ [23]. The voltage reference applied to the PV
module results as VPV = VPV + a sin(t).
As regards how an ESC controller must be tuned, some
general considerations can be applied as follows. Thus, the
excitation frequency must be sufficiently large, i.e., outside
the plants bandwidth, in order to ensure stability of the closedloop system. On the other hand, in systems having power
electronics devices, this frequency is upper limited by the
power electronics switching frequency. The amplitude a must
be sufficiently small such that the plants behavior remains
linear and the induced output variations are as sinusoidal as
possible; therefore, a must be chosen depending on the slope
of the plants VPV PPV characteristic. The integrator gain k
results from upper limiting the reference gradient to the value
of the plants main dynamic such that the plant can follow
the reference. Note that the VPV PPV characteristic is time
varying; therefore, the controller can either be tuned on a most
typical curve, or one can attempt using adaptive laws of control
parameter computation.
MATLAB/Simulink numerical simulations have been performed for a three-generator PV topology as in Fig. 1, where
each PV generator is independently set in MPPT by ESC,
under strongly and rapidly variable irradiance conditions. The
PWM signal uchi results within a voltage control loop, whose

comes from an ESC-based MPP tracker using


reference VPVi
the measured PV generator power PPVi . Each PV generator is
a series connection of two (125 125)-mm-cell PV modules,
each of 150-W peak power, 5.3-A short-circuit current, 59-V
open-circuit voltage, and 40-V typical voltage at typical power.
The well-known PV-cell five-parameter diode model has been
used [24]

IPV = Iph I0 e(VPV +Rs IPV )/Vt 1


(VPV + Rs IPV ) /Rsh

(2)

Fig. 3. Performance of the analyzed PV architecture when all the generators are simultaneously operated in MPPT under rapidly variable irradiance
conditions.

where Iph is the current due to the PV effect, I0 is the diode


saturation current, Vt is the thermic voltage, and Rs and Rsh
are the series and shunt resistances, respectively.
A requirement imposed when sizing the PV system is that the
ratio between the ideal value of the dc-bus voltage and the rated
value of the output chopper voltage Vr to approach the number
ideal
of PV generators. In this case, Vdc
= 450 V in order to allow
interfacing to a 220-V/50-Hz grid; provided that n = 3, Vr can
ideal
/n, i.e., 150 V.
be chosen around Vdc
The irradiance signals are delayed from a PV generator to
the next one by a time interval, thus simulating a cloud passing
over. A two-spectral-component dynamic model of the irradiance in analogy with the modeling of another irregular renewable energy source, the wind speed, has been used [20], [25].
Here, the variation speed has been set around 1 s, representing
faster variations than habitually occurring [see Fig. 3(a)]. Each
ESC controller has been tuned thus: the integrator constant is
k = 100; the sinusoidal disturbance has amplitude a = 0.1 and
frequency = 100 Hz.
Initially, to all the PV generators, the same voltage reference

= 65 V, which is a value arbiis imposed (in this case, VPV


trarily chosen somewhere in the middle of the voltage variation
range); then, the control is switched to MPPT. The results of the
simultaneous MPPT operation of all generators can be noted
in Fig. 3(b). When the irradiance values became maximally
unbalanced, then the output chopper voltages of the first and
the third generator, i.e., Vo1 and Vo3 , have gone well beyond
the rated value Vr = 150 V.
The values of the capacitors have been chosen as CPV =
4700 F and Cdc = 22 000 F. These quite high values are
motivated by control reasons, i.e., for smoothing the voltage
variations due to fast changes of the primary resource. Thus,
the CPV capacitors ensure a reasonably smoothed PV output
voltage variation. The value of Cdc prevents a sudden imbalance of the individual irradiation levels from propagating too
fast in the choppers output voltage. In this way, enough time is
eventually given to a supervisory structure for taking decisions,
as shown in the next section. The cost of capacitors may also

406

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 2, FEBRUARY 2011

be involved in defining the best tradeoff in order to choose the


capacitors values, but this is beyond the scope of this paper.
More details on both the ESC design and simulation results
can be found in [20]. In conclusion, the simultaneous MPPT
operation of all generators within a grid-connected cascaded
PV topology is effective only when the PV generators are
illuminated almost equally. Otherwise, constraints of voltagelimitation type can be violated. The global power optimization
requires a supervisor that detects a sufficient degree of degrading the initial optimal strategy in order to meet the constraints.
III. S UPERVISORY A LGORITHM P ROPOSED FOR
G LOBAL P OWER O PTIMIZATION
The maximum power capture irrespective of the irradiance conditions, meanwhile meeting the voltage-limitation constraints, may be ensured if each PV generator is equipped with
two control laws instead of a single one, the MPPT. The main
control law remains the MPPT, with the second one being an
OVL control law. A control system, which is able to globally
manage the different events and then to take decisions of enabling one of the two control ways for each PV generator, must
be designed. Such a systemcalled a supervisordoes not
need to run continuously, as it looks rather at variation trends
in a relevant time window than at instantaneous variations.
Thus, the supervising algorithm runs every Tsv s, where Tsv
is suitably chosen depending on the fastest dynamic of the
supervised system.
Assuming that the normal operation is MPPT on all the
generators, the supervisor must be able to detect OVL violations
due to unbalance between the power weights of generators and,
thus, between their output voltages Voi , i = 1, 2, . . . , n, as in
(1). The admissible threshold is here set to 1.2Vr . When at least
one such limitation is detected, then a new possible value of the
dc-bus voltage reference is first sought for, which results such
that to reestablish the balance between voltages, meanwhile
ideal
. If such a
needing to lay in between 20% from Vdc
value could not be found out, then the supervisor switches
from MPPT to OVL for the generators having surpassed the
admissible threshold voltagethe other generators remain in
MPPT. If the power weight of a generator operating in OVL
decreases, this means that the global power has increased. This
can only be due to the increasing of the power provided by
the generators still operating in MPPT. Furthermore, this means
that the irradiance balance is going to be reestablished, and the
MPPT is again possible. The control of the concerned generator
can then be switched back to MPPT based on estimating the
gradient of the power weight.
The block diagram of the supervisor, together with its connection with the choppers and inverters control blocks, is
shown in Fig. 4 (see also [20]). The PV power measures and
the output voltage measures PPVi and Voi , i = 1, 2, . . . , n, as
well as the dc-bus voltage measure Vdc , are conveniently lowpass filtered to reflect the trend on the desired time window.
The power weights wi , i = 1, 2, . . . , n, and their gradients
are computed inside the supervisor. The supervisor outputs n
binary decisions referring to the operating modes of the n PV
generators: Output i is 1 for MPPT and 0 for OVL. Switching

Fig. 4. Block diagram of the supervisor that implements the proposed global
power optimization strategy and its connections with the choppers and inverters control blocks.

between the two controls means, in fact, only changing the


voltage reference.
These outputs feed block 1 in Fig. 4, which generically
denotes the n chopper control structures; these latter further
provide the PWM signals uchi , i = 1, 2, . . . , n, to the choppers.
The supervisor also computes a new value of the dc-link voltage
new
, which is further provided to the inverter
reference Vdc
control block (block 2 in Fig. 4). The PWM signal uinv is finally
new
computation block also outputs a
sent to the inverter. The Vdc
new
binary variable noted Vdc found in Fig. 4, which, together
with wi and dwi /dt, i = 1, 2, . . . , n, is used within a finitestate automaton in order to provide the operating modes of the
PV generators.
new
is computed as follows. If at least one
The value of Vdc
PV generator has surpassed the maximum voltage threshold,
this means that the power available from another generator (or
maybe from many others) has decreased. A new reduced dcnew
is sought for, which allows for all
bus voltage reference Vdc
the generators to remain in MPPT, meanwhile meeting all the
new constraints Voinew 1.2 Vr , i = 1, 2, . . . , n. Taking into
account (1), it is sufficient to reduce the dc-bus reference to
new
= 1.2 Vr
Vdc

max {wi }.

i=1,2,...,n

(3)

Note that (3) is valuable in steady state, where the PV


power generators reflect exactly the irradiance levels. Thus, the
weights wi are the weights of Irri in the sum of Irri . The
value (3) can be declared the new dc-bus voltage reference
ideal
.
for the inverter control loop if it is larger than 0.8 Vdc
If not, the supervisor degrades the MPPT of the generators
having surpassed the limit by switching their control to OVL at
1.2Vr . This is a suboptimal regime from the energy viewpoint;
however, it is the less restrictive degradation as it still ensures
the constraints to be met at the limit.
In Fig. 4, it is also depicted that the supervisor is a cyclic
task, executed every Tsv s.

BRATCU et al.: CASCADED DCDC CONVERTER PHOTOVOLTAIC SYSTEMS: POWER OPTIMIZATION ISSUES

407

Fig. 5. Performance of the supervised cascaded PV system under step changes of the irradiance levels. (a) Irradiance scenario. (b) Output voltages of the three
choppers. (c) Time evolutions of the dc-bus voltage and grid current. (d)(f) Time evolutions of the power provided by the three PV generators.

The steps of the supervising algorithm are listed as follows.


For all PV generators
Compute their power weights wi , i = 1, 2, . . . , n.
Estimate the gradients of their power weights gi , i =
1, 2, . . . , n.
If there is at least one generator that has surpassed 1.2Vr , then
Compute a new dc-bus voltage reference, as to (3).
new
ideal
ideal
 [0.8 Vdc
; 1.2 Vdc
], then
If Vdc
For those PV generators having surpassed 1.2Vr
Set their control to OVL.
new
as reference to the inverter control.
Otherwise, send Vdc
Otherwise
For all generators operating in OVL that have negative
gradients of power weights
(Re)set their control to MPPT.
IV. VALIDATION BY N UMERICAL S IMULATION
A. Cascaded Case With Supervisor
The supervised cascaded dcdc converter PV system described in Section II has been numerically simulated under
the irradiance scenario shown in Fig. 5(a), with all the PV
generators initially operating at MPPT, under equal irradiance
levels, i.e., Irr1 = Irr2 = Irr3 = 900 W/m2 . The most unfavorable case has been considered, with step variations of
irradiance, which are seldom encountered in nature. After Irr1
decreased from 900 to 400 W/m2 , the output voltages Vo2 and
Vo3 reach the maximum threshold of 1.2Vr = 180 V [Fig. 5(a)].
According to (3), a new value of the dc-bus voltage reference
new
results around 1.2 150/(900/2200) = 440 V, which is
Vdc
ideal
= 360 V [Fig. 5(c)].
acceptable as it is larger than 0.8 Vdc
Relation wi = Irri /(Irr1 + Irr2 + Irr3 ), i = 1, 2, 3, has
been used for computing the weights wi . One must note that

(3) provides a theoretical level of the new dc voltage, as the


measured PV powers and, therefore, their weights wi are supposed to be at their steady-state values. The abrupt variations of
the irradiations on the concerned PV generators determine that
the optimal power levels change; therefore, the MPPTs need
some time to readjust; moreover, the PV power signals will
always exhibit ripples due to the MPPT algorithms. It is very
probable that this time cannot be waited, so the supervisor must

decide the change of Vdc


faster than the MPPTs reach their new
steady state. Thus, the weights wi will intervene in (3) with
some transient values; therefore, the computation will give a
new
slightly different value of Vdc
in relation to the theoretical
one. This difference can be noted on the zoom in Fig. 5(c).
This new reference must be low-pass filtered and then imposed
to the inverter in order to prevent too large values of the grid
current. Between t2 and t3 , all the PV generators operate in
MPPT [Fig. 5(d)(f)]. The zooms in Fig. 5(e) and (f) show that,
once the dc-bus voltage decreased, the searching noise due to
MPPT has slightly increased. At time t3 , Irr1 comes back to its
initial level, and the generators remain in MPPT until t4 , when
both Irr1 and Irr2 decrease abruptly and significantly, from
900 to 400 and to 500 W/m2 , respectively. Shortly after, only
Vo3 reaches the threshold of 180 V, but a new dc-bus voltage
reference cannot be found anymore. Consequently, the control
of the third PV generator is switched from MPPT to OVLthe
evolution of Vo3 and of the power provided by this generator
can be seen in Fig. 5(b) and (f), respectively. The other two
generators continue to operate in MPPT, obviously at reduced
power levels [Fig. 5(d) and (e)].
One can note that this latter case corresponds to a particular
situation. The new dc voltage value computed according to (3)
would be 1.2 150/(900/1800) = 360 V if using the steadystate values of wi . On the other hand, 360 V is exactly the lower
ideal
admissible threshold for Vdc , i.e., 360 V = 0.8 Vdc
.

408

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 2, FEBRUARY 2011

Therefore, the supervisor would theoretically allow the


MPPT operation of all PV generators. However, the values
of wi effectively used by the supervisor are not the steadystate ones, for reasons already mentioned. Moreover, the abrupt
decreases of irradiance on generators 1 and 2 at t4 make
PPV1 and PPV2 used for computing wi in (3) to be smaller
than the steady-state ones [Fig. 5(d) and (e)]. Therefore, the
new
maximum of wi results are larger than 900/1800, and Vdc
results are strictly smaller than 360 V. Thus, the supervisors
actual decision in this particular case is to degrade the MPPT
on PV generator 3, making sure that voltage constraints are
met. Note that slower-than-step variations of irradiance and
smaller transients are more probable in nature; in this case, the
supervisor is more likely to decide keeping the MPPT operation
by reducing the dc-bus voltage.
One can note that the dynamic performance of the MPPT
control law is better at negative than at positive variations
of irradiance, due to the fact that the VPV PPV curve is not
symmetric in relation with its maximum, so the dynamics on
the rising side are different from the ones on the falling side.
Note also that the MPPT can be speeded up/slowed down by
setting the integrator constant k larger/smaller.
B. Comparison With Parallel Configuration
Numerical simulation can be used in order to perform a comparison between the supervised cascaded case and a parallel
configuration, in which the individual choppers are connected
in parallel on a common dc link. The simulations will aim at
comparing the energy efficiencies of the two PV systems under
both balanced and imbalanced individual irradiance levels. To
this end, modeling of losses is necessary.
A correct comparison is ensured if the two PV systems
have the same power, each PV generator can be individually
controlled by means of similar choppers, and the grid inverter
is the same. Taking into account the size and ratings of the previously described cascaded PV system, a parallel configuration
allowing the same power to be provided, independent controls
of PV generators and power injection into the grid by means
of the same inverter must embed a two-stage dcdc conversion.
The block diagram of such a grid-connected PV architecture
can be seen in Fig. 6. Boost choppers 1n are controlled to
maintain the PV generators in MPPT; they are connected in
parallel on the first common dc link, denoted by dc-link 1.
Its voltage must be maintained constant at V0 by the common
chopper. The voltage of the second dc link Vdc is maintained
constant by the grid inverter, which is also in charge with the
power injection at the grids parameters.
The numerical simulation setup has been established as
follows. Both the cascaded and the parallel PV systems contain
three PV generators. The PV choppers are the same like those
in the cascaded case and the same as the common chopper;
moreover, their control parameters are the same (for details,
see the Appendix). The final dc-link voltage is the same as
that in the cascaded case, i.e., Vdc = 450 V, for reasons of
grid coupling and of using the same grid inverter, whereas the
intermediary dc-link voltage can be set to V0 = 150 V in order
to make use of the choppers step-up ratios around three. Only
the losses in choppers have been modeled such as to vary with

Fig. 6. Block diagram of a parallel dcdc converter PV architecture with two


dcdc conversion stages.

the operating point [26]because the two architectures use the


same grid inverterand thus, the comparison will refer to the
final dc-link power as the output power.
The irradiance scenario used in the simulation is shown in
Fig. 7(a), where, initially, all irradiances take the reference values Irr1 = Irr2 = Irr3 = 1000 W/m2 . On a 25-s time horizon, three events take place: at time 5 s, Irr1 decreases abruptly
to 400 W/m2 ; at time 10 s, the irradiances become again equal
and maximum; and finally, at time 17 s, both Irr1 and Irr2
exhibit sudden decreases to 400 and 500 W/m2 , respectively.
In Fig. 7(b), one can see the time evolutions of the power of
interest, namely, the optimal power totally delivered by the PV
generators and the final dc-link power in the two cases.
The chosen parallel configuration can deliver the maximum
of power irrespectively of the irradiance conditions because
there are no supplementary constraints to meet; each PV generator injects the current corresponding to its maximum power in
the dc link. As regards the cascaded case, this one can operate
under imbalanced irradiance levels with all the PV generators
in MPPT as long as the supervisor can find a reduced dc-link
voltage in order to allow the voltage constraints being met. One
can see such evolution between moments 5 and 10 s in Fig. 7(b)
and zoom 1). The optimal power level is around 840 W. The
dc-link power of the cascaded case is largeraround 790 W
than that provided by the parallel casearound 760 W
obviously because the latter one has two stages of losses. One
can also identify the moment when the dc-link voltage reference
has been reduced, as the variations of the dc-link power have
increased due to the increase of the MPPT search noise. The
efficiency of the cascaded configuration is, in this case, superior
to that of the parallel case (0.94 versus 0.9).
In conditions of equal and maximum irradiance levels, the
efficiencies of the two PV architectures are even more clearly
distinguished (0.95 versus 0.89, see zoom 2) in Fig. 7).
The global efficiency of the cascaded case becomes inferior
to that of the parallel case when the irradiances are strongly
imbalanced and the simultaneous MPPT operation of all PV
generators is not possible anymore. Such a situation is shown
after moment 17 s in Fig. 7(b) and zoom 3). The total optimal

BRATCU et al.: CASCADED DCDC CONVERTER PHOTOVOLTAIC SYSTEMS: POWER OPTIMIZATION ISSUES

Fig. 7.

409

Comparison between the cascaded case and the parallel case under the same irradiance scenario from the point of view of the energy efficiency.

power is around 655 W. In this case, the supervisor of the


cascaded system could not find a suitable reduced value of the
dc voltage, and thus, one of the MPPTs had to be degraded.
Consequently, the global efficiency has diminished to about
0.76 versus 0.93 of the parallel configuration, in which the PV
generators remain all in MPPT.
Some generalizations can be attempted at the end of this
analysis. The two PV configurations are comparable from the
point of view of the energy efficiency under relatively normal
irradiance conditions, with a more obvious advantage of the
cascaded configuration as the number of dcdc conversion
stages in the parallel one is larger. When the individual irradiance levels are sensibly different, the supervisor still potentially
ensures a superior efficiency of the cascaded case versus the
parallel one. If the degradation of MPPT operation on some
PV generators is absolutely necessary, then the order relation
between the two efficiencies is likely to change, the clearer
as the number of MPPT degradations is larger. A constant
advantage of the cascaded configuration versus the parallel
one is its lower cost due to using a smaller number of power
electronics devices and a relatively cheap digital structure to
support the supervisor implementation.
V. C ONCLUSION
This paper has investigated how the power optimization can
be achieved for cascaded dcdc converter PV architectures.
Such kinds of systems ensure good efficiency, along with quite
low cost, when compared to parallel configurations, mainly
because the individual choppers are not required to have high
boost ratios. Conversely, the simultaneous operation of all PV
generators at MPPT, in order to maximize the global power,

supposes supplementary constraints that are met as long as the


generators receive almost the same irradiance level. When this
is not the case anymore, dc-bus overvoltage may arise, as shown
by the steady-state analysis and numerical simulations. A control strategy ensuring robustness to rapidly variable irradiance
conditions and respect of the voltage-limitation constraints was
proposed. Thus, each PV generator was equipped with a second
control law, aiming at limiting the choppers output voltage.
A supervisor was designed, which, for each PV generator,
switches between the two control laws in order to establish
the best power regime possible under the given constraints.
MATLAB/Simulink numerical simulations were performed in
the case of a single-phase grid-connected PV system, where
individual MPPTs were implemented by a simple and robust
ESC method.
The main points of interest in the future concern the real-time
validation on dedicated experimental rigs and, as theoretical
development, possible generalizations envisaging global power
optimization by coordinated control strategies.
A PPENDIX A
PV generator features:
Two series-connected modules of 8 9 cells
Cell parameters: (125 125)-mm 5.3-A short-circuit
current
Module open-circuit voltage 59 V
Typical voltage 40 V at typical power 150 W
Power electronics switching frequency: f0 = 10 kHz
Choppers features:
PV capacitor: capacity CPV = 4700 F, resistance
RPV = 100 k

410

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 2, FEBRUARY 2011

Chopper circuit parameters: Lch = 8 mH, Rch = 320 m


Power MOSFETs: rising time tr = 100 ns, falling time
tf = 50 ns, Rds_on = 200 m
Schottky diodes: Vd = 0.47 V
Grid inverter features:
Grid voltage: 230 V/50 Hz
Grid parameters: Lgrid = 1 mH, Rgrid = 170 m
Two-leg power MOSFETs and Schottky diodes
Final dc-link voltage: Vdc = 450 V
Final dc-link capacitor: cascaded: three series-connected
Cdc = 22 000 F; parallel: Cdc = 4700 F
Intermediary dc-link features (dc-link 2 parallel case):
Voltage V0 = 150 V
Capacitor C0 = 4700 F
Resistance R0 = 100 k
Choppers control parameters:
Nonlinear current controller: gain = 10, hysteresis
width = 0.25
PI voltage controller: proportional gain Kp = 10, integral
gain Ki = 250
ESC MPPT control parameters:
a = 1, k = 100, = 100 Hz, filter bandwidth 50 Hz
Grid inverter control parameters:
Nonlinear current controller: gain = 10, hysteresis
width = 0.25
PI voltage controller: proportional gain Kp = 10, integral
gain Ki = 75
Supervisor: Tsv = 0.1 ms
ACKNOWLEDGMENT
This work has been partially developed within the framework
of the Solution PV Project, financed by the French National
Research Agency (ANR), to whom the authors are kindly
grateful.
R EFERENCES
[1] R. Coenraads, G. Reece, M. Voogt, M. Ragwitz, A. Held, G. Resch,
T. Faber, R. Haas, I. Konstantinaviciute, J. Krivosk, and T. Chadim,
Promotion and Growth of Renewable Energy Sources and Systems,
Final Report, European Project PROGRESS, Ctrc. No. TREN/D1/422005/S07.56988, European Commision, Utrecht, The Netherlands, 2008.
[Online]. Available: http://ec.europa.eu/energy/renewables/studies/doc/
renewables/2008_03_progress.pdf
[2] J. M. Carrasco, L. G. Franquelo, J. T. Bialasiewicz, E. Galvn,
R. C. Portillo Guisado, Ma. . Martacute;n Prats, J. I. Len, and
N. Moreno-Alfonso, Power-electronic systems for the grid integration
of renewable energy sources: A survey, IEEE Trans. Power Electron.,
vol. 53, no. 4, pp. 10021016, Jun. 2006.
[3] B. Marion, J. Adelstein, K. Boyle, H. Hayden, B. Hammond, T. Fletcher,
B. Canada, D. Narang, D. Shugar, H. Wenger, A. Kimber, L. Mitchell,
G. Rich, and T. Townsend, Performance Parameters for Grid-Connected
PV Systems, Nat. Renewable Energy Lab., Golden, CO, NREL/CP-52037358. [Online]. Available: http://www.nrel.gov./docs/fy05osti/37358.pdf
[4] S. Jemeuml;, D. Hissel, A.-S. Coince, and B. Al-Nasrawi, Optimization
and economic analysis of an hybrid fuel cell, photovoltaic and battery
electric power generation system, J. Fuel Cell Sci. Technol., vol. 3, no. 4,
pp. 410414, Nov. 2006.
[5] L. Wang and C. Singh, Multicriteria design of hybrid power generation
systems based on a modified particle swarm optimization algorithm,
IEEE Trans. Energy Convers., vol. 24, no. 1, pp. 163172, Mar. 2009.
[6] E. L. Meyer and E. D. van Dyk, Assessing the reliability and degradation of photovoltaic module performance parameters, IEEE Trans. Rel.,
vol. 53, no. 1, pp. 8392, Mar. 2004.

[7] A. Fernndez-Infantes, J. Contreras, and J. L. Bernal-Agustacute;n, Design of grid connected PV systems considering electrical, economical
and environmental aspects: A practical case, Renew. Energy, vol. 31,
no. 13, pp. 20422062, Oct. 2006.
[8] J.-H. Shi, X.-J. Zhu, and G.-Y. Cao, Design and techno-economical
optimization for stand-alone hybrid power systems with multi-objective
evolutionary algorithms, Int. J. Energy Res., vol. 31, no. 3, pp. 315328,
2007.
[9] E. Romn, R. Alonso, P. Ibaez, S. Elorduizapatarietxe, and D. Goitia,
Intelligent PV module for grid-connected PV systems, IEEE Trans. Ind.
Electron., vol. 53, no. 4, pp. 10661073, Jun. 2006.
[10] M. Kolhe, Techno-economic optimum sizing of a stand-alone solar
photovoltaic system, IEEE Trans. Energy Convers., vol. 24, no. 2,
pp. 511519, Jun. 2009.
[11] V. Salas, E. Olacute;as, A. Barrado, and A. Lzaro, Review of the
maximum power point tracking algorithms for stand-alone photovoltaic
systems, Sol. Energy Mater. Sol. Cells, vol. 90, no. 11, pp. 15551578,
Jul. 2006.
[12] W. Xiao, W. G. Dunford, P. R. Palmer, and A. Capel, Application of centered differentiation and steepest descent to maximum power point tracking, IEEE Trans. Ind. Electron., vol. 54, no. 5, pp. 25392549, Oct. 2007.
[13] C. Rodriguez and G. A. J. Amaratunga, Analytic solution to the photovoltaic maximum power point problem, IEEE Trans. Circuits Syst. I,
Reg. Papers, vol. 54, no. 9, pp. 20542060, Sep. 2007.
[14] V. V. R. Scarpa, S. Buso, and G. Spiazzi, Low-complexity MPPT
technique exploiting the PV module MPP locus characterization, IEEE
Trans. Ind. Electron., vol. 56, no. 5, pp. 15311538, May 2009.
[15] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, Optimization of perturb and observe maximum power point tracking method, IEEE Trans.
Power Electron., vol. 20, no. 4, pp. 963973, Jul. 2005.
[16] D. Sera, R. Teodorescu, J. Hantschel, and M. Knoll, Optimized maximum power point tracker for fast-changing environmental conditions,
IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 26292637, Jul. 2008.
[17] H. Patel and V. Agarwal, MATLAB-based modeling to study the effects of partial shading on PV array characteristics, IEEE Trans. Energy
Convers., vol. 23, no. 1, pp. 302310, Mar. 2008.
[18] H. Patel and V. Agarwal, Maximum power point tracking scheme for PV
systems operating under partially shaded conditions, IEEE Trans. Ind.
Electron., vol. 55, no. 4, pp. 16891698, Apr. 2008.
[19] L. Gao, R. A. Dougal, S. Liu, and A. P. Iotova, Parallel-connected solar
PV system to address partial and rapidly fluctuating shadow conditions,
IEEE Trans. Ind. Electron., vol. 56, no. 5, pp. 15481556, May 2009.
[20] A. I. Bratcu, I. Munteanu, S. Bacha, D. Picault, and B. Raison, Power
optimization strategy for cascaded DCDC converter architectures of photovoltaic modules, in Proc. IEEE ICIT, Churchill, Australia, Feb. 1013,
2009, pp. 18, [CD-ROM].
[21] G. R. Walker and P. C. Sernia, Cascaded DCDC converter connection
of photovoltaic modules, IEEE Trans. Power Electron., vol. 19, no. 4,
pp. 11301139, Jul. 2004.
[22] K. Ujiie, T. Izumi, T. Yokoyama, and T. Haneyoshi, Study on dynamic
and static characteristics of photovoltaic cell, in Proc. Power Convers.
Conf., Apr. 25, 2002, vol. 2, pp. 810815.
[23] K. B. Ariyur and M. Krsti, Real-Time Optimization by Extremum Seeking
Control. New York: Wiley-Interscience, 2003.
[24] R. A. Messenger and J. Ventre, Photovoltaic Systems Engineering, 2nd
ed. Boca Raton, FL: CRC Press, 2003.
[25] T. Burton, D. Sharpe, N. Jenkins, and E. Bossanyi, Wind Energy
Handbook. New York: Wiley, 2001.
[26] I. Batarseh, The power MOSFET, in Power Electronics Handbook,
M. Rashid, Ed., 2nd ed. New York: Academic, 2007, pp. 4169.
Antoneta Iuliana Bratcu (M07) received the M.S.
degree in electrical engineering from Dunarea
de Jos University of Galati, Galati, Romania, in
1996 and the Ph.D. degree in automatic control
and informatics from Universit de Franche-Comt,
Besanon, France, in 2001.
Between 2007 and 2009, she was a Postdoctoral
Researcher with the Grenoble Electrical Engineering
Laboratory (G2ELab), Grenoble National Institute of
Technology, Grenoble, France. She is currently an
Associate Professor with the Department of Electronics and Telecommunications, Dunarea de Jos University of Galati. Her
major fields of study include discrete and continuous optimization and hybrid
dynamical systems with application to energy conversion systems and management of industrial systems.

BRATCU et al.: CASCADED DCDC CONVERTER PHOTOVOLTAIC SYSTEMS: POWER OPTIMIZATION ISSUES

Iulian Munteanu (M07) received the M.S. degree


in instrumentation and control from Universit du
Havre, Le Havre, France, in 1997 and the Ph.D.
degree in automatic control from Dunarea de Jos
University of Galati, Galati, Romania, in 2006.
In 2002, he was a Fellow with the Marie
Curie European Framework Program, Laboratoire
dlectrotechnique de Grenoble, France, which
became the Grenoble Electrical Engineering Laboratory (G2ELab), where he was a Postdoctoral Researcher between 2007 and 2009. He is currently
a Lecturer with the Department of Electronics and Telecommunications,
Dunarea de Jos University of Galati. His research work is oriented toward
optimal control of renewable energy systems.

Seddik Bacha (M08) received the B.E. and M.S.


degrees from cole Nationale Polytechnique de Algiers, Algiers, Algeria, in 1982 and 1990, respectively, and the Ph.D. degree from the Grenoble
National Institute of Technology, Grenoble, France,
in 1993.
In
1990,
he
joined
the
Laboratoire
dlectrotechnique de Grenoble. In 1998, he was
habilitated to conduct research. He is currently the
Manager of the Power Systems Group, Grenoble
Electrical Engineering Laboratory (G2ELab),
Grenoble National Institute of Technology, and is also a Professor with
Joseph Fourier University of Grenoble. His research interests include power
electronics system modeling and control, power quality, and renewable energy
integration.

411

Damien Picault was born in Vitry-sur-Seine, France,


in 1984. He received the M.S. degree in electrical
engineering from the Grenoble National Institute
of Technology, Grenoble, France, in 2007, where
he is currently working toward the Ph.D. degree
on maximizing energy production of grid-connected
photovoltaic systems.
His main research interests are photovoltaic plants
and power electronics modeling.

Bertrand Raison (M03) was born in Bthune,


France, in 1972. He received the M.S. and Ph.D.
degrees in electrical engineering from the Grenoble
National Institute of Technology, Grenoble, France,
in 1996 and 2000, respectively.
From 2001 to 2009, he was an Associate Professor
with the Grenoble National Institute of Technology.
Since 2009, he has been a Professor with Joseph
Fourier University of Grenoble. His general research
interests are fault detection and localization in electrical systems and distribution network planning with
respect to fault management.

You might also like