Input-controlled Buck Converter for

Photovoltaic Applications: Modeling and Design

M. G. Villalva, E. Ruppert F.
University of Campinas (UNICAMP), Brazil
[email protected], [email protected]
Keywords: converter, buck, control, photovoltaic, PV.
Abstract
DC-DC converters used in some photovoltaic (PV) systems
require that the input voltage be controlled while the output
voltage is constant. This paper shows different ways of
modeling and controlling a buck converter with variable input
voltage fed by a photovoltaic array.
only in the current-source region. When Vpv > VMP this model
does not represent the array correctly. Fig. 2b shows another
linear circuit that we could think as a possible array model.
This circuit originates the line that contains the segment ~ of
Fig. Ib and describes the array in the voltage-source region of
operation. This model is rare in the literature and our
application with the input-controlled buck converter it is
especially undesirable.
(a) (b)
Rs Rs
I\,r-O
~ ) ( '
i
r-----....--......--J\/\/\J-o+.!!:::.- Ix <D
Fig. 2: PV array equivalent circuits.
(a) Current source. (b) Voltage source.
3 Modeling and controlling the converter
Fig. 3 shows a DC-DC buck converter attached to a PV array.
The output voltage of the array is the variable input voltage of
the converter. A storage device (e.g. battery) or a cascaded
converter (e.g. DC-AC inverter) may used to keep the
output voltage Va at a constant value.
In this section we will build linear models for this PV-buck
system and design feedback systems with linear compensators
to control the input voltage of the converter.
3.1 Voltage control with single feedback loop
We need a linear model of the PV-buck system of Fig. 3. The
goal is to find a small-signal transfer function that describes the
dynamic behavior of the converter input voltage near the MPP.
The control variable is the duty cycle of the transistor
switching at frequency Is 1 Considering that is
(2)
" - - - - - - ~ - - - V
(b) MP v ~
Fig. 1: (a) Equivalent circuit of the PV array and
(b) its current vs. voltage characteristic curve.
If the array is modeled in the current source region (line the
model error increases as Vpv becomes greater than VMp.
Similarly, the error increases when is bellow V if the
array is modeled as a voltage source (line ~ ) . What model
should be considered? Next section will show why the current
source model (line is the best choice.
(2)
(1)
sc
Introduction
It is quite often to fmd publications about PV systems where
the circuit of Fig. 2a is used as an array linear model. This
circuit represents the line that contains the segment Q) of Fig.
lb. However, this circuit describes the behavior of the array
Conventional DC-DC converter models generally encountered
in the literature are not suitable for the control of the input
voltage. In the following sections we study three control
systems for the buck converter employing two modeling
strategies. The first system permits the control of the average
input voltage using the transistor duty cycle as the control
variable. In the second system the average voltage and current
of the converter are controlled and two feedback control loops
are employed. In the third system an analog inductor peak
current controller and a linear feedback voltage loop are used
to achieve the control of the input voltage of the converter.
2 Modeling the photovoltaic (PV) array
A common model for PV arrays found in the literature is the
circuit of Fig. la, where I is the PV current and R and R are
the series and shunt resistances, respectively. Fig. Ib shows the
current vs. voltage curve of the PV array obtained from this
circuit. We can notice that for output voltages lower than V
the array behaves like a current source (segmentQ) and for
voltages greater than VMP it becomes a voltage source (segment
@). Fig. 2 shows that the PV array has a maximum power
point (MPP). Ideally the array should operate at this point in
order to deliver the maximum available instantaneous power.
The parameters of the model of Fig. la may be obtained from
eqs. (I) and (2) and from information encountered in
manufacturer's datasheets, which generally provide the open-
circuit voltage of the array (V
ac
), the short-circuit current (I
sc
),
the maximum-power voltage and the maximum-power
current (IMP).
1
sufficiently high, by inspection of the circuit of Fig. 3 we can
write eq. (3) for the inductor current. The symbol > denotes
the average value of the variable (voltage or current) in a
switching period
s
.
0)
;....0"'"

o
-
array Vp,o

-;..-
...

Fig. 5: Feedback control of the buck input voltage.
Fig. 4: Bode plots of G
VD
using the current source (continuous
lines) and voltage source (dashed lines) array models.
Although neither of the models of Fig. 2 represents the PV
array in its full operating range, the circuit of Fig. 2a may be
used as a linear array model for the purpose of modeling the
PV-buck system. By comparing the Bode plots of Fig. 4 we
intuitively conclude that a feedback control system designed
for the current source region of the PV array will automatically
fit for the operation in the voltage source region.
Fig. 5 shows the feedback system employed to control the
input voltage v of the converter. The compensator may
be a proportional and integral one designed with conventional
techniques of control systems to meet any dynamic and
stability specifications [1]. Fig. 6 shows the frequency response
of compensated with = (0.2+20/s), which results
a crossover frequency of approximately 34000 rad/s and a
phase margin of 86 degrees in this example system.
(7)
(5)
(3)
=-v .d
Fig. 3: Input-controlled buck converter attached to a PV array.
d<i>
L--=<v>d-Vo
dt
-_v_-sCv+IJ - i D= 0
R
EQ
From eqs. (5) and (7) we obtain the small-signal control-to-
output transfer function of eq. (8).
EQ
and
EQ
are the voltage and resistance of the equivalent
Thevenin's circuit of the models seen in Fig. 2. If the current
source model (Fig. 2a) is used, we have: V
EQ
=I R and R
EQ
=
p +
s
. If the voltage source model (Fig. 2b) is used, we have:
V
EQ
= V
oc
and R
EQ
=R
s
.
Let us write an equation for the capacitor voltage. By
inspection of the circuit of Fig. 3 we obtain eq. (6).
VEQ- v d v (6)

R
EQ
dt
Similarly, by replacing eq. (4) in eq. (6) we find eq. (7).
In order to obtain a small-signal model we introduce small
perturbations [2] in the circuit variables, as shown in eq. (4).
The symbol -- denotes small AC perturbations and capitalized
letters denote DC steady-state values.
<v>=V+v, d=D-d , <i>=I+i
From eqs. (3) and (4), by neglecting the nonlinear products and
applying the Laplace transform, we obtain the frequency
domain equation (5).
With a simple DC circuit analysis we can find the expressions
of the steady-state values I and V that appear in eq. (9). These
expressions may be substituted in eq. (8).
Let us compare the frequency responses (Fig. 4) of
when the models of Fig. 2a and 2b are used. The Bode plots of
Fig. 4 were obtained with a PV-buck system with the
parameters of the Kyocera KC 200 GT solar panel and V =12
V, L = 2 mH, C= 450 D = 0.5.
When the current source model (Fig. 2a) is used,
presents a resonance with an abrupt phase shift near 500 rad/s.
When the voltage source model (Fig. 2b) is used,
presents a low-pass response without any important
characteristics.
v(s) REQ(VD+sL I)
GVD 2 2
d s s REQLC+sL+D R
EQ
I =_VE...;;;Q;-.-_V_ , V = V
REQD D
(8)
(9)
Fig. 6: Bode plots of GVD(s) (dashed line) and of the compensated
transfer function CVD(s) GVD(s) (continuous line).
3.2 Voltage control with inner current loop
The second control system proposed employs the small-signal
transfer and From eqs. (5) and (7)
we can write the transfer function eq. (10).
i(s) -V(l sREQC) + REQDI
ID = =
d(s) s2L REQC+sL+REQD2
Fig. 7 shows the frequency response of and Fig. 8 shows
its root locus. This system has a right half-plane zero and its
root locus is mostly on the right half-plane. The Bode plot
shows that near the crossover frequency there is a phase shift
of 180 degrees. Such a marginally stable system requires
special attention.
Fig. 9 shows the current control loop with the compensator
This compensator may be a simple proportional and
integral one, but we must carefully choose the proportional
10' 10
2
I---;.....;....O"';"';";';'O-- ~ : :.:
i-
Fig. 10: Frequency responses of and of
Up to now we have analyzed two control systems where the
average capacitor voltage v and the average inductor current
are controlled. These control systems may be implemented
with analog circuitry or with a digital processor. Integrated
circuits for duty-cycle control of DC-DC converters (also know
as voltage-control mode) are widely available in the market.
With the usage of an integrated controller all we have to do is
to implement external analog or digital compensators and
feedback paths.
3.3 Voltage and inductor peak current control
Most of integrated controllers for DC-DC converters available
in the market also work in the current-control mode (also
known as current-programmed mode). In this case instead of
controlling the output voltage of the converter (or the input
voltage in PV systems) by changing the transistor duty cycle
we control the inductor current. If we wish to control the
voltage an external control loop may be used to provide a
reference current.
8
\
-100 L . . . . - . . ~ ....................L.- ........................L-_............--................_-----......-.....-.....
10 10
Fig. 7: Bode plots of (dashed line) and of the compensated
system (continuous line).
gain in order to set a crossover frequency with a good phase
margin, keeping all system poles on the left half-plane.
Because the root-locus is mostly on the right side of the s-plane
there are strong constraints in the design of the compensator
A compensator with a low proportional gain, an integrator at
the origin and a zero placed at a low frequency makes this
system stable. If we try to set a closed-loop crossover
frequency above the resonance frequency of the system
poles fall on the outside of the left half-plane, making the
system unstable. The only option to make this system stable is
keeping the magnitude at the resonance frequency bellow 0 dB.
Unfortunately the stabilization of the current loop results a very
low bandwidth. Fig. 7 shows the Bode plot of compen-
sated with =1.3 10-4(s + 836.1)/ s.
200 400 600 800 1000 1200 1<lOO 1600
Ax
-1000 L....-..--L-_.L---L-_.L---L-_.L---L-_L....-..---J
Fig. 8: Root locus of
Fig. 9: Input voltage control system with inner current loop.
A low current control bandwidth does not mean the system will
not work properly. In Fig. 9 we notice the existence of an
external voltage control loop. The combination of the inner
current loop and the outsider voltage loop results a closed-loop
system whose transient voltage response may be as fast or as
low as we wish, according to the design of the voltage
compensator Cn(s).
Our attention now is on the design of the voltage compensator
used in the scheme of Fig. 9. Fig. 10 shows the
frequency response of where =
We must design the compensator Cn(s) in order
to stabilize the outer voltage loop. A simple proportional and
integral controller may be used and a large bandwidth with a
sufficiently large phase margin may be easily obtained. Fig. 10
shows the Bode plot of the compensated loop transfer function
where
V1
40) / s .
Fig. 11: (a) Analog inductor peak current controller and
(b) waveform of the controlled inductor current.
Fig. Ila shows the simplified scheme of the peak current
controller. A square-wave oscillator sets the flip-flop at the
beginning of the switching period T
s
. The inductor current i is
measured and compared with the reference current i
REF
When
> i
REF
the flip-flop is reset and the transistor is open until the
beginning of the next cycle. Fig. 11b illustrates a typical
inductor current waveform obtained with this control scheme.
The transistor duty cycle is automatically adjusted in order to
keep the inductor current near the desired reference current. As
we know, when the duty cycle is greater than 0.5 this control
scheme tends to become unstable [2,3]. The stabilizing ramp
signal seen in the scheme of Fig. 11a is necessary for the
proper operation of the converter. This stabilization increases
the controller error, as the peak of the controlled current gets
farther from the reference [2,3]. However, if this control
scheme is employed as the inner part of an external control
loop, the current error has no influence since the reference
current will be automatically corrected, hence cancelling the
effect of the current control error.
Fig. 12 shows how the input voltage of the buck converter may
be controlled with the peak current controller of Fig. 11a. The
input voltage v is fed back and compared with the reference
voltage VREF. The reference current i
REF
is determined by the
voltage compensator
4 Results
Fig. 14: Frequency response of and of the
compensated transfer function
(12)
(11)
(13)
<i > =

--
:,

v VEQ-V
G
V1

lREF l+s R
EQ
I C
<v>=v+v, <i>=1+7,
<i
E
IE + i
REF
l
REF
+ 7
REF
With eq. (11), using the defmitions of eq. (12), we can write
the transfer function of eq. (13).
Switching converter models with the proposed control systems
were simulated. The frequency responses of the compensated
systems were presented in the previous section. Now let us see
how the converters behave in the time domain.
Fig. 15a shows the open-loop step response of the transfer
function superimposed with the response of the
simulated switching converter. This transfer function almost
exactly represents the behavior of the switching converter
regarding the duty-cycle of the transistor and the capacitor
voltage.
Fig. 15b shows the open-loop step response of the transfer
function superimposed with the response of the
simulated switching converter. This transfer function
represents the behavior of the switching converter regarding
the duty-cycle of the transistor and the inductor current.
Fig. 16a shows the result of the simulated switching converter
controlled with the feedback system of Fig. 5, designed in
section 3.1. The input voltage v rapidly and accurately reaches
the 30 V reference.
Fig. 16b shows the response of the converter controlled with
the control scheme of Fig. 9, designed in section 3.2. The
presence of the inner current loop speeds up the transient
response. The input voltage v of the converter accurately
reaches the reference voltage. Fig. 17 shows the behavior of the
inductor current with this control scheme.
Fig. 18 shows the open-loop step response of the transfer
function superimposed with the response of the
simulated switching converter. This transfer function
aproximately represents the system behavior for small and even
large voltage oscillations near the nominal operating point
(V=24 V in this example). The transfer function exactly
describes the approximated circuit model of Fig. 13. The error
Fig. 14 shows the frequency response of the transfer function
and of the compensated transfer function G
may be a simple proportional and integral compensator.
In this example we have =
Fig. 13: Equivalent circuit of the PV-buck system
with peak current control.

cV :

Fig. 12: Voltage control employing inner
peak current control system.
The simplicity of this scheme hides an important question: how
is the dynamic response of the converter and the current
controller? We need to develop a model that describes the
behavior of the converter and helps to properly design the
compensator
In order to obtain a converter model we assume that the control
of the inductor current is instantaneous. If the current has a
small ripple Ai and a small error due to the stabilizing ramp,
the average inductor current <1> approximately follows the
reference
REF
These assumptions are good enough for the
purpose of modelling the system.
When the reference i
REF
is perturbed the inductor average
current <1> is instantaneously adjusted except by the time AT it
takes for the inductor current to rise till the value of the
reference signal at the (+) terminal of the comparator - this is
illustrated in Fig. 11b.
It would be difficult to develop a sophisticated dynamic model
for the current controller, like models found in references [4]
and [5] for the conventional converter with output voltage
control. If we pay attention to the inductor current waveform in
Fig. 11b we can notice that the current rises linearly during the
interval beginning when a reference step occurs. During
this interval T
s
) the transistor remains closed and the
duty cycle is d = 1. In other words, the converter is out of
control during the interval The transistor is simply turn on
and so it remains until the flip-flop receives a reset signal. This
behavior is extremely nonlinear and difficult to model, since
the duration of AT depends on several variables such as the
inductance the size of the reference step, and the
instantaneous input voltage. Models developed in [4] and [5]
take in account very small signal variations and do not
represent the system behavior when
s
.
For the purpose of modelling and controlling our system we
can neglect AT and consider that the peak current control is an
instantaneous process. Even if AT T
s
this is a feasible
assumption because the dynamic characteristic we are
interested in depends mainly on the capacitor charging time.
This considerably simplifies the task of modeling the PV-buck
system and permits to draw the equivalent circuit of Fig. 13.
Current is drained from the PV array and from the capacitor by
a controlled current source. It does not matter what exists after
the current source. Provided that the analog current controller
of Fig. 11a is capable to maintain the inductor current under
control the current flowing through the current source of Fig.
13 may be expressed as eq. (11), where D is the steady state
duty-cycle of the transistor.
EQ
<iB>
Fig. 15: Open-loop step responses of (a) and (b)
compared to their respective physical systems with duty-cycle control.
5 Conclusions
References
[1] Villalva, M. G., Ruppert F., E. (2007), Buck Converter with
Variable Input Voltage for PV Applications, Proc.
Brazilian Power Electronics Con! (COBEP), 2007.
[2] Erickson, R. W., Maksimovic, D., Fundamentals of Power
Electronics, Kluwer Academic Publishers, 2nd Edition,
ISBN 0-7923-7270-0.
[3] Pressman, A. (1998), Switching Power Supply Design,
McGraw-Hill, 2nd Ed., ISBN 0-07-052236-7.
[4] Tan F. D., Middlebrook R. D. (1995), A Unified Model for
Current-Programmed Converters, IEEE Transactions on
Power Electronics, Vol. 10, No.4.
[5] Middlebrook, R. D. (1989), Modeling Current-Programmed
Buck and Boost Regulators, IEEE Transactions on Power
Electronics, Vol. 4, No.1.
In section 2 we have demonstrated why the current source PV
array model is better than the voltage source one for the
purpose of modeling this PV-buck system. Although neither of
the models represents the array in its full operating range, the
current source model may be adopted and considered as a
suitable representation of the PV array. This becomes evident
when we analyse the frequency responses of the open-loop
converter using both models. The converter response is worse
when the PV array operates in the current mode, so we must
design the control system aiming this region of operation.
In section 3 we have presented several ways to obtain converter
transfer functions for the input voltage control of the buck
converter operating with duty-cycle control (voltage mode) or
with inductor peak current control (current-programmed
mode). The transfer functions show important dynamic
characteristics of the input-controlled buck converter attached
to the PV array. With these transfer functions we were able to
design compensators and feedback controllers for the control of
the input voltage of the buck converter.
The development of the linear models and transfer functions
presented in section 3 is rare in the literature. The control of the
input voltage of DC-DC converters has not been well explored.
Most of publications about PV systems which employ input
control usually show experimental results but seldom show
how the control systems were designed.
In section 4 we have presented results obtained with simulated
switching converters. These results show that the proposed
models and transfer fuctions correctly describe the dynamic
behavior of the studied PV-buck system. The open-loop
transfer functions
VD
,
D
, and
v
/) describe the behavior of
the converter with good accuracy. The results presented in the
previous section show that the compensators designed in
section 3 permit the construction of accurate, fast and stable
closed-loop feedback control systems, making feasible the
control of the input voltage of the DC-DC buck converter
employed in PV applications.
010
a


"'00

1000
....
..,.w
-
0.10
Is)

::I:J:::.:l::::::::::::::r::::.:::::::I::::::::::':::.'.':::::::1
..


Fig. 17: Inductor current of the buck converter obtained
with the inner control loop designed in section 3.2.
Fig. 18: Response of to a step at 10 ms and open-loop input
voltage of the buck converter with peak current control (iREF 15 A).
Fig. 16: (a) Input voltage (VREF 30 V) of the buck converter with
the control system designed in section 3.1 (single voltage loop) and
(b) with the control system designed in section 3.2 (voltage loop
combined with inner current loop).
between the responses of the tranfer function and of the real
system is mainly due to the approximation <1> i
REF
and
due to the assumption that O. We must pay special
attention to the fact that this system has a nonzero initial
condition, since the initial capacitor charge does not depend on
the existence of current flow through the controlled current
source of Fig. 13. So we must remark that the DC value of the
transfer function does not necessarily coincide with the
steady state input voltage
Fig. 19 shows the result of the PV-buck system operating with
the current controller of Fig. 11, with the control system of Fig.
12, and with the compensator designed in section 3.3. This
system is considerably stable due to the robustness of the
analog current controller and its dynamic response is excellent.
J1I.;Jt,r-l-
.
Hi ;.
fO _ ; (' ; r,,.. .,, ..


000
Fig. 19: Input voltage (VREF 30 V) with the control system
designed in section 3.3 (voltage loop with peak current controller).