A Digitally Implemented Photovoltaic Simulator with

a Double Current Mode Controller

Jie Zhao and Jonathan W. Kimball
Electrical and Computer Engineering Missouri S&T
301 W. 16
th
St., Rolla, MO 65409
jazfx7@mail.mst.edu; kimballjw@mail.mst.edu
Abstract This paper presents a microcontroller- and DC-DC
converter-based photovoltaic (PV) simulator which emulates the
output characteristics of a real solar panel. In particular, the PV
simulator designed is portable and has a maximum output power
of 85W. The control mechanism is based on a double current
mode controller in combination with an IV curve look-up-table
that generates the output reference current based on the meas-
ured output voltage. Stability analysis is performed based on the
classical phase margin and gain margin criteria. Finally, the per-
formance of the PV simulator is examined in terms of its steady
state IV curve matching capabilities and the convergence time to
the PV module IV curve of interest corresponding to a step
change in load. The result shows a well behaved and responsive
PV simulator that can be treated just like a real PV module in
most situations.

Keyword digital control, photovoltaic simulator, predictive current
mode control, solar panel IV curve.
I. INTRODUCTION
Solar energy is one of the few renewable energy sources
that have gained popularity around the world due to its high
predictability and availability. One way to harness solar ener-
gy is by converting it to electrical energy using photovoltaic
(PV) modules. One challenge when interfacing devices to a
PV module is that it behaves like a current source at high out-
put current and a voltage source at low output current. This
special output characteristic of PV modules requires devices
that interface with them to have some kind of maximum power
point tracking (MPPT) capabilities in order to maximize the
efficiency of PV modules. Field testing these devices with real
PV modules can be challenging. First, PV modules are still
expensive despite the fact that their price has been on a steady
decline over the past decade. Installation of PV modules is
also an expensive and time-consuming process. Second, the
output of a PV module is dependent on the insolation level,
which varies depending on the time of year and weather of the
day. Moreover, the user has no control over the temperature,
which affects the output characteristics of PV modules as well.
A PV simulator is a device that has the same output char-
acteristics of a real PV module. It can be used within a lab at
any time of the year. There have been a few methods devised
to simulate a real PV module. The first method is to amplify
the output of a PV cell or photodiode using analog circuits [1].
This requires the use of a device that can produce or simulate
natural sunlight. The second method is to build an equivalent
circuit of a PV module using transistors and resistive networks
[2]. The last method is to digitally implement the characteris-
tics of a PV module using a DC-DC converter and a microcon-
troller [3]. This method has gained increased popularity due to
the ever increasing speed and capabilities of microcontrollers.
Digitally implemented PV simulators give users the most con-
trol. They provide a controlled environment where users can
set conditions such as temperature, insolation level, the type of
PV modules and the shading scenario.
This paper presents a portable microcontroller based PV
simulator that consists of a TMS320F28335 DSP from Texas
Instrument, a 14.8V Li-ion battery pack and a two-stage syn-
chronous DC-DC converter. A look-up-table (LUT) that stores
the IV curve data and a double current mode controller that
regulates output current is implemented in the microcontroller.
The controller is carefully designed to avoid gain scheduling
as proposed in [4].
II. HARDWARE IMPLEMENTATION
A. Power Stage of the PV Simulator
To allow for the use of a battery that has a lower terminal
voltage than the open circuit voltage of the PV module to be
simulated, a two-stage synchronous DC-DC converter is de-
vised as shown in Fig. 1. The first stage is a step-up synchro-
nous boost converter which outputs a standard voltage of 30V
from a nominal input voltage of 14.8V from the Li-ion battery.
The second stage is a step-down synchronous buck converter
which has an output voltage range of 0-30V from a standard
input voltage of 30V. One major advantage of using a syn-
chronous topology is high efficiency, which is essential for
portable applications. Another major advantage of using syn-
chronous topology here is that the DC-DC converter will nev-
er enter discontinuous conduction mode (DCM). This allows
for a much simpler controller implementation since it does not
have to make the transition between continuous conduction
mode (CCM) and DCM. However, with a synchronous topol-
ogy, the number of MOSFETs used is increased from two to
four. Dead-bands also must be introduced between Q1 and Q2,
and between Q3 and Q4 to prevent cross conduction. The in-
ductor and the capacitor for the second stage buck converter
are calculated according to
I =
v
in
4I
PP
]
sw
(1)
C
I
PP
8]
sw
v
pp
, R
LSR

v
pp
I
pp
(2)
978-1-4577-1216-6/12/$26.00 2012 IEEE 53
to satisfy current ripple and voltage ripple requirements in the
worst case scenario. The inductor and capacitor values chosen
in Section V are to satisfy a peak-to-peak ripple current of 1A
and a peak-to-peak ripple voltage of 50mV. As long as the
filter capacitor, C
B
, is relatively large, the ripple from the first
stage boost converter can be ignored and has no effect on the
output ripples of the second stage buck converter.
B. The DSP
The microcontroller shown in Fig. 1 is the 32-bit
TMS320F28335 by Texas Instrument. The microcontroller
runs at 150MHz, has a built-in floating-point-unit, 16 channels
of 12-bit ADCs, six PWM modules with a total of 12 PWM
output channels. The ADCs are used to measure the output
voltage, output current and the buck converter inductor current
by the use of sensing resistors, as needed by the double current
mode controller to be discussed later. Each PWM module has
two channels that can run in complementary mode with preset
dead-band, which is perfect for the synchronous boost and
buck converter. Hence, the boost converter will be connected
to the first PWM module and the buck converter will be con-
nected to the second PWM module.
III. PV MODULE IV CURVE IMPLEMENTATION
A. PV Module Modeling
A PV module can be modeled as a network that consists
of a current source, a diode, a shunt resistor and a series resis-
tor configured as shown in Fig. 2. The diode equation is given
as [5]
I = I
0
(c
v
D
nv
t
-1). (3)
Using simple circuit analysis, the output equation that re-
lates the output current and output voltage can be derived as
I
out
= I
ph
-I
0
_c
v
cut
+I
cut
R
s
nv
t
-1] -
v
cut
+I
cut
R
s
R
sh
. (4)
From (4), it can be seen that the IV curve of a PV module
is affected by temperature, insolation level, which determines
the short circuit current, shunt resistance and series resistance.
By varying the parameters in (4), one can model many differ-
ent kinds of PV modules under different environmental condi-
tions.
B. IV Curve Implementation
The PV module IV characteristic is represented by a look-
up-table (LUT), which is a physical memory block inside of
the microcontroller. In order to store the IV curve as a look-
up-table in the memory of the DSP, a set number of matching
output current and output voltage pairs has to be determined
according to (4). For the double current mode controller to be
presented in this paper, an output current reference has to be
generated once the output voltage is known. However, (4) has
no analytical solution because I
out
is on both sides of the
equation and is in a transcendental function. Here, Newtons
method is used to find the numeric solutions to (4). Newtons
method is given as [6]
I
out(n+1)
= I
out(n)
-
](I
cut(n)
)
]
|
(I
cut(n)
)
(5)
(I
out
) = I
out
-I
ph
+I
0
_c
v
cut
+I
cut
R
s
nv
t
-1] +
v
cut
+I
cut
R
s
R
sh
.
(6)
The derivative of (I
out
) is given as
d](I
cut
)
dI
cut
= I
0
_c
v
cut
+I
cut
R
s
nv
t
]
R
s
nv
t
+
R
s
R
sh
+1. (7)
The LUT is a 1000-element, one dimensional array that
stores the output currents corresponding to output voltages.
Here the size of the array is limited by the random access
memory (RAM) size of the microcontroller and can be in-
creased if higher resolution is necessary and RAM permits.
The LUT is updated whenever the controller receives a com-
mand to change the operating conditions, such as a different
insolation level.
IV. CONTROLLER IMPLEMENTATION
A. Converter Small Signal Modeling
With the first stage boost converter running open loop, all
control efforts are focused on the second stage buck converter.
The control block diagram is shown in Fig. 3. In order to de-
sign a controller for the simulator, the small signal transfer
functions of the second stage buck converter have to be de-

Figure 2. Equivalent Circuit of a PV module

Figure 1. PV simulator block diagram
54
rived. These small signal transfer functions can be derived
using averaged switch modeling or state space averaging; both
of them produce the same transfer functions. Here, averaged
switch modeling is used because it is easier to understand. The
process accounts for the inductor DC resistance and ESR of
the capacitor. Assuming a pure resistive load and the inductor
DC resistance and the ESR of the capacitor is a lot less than
the load resistance, the small signal transfer functions are de-
rived as
0
d(s)
=

cut(s)
d

(s)
_

in(s)=0

=
v
in
R(1+CR
csr
s)
LC(R+R
csr
)s
2
+|L+CR
csr
R
L
+CR(R
csr
+R
L
)]s+R+R
L

=
v
in
R(1+CR
csr
s)
LCRs
2
+|L+CR(R
L
+R
csr
)]s+R+R
L

0
d(s)
=
i
L(s)
d

(s)
_

in(s)=0

=
v
in
|1+C(R
csr
+R)s]
LC(R+R
csr
)s
2
+(L+CR
csr
R
L
+CR(R
csr
+R
L
))s+R+R
L

=
v
in
(1+CRs)
LCRs
2
+|L+CR(R
L
+R
csr
)]s+R+R
L

0
(s)
=
i
cut(s)

cut(s)
=
1
R

where
0
d(s)
is the small signal transfer function of output
voltage over duty cycle;
0
d(s)
is the small signal transfer function of inductor
current over duty cycle;
0
(s)
is the small signal transfer function of output cur-
rent over output voltage.
With bilinear transform, the discrete time equivalent of (8)
to (10) can be derived as
0
d(z)
= 0
d(s)
|
s=
2(z-1)
T
s
(z+1)
(11)
0
d(z)
= 0
d(s)
|
s=
2(z-1)
T
s
(z+1)
(12)
0
(z)
= 0
(s)
|
s=
2(z-1)
T
s
(z+1)
. (13)
B. Inner Predictive Current Mode Controller
Predictive current control is an accurate digital control
technique that is based on inductor current predicted by sam-
pled inductor current and output voltage. There are three kinds
of predictive current mode controls peak current control,
average current control and valley current control. There are
four kinds of modulation methods trailing edge, leading edge,
trailing triangle and leading triangle. Each of the three current
control methods has to be paired with the correct modulation
method in order to be stable over the whole range of the duty
cycle. For the buck converter used, its duty cycle range is 0-1,
so it is important to pair the current control method with the
right modulation method. Here average current control is used,
and it is paired with trailing triangle modulation to give stabil-
ity over the whole duty cycle range of the buck converter. Re-
gardless of the current control method, the predictive current
mode control law for a buck converter is given as [7]
J|n +1]
= -J|n] +
L

in
1
s
(i
c]
|n] -i
L
|n]) +
2
c

in
. (14)
The small signal discrete time transfer function of duty
over current error is given as [8]
J
(z)
= F
m(z)
(i
c](z)
-i
L(z)
) (15)
where
F
m(z)
=
L
v
in
1
s

1
z+1
. (16)
The switching diagram is shown in Fig. 4. Sampling oc-
curs at the beginning of each switching period. It can be seen
that by the end of the second switching period, the average
inductor current reaches the reference current set at the first
switching period.
Figure 3. Control block diagram of the PV simulator

Figure 4. Switching diagram

55
C. Outer PI controller Design
From the block diagram shown in Fig. 3, the loop gain of
the plant without the compensator 0
c(z)
can be found using
loop reduction or Masons Law. Assuming that the gains from
the ADCs have been adjusted to one, the loop gain of the plant
is found as
I
(z)
=
P
m(z)
u
d(z)
u
i(z)
1+P
m(z)
u
id(z)
. (17)
It is possible to design a PI controller in the z domain, but
the system bandwidth, gain margin and phase margin will not
be obvious. In order to see the system bandwidth, gain margin
and phase margin, (17) is converted to the continuous time s
domain by bilinear transform as
I
(s)
= I
(z)
|
z=
1+
T
s
2
s
1-
T
s
2
s
=
2L(1+CR
csr
s)[1+
T
s
2
s

(s)
(18)

(s)
= SCRI
s
Is
2
+(2CRI +SII
s
+4CRR
cs
I
s
+
4CRR
L
I
s
)s +2I +4RI
s
+4R
L
I
s

= SCRI
s
Is
2
+(2CRI +SII
s
+4CRR
cs
I
s
+
4CRR
L
I
s
)s +2I +4RI
s
. (19)
The compensator, 0
C(s)
, is designed based on the common
PI controller as
PI = K
P
+
K
I
s
=
K
p
_s+
K
I
K
P
]
s
. (20)
If the zero of the compensator is placed over the dominant
pole of I
(s)
, the band width of the system will be equal to the
cross over frequency of the compensated loop gain. The DC
gain of the PI controller will be chosen so that the bandwidth
is maximized while satisfying the phase margin greater than
75 and gain margin greater than 6 dB requirements.
However, the small signal transfer functions of the buck
converter are dependent on the load condition. In order to
avoid having to implement mode switching or gain scheduling
in the controller, it is necessary to find the load conditions that
place the most stringent requirements on the PI controller. If
the load resistance is assumed to be between 0.25 and 200,
then a PI controller can be designed to satisfy both load condi-
tions. Using Matlab, a range of operating conditions can be
considered. The location of the zero is determined primarily
by the light load condition, while the choice of gain is deter-
mined primarily by the heavy load condition. Considering
both, a stable controller may be found for the entire operating
range. The PI controller is chosen to be
0
c(s)
=
2000(1+0.0039s)
s
. (21)
The discrete time equivalent of 0
c(s)
is transformed as
0
c(z)
= 0
c(s)
|
s=
2(z-1)
T
s
(z+1)
=
7.61z-7.59
z+1
. (22)
The bandwidth of the system increases as load resistance
decreases. The bode plots of the close loop gains when the
system is operating at minimum load resistance, maximum
load resistance and maximum power point load resistance is
shown in Fig. 5.
D. Stability Analysis
The loop gain of the whole closed system, which includes
the LUT and the plant, is found to be
H
ss(z)
= -0
L01
1
((z))
u
c(z)
1+1
((z))
u
c(z)
(23)
0
L01
is the gain from the LUT. From (23), one can tell
that the stability of the whole system depends on the gain of
the LUT since the plant is designed to be stable. The IV curve
of the PV module to be simulated in Section V is shown in Fig.
6 along with the corresponding 0
L01
. From the figure, one can
tell that the absolute value of the IV curve slope increases as
the voltage increases, and the worst case gain is -1.797. At this
Figure 5. Compensated plant loop gains for multiple load conditions

Figure 6. PV module IV curve and slope

56
worst case gain, the loop gain of the whole system is shown in
Fig. 7. As seen in Fig.7, the system still has adequate gain and
phase margins and is stable.
V. EXPERIMENTAL RESULTS
The PV simulator has been built with the circuit parame-
ters shown in Table 1. The PV simulator is set to simulate the
multi-crystalline PV module model SW-S85P from SunWize.
The outputs of the PV simulator when connected to constant
current loads are shown in Fig. 8 for 100%, 60% and 20%
insolation levels. From Fig. 8, it can be seen that the output of
the PV simulator follows the ideal IV curves consistently
without encountering stability issues. The load step response
of the PV simulator corresponding to five different operating
points on the IV curve under 100% insolation is shown from
Fig. 9 to Fig. 13. The top trace is the output voltage, the mid-
dle trace is the output current and the bottom trace is the load
switch. Therefore, the PV simulator can be perturbed, as by a
MPPT controller, with a sampling frequency of up to 2 kHz.

Figure 7. Bode plot of worst case system loop gain
TABLE 1. PV SIMULATOR PARAMETERS
DC-DC converter parameters
Switching Frequency (f
sw
) 100 kHz
Inductor (L
1
) 138 H
Inductor (L
2
) 138 H
Input Capacitor (C
in
) 560 F
Filter Capacitor(C
B
) 1 mF
Output Capacitor (C
out
) 560 F
Photovoltaic module parameters
Short Circuit Current (I
sc
)(100% Insolation) 5.4 A
Open Circuit Voltage (V
oc
)(100% Insolation) 22 V
Internal Series Resistance (R
s
) 342 m
Internal Shunt Resistance (R
sh
) 1.11 k
Max. Output Power (P
max
)(100% Insolation) 85 W
Temperature 300 K



Figure 8. PV simulators performance with constant current load

Figure 9. Step response when load changes between 1 and 0.9.


Figure 10. Step response when load changes between 2.22 and 2.

Figure 11. Step response when load changes between 3.2 and 2.9.

57
VI. CONCLUSION
This paper presents a photovoltaic simulator that emulates
a real PV module. The special output characteristics of a solar
panel can be emulated by combining a microcontroller and a
DC-DC converter. With a double current mode controller, the
stability of the system can be guaranteed over a wide range of
load conditions. The speed and convergence path of the simu-
lator is also examined to allow users make informed decision
to maximize the effectiveness of the simulator.
ACKNOWLEDGMENTS
This work was supported in part by the National Science
Foundation under grant ECCS-0900940.
REFERENCES
[1] H. Nagayoshi, Characterization of the module/array simulator using I-V
magnifier circuit of a pn photo-sensor, 3
rd
World Conference on Photo-
voltaic Energy Conversion, vol. 2, May 2003, pp. 2023-2026.
[2] A. Koran, K. Sano, R. Y. Kim and J. S. Lai, Design of a photovoltaic
simulator with a novel reference signal generator and two-stage LC out-
put filter, in IEEE 2009 ECCE, Sept. 2009, pp. 319-326.
[3] E. Koutroulis, K. Kalaitzakis and V. Tzitzilonis, Development of a
FPGA-based system for real-time simulation of photovoltaic modules,
in proc. IEEE RSP06, Jun 2006, p. 200.
[4] Y. Li, T. Lee, F.Z. Peng and D. Liu, A hybrid control strategy for pho-
tovoltaic simulator, in IEEE 2009 Applied Power Electronics Confer-
ence, 2009, pp. 899-903.
[5] D. A. Neaman, "Microelectronics Circuit Analysis and Design 3
rd
edi-
tion," McGraw-Hill, New York, 2007.
[6] J. Stewart, Calculus, 5
th
edition, Thompson, 2003, pp. 294-295.
[7] J Chen, A. Prodic, R. W. Erickson and D. Maksimovic, Predictive
digital current programmed control, in IEEE trans. Power Electronics,
vol. 18, issue 1, Jan 2003, pp. 411-419.
[8] Y. S. Lai and C. A. Yeh, Predictive digital-controlled converter with
peak current-mode control and leading-edge modulation, in IEEE trans.
Industrial Electronics, vol. 56, no. 6, June 2009, pp. 1854-1863.


Figure 12. Step response when load changes between 4.4 and 4.

Figure 13. Step response when load changes between 9.6 and 8.7.
58