Sie sind auf Seite 1von 6

Adaptive Distributed MPPT Algorithm for

Photovoltaic Systems
F. Scarpetta, M. Lisere, and R. A. Mastomauro
Department of Electrical and Electonic Engineering
Politecnico di Bari, Italy
fscarpetta@hotail.com, liserre@poliba.it, mastomauro@deemail.poliba.it
Abstact - Fast change of the irradiance conditions entails
failure of the maximum power point tracking (MPT) algorithm
in a photovoltaic system (PVS). In this paper it is proposed the
combination of a modifed Perturb and Observe (P&O)
distributed MPT and an adaptive algorithm. The MPT has
been tested in case of hard mismatch conditions through
simulation showing high performance in terms of efciency.
I. INTRODUCTION
Partial shading, and consequently mismatch, among
photovoltaic (PV) modules represents one of the main causes
of PVS mlssmg power production [1-2], [4]. Since
mismatching leads to a MPPT effciency decrement, the
integration of the MPPT inside each PV panel is the fontier
of this technology. There are several MPPT methods, which
require different implementations depending on PVS
topology: single-stage or double-stage. I case of a double
stage conversion system the MPPT technique is used to
control the DC/DC converter (the parallel connection of the
MPPT can allow higher effciency [9]), while in case of
single-stage conversion system the MPPT is included in the
DC/ AC converter contol. A distibuted MPPT (DMPPT) can
be performed by means of a DC/DC converter inside each PV
panel [10-12] or on the DC/AC converter unit integrated in
each PV panel. I allows an improvement of the PVS power
effciency. Among the various MPPT techniques, the Perturb
and Observe algoritm is one of the best compromises
between implementation simplicity and reliability [8].
Otherwise, in case of rapid irradiance changing conditions,
the P&O algoritm may result unstable. Many solutions, such
as the Estimate-Perturb-Observe method [4], the dP-P&O
method [2] and the adaptive algorithm [ I], can be found in
literature tying to improve the tracking performance. The
application is on the PVS DC/AC converter. I this paper the
dP-P&O MPPT is applied to the contol of the DC/DC
converter in case of distributed MPPT. By combining this
MPPT with an adaptive algoritm, it is possible to obtain a
fast response and a high effciency also in case of rapid
change of irradiance conditions. The behavior of a PVS
where the panels are working under different level of shading
is simulated in this paper and a performance comparison
among taditional P&O, adaptive P&O and the adaptive dP-
P&O MPPT for fast changes of irradiance conditions is
operated. I particular in Section IT the effects of mismatch
on the PVS performance are analyzed; a brief intoduction of
DMPPT is reported in Section TIT; suggestions about the
possible modifcation of the traditional P&O are detailed in
Section IV; fnally, the proposed adaptive dP-P&O MPPT for
fast changes of irradiance conditions is described in Section
V. The PVS model and the comparison results are shown,
respectively, in Section VI and in Section VIT.
IT. THE MISMATCH EFFECT
PV panels are usually equipped with one or more bypass
diodes (Fig. I ). Although the bypass diodes allow higher
currents by adding a path in the circuit, they alter the power
curve shape in case of irradiance mismatch. Fig.2 shows how
the mismatch irradiation affects the P-V curve shape of the
PVS. Assuming the operating point located on the maximum
peak, the power extaction is not optimal because happens
that the shadowed panel needs a lower current to work in its
maximum point thus reducing the power fom the irradiated
panel (Fig. 3). The reduction of power is higher (up to 60%)
in case of hard mismatch.
lill
' v
Fig. I. PVS with panels connected in series and equipped with bypass diodes.
Voltage [Vj
Fig. 2. Mismatch efct on the power-voltage curve of a PV array
connected in a 2x I matrix.
978-1-4673-2421-2/12/$31.00 2012 IEEE b708
(
"
o
"
Voltage [VI Voltage [VI
Fig. 3. Power extracted from the irradiated (lef) and shadowed (right)
panel.
m. DISTRIBUTED MPPT
Tf the PVS is connected directly to an inverter with MPPT
fnction included (Fig. 1) or in case a centalized DC/DC
converter is used, the MPPT will not extract the maximum
power under mismatch conditions due to the reasons shown
before. The MPPT algoritm fails to tack the right maximum
power point due to the presence of relative peaks. T order to
improve power extraction a distibuted topology, with a boost
converter for each panel and with its own MPPT algoritm,
should be adopted [10] (Fig. 4). The main effect provided by
the addition of a boost converter for each panel is the exteral
P-V curve smoothing, hence there is just a sole maximum
power point. Furthemore, assuming to work in this point, the
power extacted fom each panel is the maximum possible
(Fig. 5).
10 20
Fig. 4. PVS with DMPPT.
30 40 50 60
Voltage [Vj
70 80
Fig. 5. P-V curve smoothing provided by the distributed DCIDC
converters.
TV. P&O MPPT IMPROVEMENTS
The P&O method is based on the following criterion: the
duty cycle of the boost DC/DC converter is perturbed and, as
long as the power drawn fom the system increases, the
direction of perturbation does not change. I results:
1(/?)=1(/!)u, J(/?)J(/!),1(/!)1(/)
(1)
1(/?) = 1(/ l)u, J(/?)> J(/ l),1(/ l)1(/)
where D is the duty cycle, a is the increment/decrement step
and P the measured power. Otherwise, if the power drawn
decreases, the sign of perturbation changes:
1(/?)=1(/!)u, J(/?)J(/!),1(/!)1(/)
(2)
1(/?)= 1(/ l)u, J(/ ?) J(/ l),1(/ l)> 1(/)
A. Adaptive MPPT algorithm
Since the P&O method needs a good compromise
between dynamic response and steady-state performance, the
incremental step of the duty cycle can be automatically
adjusted. T [1] the tuning of the step is contolled as follows:
u(/)=
u(/-l)
(3)
where ar is the step of duty cycle at time k. The incremental
step depends upon the historic value of step ar-i), the
measure of the power variation LP = P(-P(-i) and the gain
M By choosing the right value of M and limiting the
variation range of the step, it can be obtained the best
performance, with the highest value of the increment when
the working point is far fom the target, and the lowest one In
steady-state conditions.
B. Fast trackingfor irradiance changing conditions
It is kown that the P&O methods may fail to track the
optimal point by choosing the wrong direction during rapidly
changing irradian ce, especially on cloudy days. Tn [2] it is
proposed a solution, called "dP-P&O" method, which
determines the right direction by adding a measurement (Px)
taken in the middle of the MPPT sampling period (Fig. 6).
The application which can be found in literature is in case of
PVS DC / AC power conversion stage. The following
assumptions are considered:
1) The effect of perturbation on power curve is over when
Px is taken;
:
"
o
"
t
kT kH T/2
Time [sl
(k+1)T
Fig. 6. The dP-P&O method principle.
b709
2) The variation of environmental phenomena (irradiance,
temperature) is constant during a sample time.
The power variation dP is evaluated as the difference
between the change due to the perturbation of the duty cycle
plus the environmental variations (dP]) and the variation due
only to environmental conditions (dP2):
In the example shown in Fig. 6, the irradiance (and
consequently, the power) increment does not affect the value
of dP, which is negative. The advantage of this method is the
possibility, in case of rapidly irradiance changing conditions,
to switch rapidly between two control modes as illustated in
the next Section.
V. ADAPTIVE dP-P&O MPPT FOR FAST CHANGE
TRADIANCE CONDITIONS
A adaptive fast changing dP-P&O algoritm is
implemented in this paper in case of application to a DC/DC
distibuted MPPT: as long as dP2<dP+Thr (where Thr is a
fxed threshold) the taditional P&O MPPT method is used,
since the environmental changing conditions are negligible, in
this case the sign of perturbation is chosen equal to that of
dP2 T case of high irradiance variations, the MPP impedance
tends to lower values thus the duty cycle must increase
(positive slope) in order to follow the new target and vice
versa. T the example shown in Fig. 7, the tacking point
moves fom' l' to '2' along a low-irradiation power curve: in
case of sudden irradiance increasing, this point shifs fom '2'
to '3' on the right side of the new, high-irradiation power
curve, hence the correct tacking direction changes.
Combining the advantages of an adaptive algoritm
with the dP-P&O, it is possible to pursue the MPP tacking
despite environmental variations, for example during cloudy
days. Tn Fig. 8 the complete fowchart of the proposed
algoritm is illustated. Voltage and current of the panel are
sampled at TI2 seconds with respect to MPPT algoritm,
sampled at T seconds, in order to take the additional sample
(Px) during a period. The slope defnes the tacking direction:
if the environmental variations are predominant IdP21> IdPI
+ Thr, the slope sign of duty cycle is maintained.

:
;
0
0
225
200
P-V curve variation with
175
irradiance increasing
150
125
100
75
50
15 20 25 30 40
Voltage [Vj
Fig. 7. Efct of sudden irradiance increasing on the MPP tracking
direction.
The automatic tuning of the increment/decrement step a
takes place independently.
1. PVS DC POWER STAGE MODEL
The operation of the PVS shown in Fig.9 has been
emulated testing both the taditional P&O and the proposed
modifed adaptive dP-P&O algorithm in order to compare the
effciency performances. The PVS is directly connected to a
DC load of (12 0), hence the DC/AC power stage IS
avoided in the model.
The boost DC/DC converter used for testing is a
Synchronous boost with two power mosfets Ixys n-chanel,
Coilcraf inductor of IOOIH (DC Resistance 32mO,
Panasonic electrolytic input capacitor of 120llF (interal
resistance of 45mO), output capacitor of 220llF (resistance
30mO).
Fig. 8. Flowchart of the adaptive dP-P&O MPPT fr fst changes of
irradiance conditions.
Fig. 9. Operation of the PVS connected to a DC load.
b710
The commutation deadband time between the mosfets has
been set at (123 ns), taking into account the delay and
rise/fall time of the tansistors. The switching fequency is 40
kHz. A "average" model of the DC/DC converter has been
used for simulation (Fig.lO) in order to reduce the high
computational burden due to the high switching fequency,
hence, for this purpose the power mosfets have been replaced
by a current and a voltage contolled sources.
inductor Y-,Y.-D(1-D)
p
Fig. 10. Average model of the Synchronous Boost DCIDC converter.
The node a (active), p (passive) and c (common) of the
circuit of Fig.1 0 are related to each other by the contolled
sources; D indicates the duty cycle (amount of power
tansmitted to the output port); Req takes into account all the
dissipative effects of capacitor, inductor and switches and it
depends on duty cycle value.
The solar panel considered for simulations is one 220W
Solsonica610 with 60 cells in series and 3 bypass diodes. The
output current T of the panel, as kown, is ("one diode
model"):
[ q(V+T,)
) V + TR
T = T - T e
nKT _\ __ S
ph 0
R
p
(5)
Tph is the photocurrent (which depends on irradiation and
temperature [5-6]); To is the reverse saturation current and it
depends on temperature and n (idealit factor) [7]. The
parameters characterization of R" Rp and n, proposed in [3], is
based on a iterative method which gives the following results:
R, = (0.390), Rp = (25330), n = (1.18), To stc (reverse
saturation current in standard condition) = (9.11 e-9 A).
IT. STMULATTONRESULTS
The dynamic response of the PVS shown in Fig.9 is frstly
compared to a taditional, centalized MPPT algoritm
performed into the inverter as shown in Fig.l. The
assumption is that both the panels are flly irradiated (1000
W/m
2
) ad they are at the same temperature (25C). The
impedance (at the inverter input) is set at 2 0.
When a mismatch condition, as shown in Fig.2 happens,
the inverter fails the MPP tacking, matching the frst relative
maximum (about 200 W) without reach the absolute
maximum (Fig. I I ).
zze

Power
O
;
0
a
a
L
x
S
<
u
c
r
"
<
Q

z ,
Voltage [V]
Fig. 11. Power / Impedance versus Voltage of a Centralized MPPT
algorithm.

ee

oe
_ ze
e
u
se
u
t
^
e
w
ae
ze

e
e

z

^

e
,
Duty Cycle [%]
Fig. 12. Etciency versus Duty cycle of the synchronous Boost
DCIDC converter.
With a distibuted topology (Fig.9) it is possible to match
the absolute maximum starting fom the same conditions.
Simulation results highlight that it is possible to extract all the
available power (220 W fom each panel, as reported in the
panel datasheet) with duty cycle equal to just 19%. I Fig.12
the effciency of the synchronous DC/DC converter is shown
for different values of its output impedance: when it operates
with duty cycle equal to 19% the losses are about 3%.
Fig. 13 shows the extacted power and the duty cycle in
case of DMPPT P&O algoritm: at 1.3 seconds, one of the
two panel results shaded at 800 W/m
2
, then the P&O
algoritm continues to work independently following the new
MPP 400W which is double in comparison with the
taditional centalized MPPT. What it happens can be
described by means of the power-voltage curve (Fig. 14).
When both panels are flly irradiated, the load parabola
intersects the power curve in the linear part (' I ' ). If one panel
is shaded, the new working point jumps on a temporary
dashed curve ('2 ' ) at 19% of duty cycle and fnally reaches
the new steady state ('3 ' ): in both conditions the power
curves present only one peak.
b711

O
;
0
a

C
U
>
u
.
:
0
250
200 Panels fully irradiated
at lOOOW/m
'
150 Power recovery when
one panel is shaded
(800W/m
'
)
100
04 0.6 0.8 1.2 14 1.6 1.8 2.2
Time [sl
a)
30
04 06 08 12 14 16 18 22
Time [sl
b)
Figs. 13. DMPPT P&O algorithm performance in case of 20%
mismatch: a) power; b) DCIDC converter duty cycle.
O
;
o
a
500
Voltage [VI
Fig. 14. Power-Voltage curve in case of 20% mismatch.
Respect to the taditional P&O MPPT, the adaptive P&O
algoritm has the advantage to speed up the MPP tacking
when the working point is far fom the target. However, in
case of fast irradiance variations, a larger step of the duty
cycle may cause tracking failure. Considering a hard
mismatch case by choosing very different irradiance signals
for the two panels, as shown in Fig. 15, it can be seen that the
adaptive algorithm can choose the wrong direction leading to
power losses (Fig. 16).
The proposed adaptive dP-P&O MPPT algoritm for fast
changes of the irradiance conditions, shown in Fig. 8, is
tested under the following assumptions: the step 'a' is limited
in the range 0.01 - 0.05; the constant parameter M is chosen
equal to 5-5 in order to compensate the power variation. It
results that the effciency can reach 99.7%, while in case of
the adaptive P&O the effciency is 98.2%, (Fig. 17)
allowing an optimal and fast MPP tacking in comparison
with the previous cases.
O
;
o
a

O
;
0
a

: : : .

,...., ....;: .
|
Time [sl
Fig. 15. Trradiance signal profle chosen for the two panels.
400
350
300
Extracted
power
250
0 3 4 5
Time [sl
Fig. 16. PVS maximum power versus extracted power in case of
adaptive P&O (sample time 30ms).
400
6
O
350
;
o
a
300
power
250
o 4 6
Time [sl
Fig. 17. Power extracted with the adaptive dP-P&O MPPT algorithm
for fst changes of the irradiance variations (sample time 30ms).
Finally, it is considered a taditional P&O DMPPT
working with a half sample time: 15 ms in despite of 30 ms.
The aim is to demonstate how the choice of a higher sample
fequency is irrelevant and the system keeps to follow wrong
tacking direction during rapid variations of irradiance. Fig.
18 shows the irradiance profle chosen for this test: one of the
two panels is constantly irradiated until 3s, when shadows
cover the panel and causes an abrupt decrement.
b712
O
;
o
a
- -

..
...
U

"'
us
...
...
o_

PAAEl2(SHADOWJ

::/
/ / / /
/
>
/
.. / /
"' V ; /
,
'. , , ,
Time [s]
Fig. 18. Irradiance signal profle chosen for the two panels.
50

0
L----
2
---
3
---
4
--
5
--
6
Fig. 19. PVS maximum power versus extracted power in case of
adaptive, high fequency P&O (sample time 15ms).

T
Fig. 20. Improved tracking of power with the adaptive dP-P&O
algorithm (sample time 30ms).
The MPP tacking of the adaptive, high fequency
algoritm without dP-P&O is shown in Fig.19: as it is
expected, the sudden irradiance variations shif the
operating point back and forth around the MPP (as seen
in Fig. 7).
Tn the same mismatch conditions but with an higher
sample time (30ms), the dP-P&O algoritm has been
enabled in the boost dc/dc converters: it has been
demonstated that the tacking performances are better
than in the previous case and the effciency is around
99% (Fig. 20). This result is possible since, even if the
duty cycle step width is high, the algoritm allows the
tacking of the right direction ensuring a fast and
accurate power recover.
.
Ill. CONCLUSIONS
In mismatch condition the use of a DMPPT algorithm
allows better performance in comparison to a taditional
MPPT applied to a centalized DC/DC converter topology. In
this scenario the choice of an adaptive P&O MPPT can
represent, respect to the well-known P&O, a good
compromise between speed accuracy and power losses during
the steady-state operation. However, it is verifed that, the use
of the adaptive P&O can fail the MPP tracking in case of hard
mismatch with irradiance sudden variations.
Differently, an adaptive dP-P&O algorithm applied to
the DC/DC converters in case of DMMT is proposed in this
paper; it allows to switch between two contol modes: a
simple P&O MPPT in case of very limited environmental
conditions variations, or redefning the tacking direction in
case of fast and high variations. I is demonstated that the
adaptive dP-P&O improves effciency in case of hard
mismatch, which is around 99%, despite the MPPT sample
time.
REFERENCES
[I] w. Xiao, W.G. Dunfrd, "A modifed adaptive Hill Climbing MPPT
method fr photovoltaic power systems," 35
1h
annual IEEE power electronics
specialist conference, pp. 1957-1958,2004.
[2] D. Sera, R. Teodorescu, "Optimized maximum power point tracker for
fst-changing environmental conditions," IEEE Trans. Ind Electronics, vol.
55, Issue 7, 2008, pp. 2629-2637.
[ 3] M.G. Villalva, 1.R. Gazoli, E.R. Filho, "Comprehensive approach to
modeling and simulation of photovoltaic arrays," IEEE Trans. Power
Electronics, vol. 24, no. 5,2009.
[4] C. Liu, B. Wu, R. Cheung, "Advanced algorithm fr MPPT control of
photovoltaic systems," Canadian Solar Buildings Conference, Montreal,
2004.
[5] D. Sera, R. Teodorescu, P. Rodriguez, "PV panel model based on
datasheet values," Proc. IEEE Int. Symp. Ind Electron. (ISIE 2007), pp.
2392-2396.
[6] A Driesse, S. Harrison, P. Jain, "Evalutating the efctiveness of
maximum power point tracking methods in photovoltaic power systems
using array perfrmance models," Proc. IEEE Power Electron. Spec. Con
(PESC 2007), pp. 145-151.
[7] W. De Soto, S.A Klein, W.A Beckman, "Improvement and validation of
a model fr photovoltaic array perfrmance," Solar Energ, vol. 80, no. I, ,
2006, pp. 78-88.
[8] R. A Mastromauro, M. Liserre, A Dell' Aquila, "Control Issues in
Single-Stage photovoltaic Systems: MPPT Current and Voltage Controf',
Trans. Ind Inormatics, frthcoming issue.
[9] R. Gules, 1. De Pellegrin Pacheco, RL. Hey, 1. Imhot "A Maximum
Power Point Tracking System With Parallel Connection for PV Stand-Alone
Applications," IEEE Trans. Ind Electron., vo1.55, no.7, pp.2674-2683, July
2008.
[10] N. Femia, G. Lisi, G. Petrone, G. Spagnuolo, M. Vitelli, "Distributed
Maximum Power Point Tracking of Photovoltaic Arrays: Novel Approach
and System Analysis," IEEE Trans. Ind Electron., vo1.55, no.7, pp.2610-
2621, July 2008.
[II] G. Petrone, G. Spagnuolo, M. Vitelli, "A Multivariable Perturb-and
Observe Maximum Power Point Tracking Technique Applied to a Single
Stage Photovoltaic Inverter ," IEEE Trans. Ind Electron., vol. 58, no. I, pp.
76 - 84, Jan 2011.
[12] Wu Wenkai, N. Pongratananukul, Qiu Weihong, K. Rustom, T.
Kasparis, 1. Batarseh, "DSP-based multiple peak power tracking for
expandable power system," Eighteenth Annual IEEE Applied Power
Electronics Conerence and Exposition, APEC '03, Miami Beach, Florida,
USA, 9-13 Feb. 2003, vol. I , pp. 525- 530.
b713

Das könnte Ihnen auch gefallen