Upload
nguyen-thanh-tuan
View
198
Download
0
Embed Size (px)
Citation preview
A fast current-based MPPT technique based on sliding mode control
Enrico Bianconi(l), Javier Calvente(2), Roberto Giral(2), Giovanni Petrone(3) Carlos Andres Ramos-Paja(4), Giovanni Spagnuolo(3) and Massimo Vitelli(5)
(1) Bitron Industrie S.p.A., Grugliasco (TO) - Italy
(2) Departament d'Enginyeria Electronica, Electrica i Automatica - Universitat Rovira i Virgili - Spain
(3) Dip. Ingegneria dell'Informazione e Ingegneria Elettrica (D.I.I.I.E.), University of Salerno - Italy
(4) Facultad de Minas - Universidad Nacional de Colombia - Sede Medellin - Colombia
(5) Dip. Ingegneria dell'Informazione (D.!.I.), Second University of Naples - Italy
E-mail: [email protected]. [email protected], [email protected]
[email protected], [email protected], [email protected], [email protected]
Abstract-This paper introduces a novel maximum power point tracking technique aimed at maximizing the power produced by photovoltaic systems. The largest part of the approaches presented in literature are based on the sensing of the photovoltaic generator voltage. On the contrary, in this paper a current-based technique is proposed: the sensing of the current in the capacitor placed in parallel with the photovoltaic source is one of the innovative aspects of the proposal. A dual control loop based on the sliding mode control ensures a very fast tracking of irradiation variations. Features of the proposed algorithm are supported by a theoretical analysis and simulations. The technique described in this paper is patent pending.
I. INTRODUCTION
Maximum Power Point Tracking (MPPT) function is one of
the key factors in any PhotoVoltaic (PV) system. Perturb and
Observe (P&O) MPPT technique is often preferred in com
mercial products because it makes the implementation in low
cost digital controllers quite simple. Significant efforts have
been producing in order to improve the electrical efficiency
of the PV power processing system, often forgetting that the
total efficiency is the product of the electrical efficiency by
the MPPT efficiency. The latter is affected by the steady
state and transient performances of the algorithm, thus in
presence of an almost constant and of an abrupt change in the
irradiation level. As demonstrated in [1], the P&O performance
optimization can be achieved by means of a suitable choice
of the values of the two parameters affecting the tracking
performances, that are the amplitude and frequency of the
perturbations applied to the PV voltage by means of a switch
ing converter. In [1] it was demonstrated that in steady state
conditions the best performances are ensured by choosing the
smallest possible value of the perturbations amplitUde. This
choice minimizes the displacement from the maximum power
point when irradiation level is almost constant, but penalizes
the tracker promptness in presence of abrupt changes in the
irradiation level. In fact, the analytical result shown in [1]
clearly puts into evidence that the greater irradiation variation
to be tracked, the larger the perturbation amplitude must be
978-1-4244-9312-8/11/$26.00 ©2011 IEEE
chosen. In [1] it has also been shown that a further increase
in the value of the perturbation amplitude becomes mandatory
if the PV voltage is affected by some disturbance. This is
the case, for instance, of the low frequency oscillation, at a
frequency that is twice that one of the grid, backpropagating
to the PV voltage in systems connected to the AC mains.
In this paper a current-based P&O MPPT strategy is pro
posed. It allows to have a very prompt tracking of the varia
tions in the irradiance value and, additionally, it guarantees the
rejection the low frequency voltage oscillations affecting the
PV array and due to the inverter operation in grid-connected
systems.
The main peculiar aspect of the proposed approach is in the
fact that it is a current-based technique. The largest part of
the MPPT algorithms presented in literature is voltage-based,
because of the logarithmic dependency between the irradiation
level and the PV voltage. In fact, the linearity between the
irradiance level and the PV current would be very useful for
a fast MPPT, but irradiance drops would require a suitable
control.
In literature, few examples of current-based MPPT ap
proaches can be found. For instance, in [2] a model-based
approach, depending on the assigned PV module characteris
tic, is presented. In [3] the PV array current is not sensed, but it
is reconstructed through a sliding-mode observer of the dc/dc
converter's input inductor current and fed into the controller
to generate the maximum power point reference voltage.
II. BASICS OF THE PROPOSED CONTROL TECHNIQUE
Fig. 1 shows a possible control strategy operating on the
inductor current of the DCIDC converter in order to regulate
the PV current value. The role of the input capacitor Gin is in
absorbing the switching ripple affecting the inductance current,
so that the PV current is controlled through the average
inductor current value.
In the circuit node connecting the PV array and the boost
converter input inductance L and capacitance Gin (see Fig. 1),
59
PVarray
��-----�I"'I i J.:' 41 ' ,. Vpv: 1 -'-:.)l.::.J
: : l... ............ :-''''' {�y,�-EJ
, , , , , , L�!!.
Figure I. System scheme.
the Kirchhoff current law holds:
ipv = iCin + iL
The current controller gives the reference current irer
so that the steady-state is ensured when ivr = O.
(1)
(2)
According to the classical DC/DC converters current control
theory, the signal u(t) driving the MOSFET is a function of
the error signal ei(t) = ire!(t) - idt), so that the inductor
current iL is regulated according to the current reference signal
ire!' By using (1) and (2), the following simplification is thus
possible:
ire! = iCin + iL + ivr
ei = iCin + iL + ivr - iL
ei = iCin + ivr
(3)
(4)
(5)
where the control objective ire! = iL (see Fig. 1) leads to
ei = 0, so that it is equivalent to:
and the steady-state condition results in iCin = O.
PVarTay Inver1er
L-__ Jh:-r:�--�----�---Lt:-L�� : I \ :
Figure 2.
'N ! ,�!" v ·
8 j�� ' .. :-6��-EJ
System scheme based on input capacitor current control.
(6)
The simplified control objective (6) reveals that the control
structure of Fig. 1 can be simplified as in Fig. 2, with the
inner control loop which is now aimed at regulating the input
capacitance current icin. This simplification reveals important
because the practical implementation of the scheme shown in
Fig. 1 would require two high-bandwidth current sensors for
ipv and iL, both affecting the MOSFET control signal u(t) value. Instead, in the scheme of Fig. 2, u(t) depends on the
Isc 1 G
I.v
DCIOC converter
Figure 3. Small-signal model of the system.
iCin instantaneous value only. Consequently, only one high
bandwidth current sensor is required because the other one
is a low-bandwidth one dedicated to ipv, whose variations
are dictated by the slow MPPT controller dynamics. Another
advantage of the system shown in Fig. 2 is in its easier analysis
with respect to that one depicted in Fig. 1.
III. SLIDING-MODE-BASED MPPT
The sliding surface SURF is given in (7):
SURF = -icin - ivr = 0 (7)
It is defined in order to fulfill the control objective (6). The
current loop reference ivr in the scheme of Fig. 2 is given
by Gv (s) . The definition of the sliding surface SURF as
in (7) suggests that, in sliding-mode operation, the capacitor
current iCin changes in order to reject the perturbations on the
bulk capacitor voltage Vb and to track the perturbations in the
irradiance level. Thus, the fact that SURF does not depend on
those variables ensures the MPPT operation and the rejection
of the low frequency disturbances, at a frequency that is the
double of the grid frequency, in single phase AC applications.
Two conditions must be fulfilled in order to ensure the
sliding-mode operation [4]:
SURF = 0 dSURF
= 0 dt
(8)
(9)
From the first condition (8), and by accounting for the
characteristic equation of the input capacitance:
. dvpv ZCin =Cin· �
the following condition is obtained:
dvpv dt
(10)
(11)
Equation (11) is valid in sliding-mode condition and gives a design criterion for the external voltage controller.
From the second sliding-mode condition (9), and by con
sidering that iCin = ipv - iL, it results that:
dSURF =
diL _ dipv _ divr = 0
dt dt dt dt (12)
By neglecting the series and parallel resistances of the PV
array, namely Rs = 0 and Rsh = 00, the PV current can be
approximated by the expression:
60
mode control of the converter is preserved:
(13) disc vpv- L- > 0 (20)
where Vpv is the PV voltage, IR and a are parameters depend
ing on the PV modules used, isc is the short-circuit current
for a given irradiance level, which has also an approximated
proportional relation with the irradiance isc = ks . S [5]. In
fact, it is:
isc = isc,STC· -S
S . (1
+ G.[ . (Tpv - TpV,STc)) (14)
STC
The analysis of the equation (12) can be done by using the
small-signal model shown in Fig. 3, wherein the switching
converter is represented by a current source and the PV gen
erator is given as a Norton model including the photoinduced
current source and the differential conductance G which is
calculated in the generator's operating point. By looking at
Fig. 3 and (13), it results:
(15)
As a result, the sliding-mode condition given in (12) can be
thus rewritten as: d�; + G. d:v _ d;;c _ d;�T = 0 (16)
In addition, the classical boost converter model gives the
following relationship:
diL VPV
dt L
Vb· (1 - U) L
(17)
where the value u = 1 is used in the MOSFET ON state and
the value u = 0 in the OFF state.
In order to obtain the constraints that must be fulfilled
in order to ensure the sliding-mode operation, the equivalent
control technique [6] can be used. The constraints on the inputs
and states are defined by ensuring that the average control
signal ueq fulfills the inequalities 0 < ueq < 1. From equations
(11 )-( 17) the following equivalent control equation is obtained:
VPV _ Vb· (1 - Ueq ) + G. dvpv _ disc _ divT = 0 (18)
L L dt dt dt
From (7) and (12) it is deduced that the sliding-mode
equilibrium point is defined by {ipv = iL, iVT = O}. In (18),
at the equilibrium point it is dVd�v = 0, while the constraints
to be fulfilled in terms of maximum slope values of iVT and
irradiance, 4ftt and �� = ks· 4!tt, ensuring the sliding-mode
operation are calculated by using the superposition principle.
Thus, by putting d�tr = 0 in (18), the effect of the isc variations ensuring that 0 < ueq < 1 leads to the inequality:
vpv- L� 0< dt <1
Vb (19)
so that the two following inequalities ensure that the sliding
dt disc
vpv - L-- < Vb dt
(21)
Finally, the constraint to be fulfilled on the isc slope in order
to guarantee the proper sliding mode operation is obtained:
Vpv - Vb disc Vpv
L < dt < L (22)
It is worth noting that the lower bound for the isc slope
in (22) is negative because of the adoption of a boost con
verter. Especially if the step-up dc/dc converter is designed
for guaranteeing a high boosting factor, e.g. in PV module
dedicated applications, the quantity vPvL-
vb is deeply negative.
This means that the larger the converter's voltage boosting
factor, the faster negative short circuit current variation can be
tracked without loosing the sliding mode behavior. Due to the
proportionality between isc and the irradiance S, condition
(22) can be translated into a constraint for the maximum
irradiance variation the MPPT technique is able to track:
Vpv - Vb dS Vpv --=-=:--:--..:.. < - < --
L· ks dt L· ks (23)
Inequality (23) reveals that the maximum irradiance vari
ation that can be tracked without losing the sliding-mode
control is bounded by the inductor current derivatives in the
OFF and ON MOSFET states. This means that the inductance
value L can be properly designed in order to follow the
expected irradiance profile and variations, according to the
specific applications. For instance, stationary PV power plants
will be subjected to slow irradiance variations, while in PV
applications dedicated to sustainable mobility fast irradiance
variations would be tracked.
By fixing 4!tt = 0 in (18), the effect of variations on iVT
can be accounted for, so that the following constraint on 4ftt is obtained:
(24)
which again shows that the maximum iVT slope value that
can be tracked without missing the sliding-mode control
depends on the inductor current derivatives in the OFF and
ON MOSFET states. In this case, the voltage controller can
be designed to fulfill this dynamic constraint.
It is worth noting that, due to the symmetry of the expression
(18), constraints (23) and (24) show the same boundaries.
When both (22) and (24) conditions are fulfilled, the con
verter is in sliding-mode control and therefore the dynamic
of the system is given by (11). The transfer function Gv/i(s) between the input capacitor voltage, that is the PV voltage,
vCin and the current reference iVT provided by the voltage
controller is:
(25)
61
A. PI controller design The Gv design is performed by means of a traditional PI
compensator. By taking into account the PI transfer function GPI(s) given in (26) and the voltage error Ev(s) definition
(27) that compensates the negative sign appearing in (25),
the closed loop transfer function T( s) of the system can be
expressed as in (28):
ki PI(s) = kp +
s Ev(s) = -(Vref(s) - Vcin(s))
T(s) = kps + ki
GinS2 + kps + ki
(26)
(27)
(28)
The transfer function T( s) in (28) is designed by accounting
for a classical relation between the rising time t R of the closed
loop voltage and the minimum switching period Tsw.
The T( s) structure gives the following relations:
kp = 2GinPWn
ki = GinWn2
(29)
(30)
so that, by using the equivalent time constant definition T = _1_
PWn In sliding-mode, the duty cycle D(t) and switching fre-
quency fsw(t) will oscillate around their nominal values Do
and fswo, respectively, in order to cancel out the bulk ca
pacitor voltage oscillations !:,.Vb(t) according to the following
formulas:
D(t) = 1-VPV
VbO + !:,.Vb(t)
f () - VPV . D(t) sw t -
H. L
VPV Do=l-
VbO
f - Vpv· Do swO - H. L
B. MPPT refinements: input and output signals filtering
(31)
(32)
The MPPT controller adopted in this example is a traditional
Perturb and Observe (P&O) one. This means that the P&O
output generates a step change in the vref signal of the voltage
controller, thus violating the constraint on �. In order to
avoid this drawback, the P&O output is filtered thus generating
a vref dynamic behavior that is comparable with that one of
the closed loop system. In this way, the first order filter G fv (s) given in (33) is used,
G () Vref(s) 1
fv s = P&O(s) = Tfs + 1
where P&O(s) is the P&O control signal.
(33)
There is a need to filter signals VPV and ipv at the MPPT
input too. In fact, the possibility of reducing the amplitude of
the perturbations given by the P&O algorithm determines the
corresponding reduction of the vpv and ipv perturbations,
with a significant detrimental effect of the switching converter
operation. As for the MPPT output, the PV current and voltage
measurements used to calculate the PV power needed for the
P&O controller can be filtered by using the same G fv (s) to re
move the switching frequency components without degrading
the dynamic response of the system, thus avoiding the P&O
controller to be confused by the switching ripple. The !:,.vref perturbation generated by the P&O controller can be adopted
equal to the voltage ripple, which is mitigated due to the PV
voltage filtering. The PV voltage ripple !:,.vPv is calculated as
[7]:
!:,. _ H . Tsw,max
vpv -8Gin
IV. SIMULATION RESULTS
(34)
Some simulation results have been obtained in PSIM en
vironment. In Fig. 4 the complete PSIM schematic has been
shown.
Figure 4. Complete PSIM schematic.
The input capacitance value has been assumed equal to
Gin = 50 p,F, the input inductance L = 410 p,H, with
a desired inductor current ripple H = 4 A. The minimum
PV voltage value has been fixed at VPV,min = 100 V, and a bulk capacitor Gb = 22 p,F with an average voltage
VbO = 450 V has been considered. According to (32), the
minimum switching frequency is equal to 40 kHz. With such
values, from (34) it results that !:,.vref = !:,.vPv = 0 .2 V. In order to put into evidence one of the main features of
the proposed approach, i.e. the ability of rejecting the low
frequency voltage variations backpropagating from the bulk
voltage towards the PV voltage, !:,.Vb(t) has been assumed to
oscillate in the range [-160, 160] V, so that !:,. Vb = 160 V. This can be obtained by considering a small bulk capacitance
Gb = 22 p,F. Considerations done above about the PI controller design
lead to a maximum switching period Tsw,max = 25 p,s so that, by choosing a traditional tR/Tsw,max = 8 value,
the closed loop rising-time is tR = 200 p,s, which is the
first design consideration for the PI controller. The second
design consideration is related to the damping of the system,
which can be adjusted to the p = 0 .7 traditional value.
By approximating tR as 4T, where T is the equivalent time
constant of the closed loop system, it results that the PI
controller that ensures the defined t Rand p is the following
62
one:
( ) s + 2040 8 .16 PI s = 2· --.:.--
s (35)
which also provides an infinite gain margin and a phase margin
equal to 65 .2°.
The PI design from the t R specification also allows to define
the MPPT controller period Ta equal to the Vpv stabilization
time, that is 1 .5· tR approximatively. This Ta value ensures
that the PV power has reached its steady state when the MPPT
controller measures it, thus avoiding the MPPT deception [1].
In this example Ta = 300 f.Ls. As for the signal filtering, the time constant in (33)is set to
Tf = T = 50 f.Ls. Fig. 5 shows the results of the system simulation (Fig.
2) by using the sliding-mode and Gv controllers. First of
all, simulation results put into evidence that the large low
frequency oscillations affecting the bulk voltage have been
rejected at the PV terminals. Moreover, the sudden irradiance
variations, having an instantaneous effect on the PV short
circuit current, have been correctly tracked, with the track
ing system permanently tracking the maximum power point
even at the very high rate of irradiance variation the control
technique has been subjected to. The two features listed here
above make the control strategy shown in Fig. 2 suitable for
all the PV applications in which the adoption of electrolytic
capacitors at the dc bus must be avoided andlor wherever an
excellent MPPT performance is required. Current literature
(e.g. in ( [8])) puts into evidence that electrolytic capacitors
are the bottleneck of any PV power processing system because
they affect its lifetime significantly. The extraordinary fast
MPPT capability opens to the PV generators controlled by
means of the proposed technique the doors of a large range of
applications for which the sudden irradiation changes are the
rule, e.g. in sustainable mobility and for the PV integration on
cars, trucks, buses, ships and so on.
Fig. 6 shows the magnification in two different time in
tervals of the waveforms shown in Fig. 5. The first time
window is [23.5, 27.2] ms: plots show the system behavior
while approaching the steady state operation: Fig. 6(a) puts
into evidence the proper design of the P&O parameters leading
to a three points behavior of the PV voltage. The same figure
shows the waveform of the closed loop voltage reference vref and the accurate tracking performed by both the Gv(s) and the
sliding-mode controller. Fig. 6(b) shows that the PV current
waveform is free of the 100Hz oscillations due to the inverter
operation, thus demonstrating that the control technique has
been able to stop the back propagation of the large oscillations
of amplitude LlVb affecting the converter's output voltage. Fig.
6(c) shows the input capacitor current and the current reference
control signal waveforms. Finally, Fig. 6(d) shows that the
MPPT technique has been able to drive the system toward the
maximum PV power corresponding the short circuit current
of lO A settled for this simulation example.
Similarly, Figs. 6(e)-(h) show a magnification of the system
behavior in the time interval [333.6, 337.9] ms, when a high
irradiance transient generating a fast change in the short circuit
current of 80 % has been imposed. At the beginning, with
isc = 10 A, the converters works in continuous conduction
mode and the proper operation of the control algorithm, that is
the vref and the PV maximum power point tracking, is evident.
Afterwards, at t = 335 ms, the PV short circuit current has been
forced to drop suddenly at isc = 2 A, thus simulating a steep
reduction in the irradiance level. As a consequence of this, the
converter enters in discontinuous conduction mode at t > 335
ms (Fig. 6(f)): in this case the switching frequency of the
system is reduced, but the voltage controller Gv (s) still drives
the PV voltage to follow the MPPT controller reference. Fig.
6(e) shows the satisfactory PV voltage behavior in continuous
and discontinuous conduction modes, and Fig. 6(h) puts into
evidence the fast MPPT response under a constant and time
varying irradiance levels.
As a final consideration, the upper and lower limits of
the slopes � and � as appear in (22) and (24) have
been calculated for the numerical example considered in this
section. It results that:
disc divr I -805 Alms < - + -
d < 293 A ms
dt t (36)
This inequality shows that the most restrictive condition
appears when the irradiance level, and thus the short circuit
current, is subjected to a positive variation. By keeping into
account the transfer function (35) of the PI voltage controller
adopted in this numerical example and by considering the
sliding-mode eqUilibrium point (Vein = vref, V�t = 0), the derivative of ivr is given as follows:
(37)
The perturbation amplitude imposed by the adopted P&O
MPPT controller is Llvref = 0 .2 V, which results, by account
ing for the filter dynamics (33), in a maximum slope of 2.1
V Ims. This can be expressed in terms of the boundaries of
� as follows:
min (Tt) = -4 .2 Alms ,Llvref > 0 } (38) max (Tt-) = 4 .2 Alms ,Llvref < 0
which is two orders of magnitude lower than the limits (36)
that ensure that the sliding-mode control is still operating
correctly. This means that this system is able to track PV short
circuit current perturbations with slopes between [-800, 290]
Alms, thus corresponding to very fast irradiance perturbations
that are approximately in the range [-80, 29] W l(m2f.Ls). Such
remark confirms the inherent bent of the proposed technique
for applications characterized by uncommon irradiance slopes.
V. CONCLUSIONS
In this paper a novel, patent pending technique for the
maximum power point tracking of photovoltaic systems has
been introduced. The approach is based on the sliding mode
control technique and is based on the sensing of the current
drained by the capacitor which is usually put in parallel
with the photovoltaic generator. The technique implementation
63
i IL-:�:�:�:;;;;;;:;;;;;;:;;;;;:"-------:�EiJ�:� 0.1 0.2 0.3 0.
4 0.5 0.6 0.7 0.8 0.9
Time(s) (e)
r�t ' '" : ' : · o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
(d) 600 �1II!I!!IUI!IIl!!Il!l!ll!IlIlU! ...... � ......... � ......... � ...... JI!!1!11!I1I1I1!I1!I!!lIII!I!�
:D !�� �imiiiiliiliimilimiliil\lir"'''�'''''''''�'''''''''�'''''''''�miiiili�liimm�mliiml 3 00'-
--0
,.... '--..
,.
0."-
2 --
0...
,.3--
--'
0.4----,O.
L.5
--.,..
0.6,.--...,.O
L
.7
--0
,....
8--..
,
.0
.
�
g-
---'
Time(s)
Figure 5. Simulation of the system of Fig. 2 using the sliding-mode and Gv controllers.
('1
L:�I�.� II Ii er! 23.5 24 24.5 25 25.5 26 26.5 27
Time[ms] (dl
(0)
:�8f:S 334 334.5 335 335.5 336 336.5 337 Time (ms)
(II
1':1 � 334 334.5 335 335.5 336 336.5 337
Time [msj
334 334.5
(gl
f�1 \ c=g: 334 334.5 335 335.5 336 336.5 337
Time [ms] Time [msJ
Figure 6. Zoom of of Fig. 5.
requires few components and allows to track uncommonly fast
irradiance variations and is able to reject the low frequency
disturbances affecting the bulk voltage in grid connected
applications and backpropagating towards the photovoltaic
generator. Simulation results confirm the attractiveness of the
proposed method. The authors are confident that in the final
paper also some experimental results will be available.
REFERENCES
[1] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, "Optimization of perturb and observe maximum power point tracking method," Power Electronics, IEEE Transactions on, vol. 20, no. 4, pp. 963 - 973, 2005.
[2] H. T. Duro, "A maximum power tracking algorithm based on impp =
f(pmax) function for matching passive and active loads to a photovoltaic generator," Solar Energy, vol. 80, no. 7, pp. 812 - 822,2006.
[3] E. limenez-Brea, A. Salazar-L1inas, E. Ortiz-Rivera, and 1. GonzalezLlorente, "A maximum power point tracker implementation for photovoltaic cells using dynamic optimal voltage tracking," feb. 2010, pp. 2161 -2165.
[4] S.-c. Tan, Y. Lai, and C. Tse, "General design issues of sliding-mode controllers in dc dc converters," Industrial Electronics, IEEE Transactions on, vol. 55, no. 3, pp. 1160 -1174, 2008.
[5] U.Eicker, Solar Technologies/or Buildings. Wiley, 2003. [6] H. Sira-Ramirez, "Sliding motions in bilinear switched networks," Cir
cuits and Systems, IEEE Transactions on, vol. 34, no. 8, pp. 919 - 933, Aug. 1987.
[7] R. W. Erickson and D. Maksimovic, Fundamentals 0/ Power Electronics, 2nd ed. Springer, 2001.
[8] G. Petrone, G. Spagnuolo, R. Teodorescu, M. Veerachary, and M. Vitelli, "Reliability issues in photovoltaic power processing systems," Industrial Electronics, IEEE Transactions on, vol. 55, no. 7, pp. 2569 -2580, july 2008.
64