com
ScienceDirect
Solar Energy 97 (2013) 285–292
www.elsevier.com/locate/solener
Received 3 March 2013; received in revised form 19 July 2013; accepted 15 August 2013
Available online 18 September 2013
Abstract
This paper presents a systematic way of designing control scheme for a grid-connected photovoltaic (PV) inverter featuring rapid
maximum power point tracking (MPPT) and grid current shaping without using any phase-locked loop (PLL) circuitry. A simple integral
controller has been designed for the tracking of the maximum power point of a PV array based on the extremum seeking control method
to provide a reference PV output current for the grid interface. For the grid-connected inverter, two current loop controllers have been
designed. One of the current loop controllers is designed to shape the inverter output current while the other current control loop is to
follow the reference received from the maximum power point tracking algorithm and to provide a reference inverter output current for
the PV inverter without largely disturbing the maximum power point of the PV array. Four power switches are used to achieve three
output levels and MPPT. Moreover, it is unnecessary to use any energy storage such as rechargeable battery. Experimental results
are included to demonstrate the effectiveness of the tracking and control scheme.
Ó 2013 Elsevier Ltd. All rights reserved.
Keywords: Photovoltaic; Three-level grid-connected inverter; Maximum power point tracking; Grid current shaping
0038-092X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2013.08.017
286 K.M. Tsang et al. / Solar Energy 97 (2013) 285–292
from the plot. As the grid voltage is used to derive the maximum power under varying irradiance. An improvement
required inverter output current, it can be regarded as on the constant voltage method uses the open circuit voltage
external disturbance to the PV output current control loop. to estimate the maximum power output voltage while the
One simple way to reduce its effect on the PV output cur- short circuit current method uses the short circuit current
rent is to set the bandwidth of the PV current control loop to estimate the maximum power output current. Perturb
to be very much lower than the grid supply frequency such and observe method searches for the maximum power point
that the effect of the supply voltage on the PV current con- by changing the PV voltage or current and detecting the
trol loop will be substantially attenuated by the lowpass change in PV output power. The step size for the search
characteristic of the control loop. The closed-loop charac- affects the rate of convergence of the tracking. Also, the
teristic equation can be approximated as: method may fail under the rapidly changing atmospheric
V rms K D2 L 2 V rms K I2 conditions (Petrone et al., 2011). A faster searching technique
DðsÞ ¼ LCs3 þ s þsþ ð12Þ for the PV array can be realized using the extremum seeking
Um Um
control (Heydari-doostabad et al., 2013). From the power
If one of the poles is placed at: versus current PV characteristic curve, four cases can be dis-
V rms K I2 tinguished. If DP > 0 and Dip > 0, where DP is the change of
s¼ ð13Þ power and Dip is the change of current from the PV array, the
Um
maximum power point can be obtained in increasing ip. If
and extracted from (12), the resulted second order polyno- DP < 0 and Dip < 0, the maximum power point can be
mial becomes: obtained in increasing ip. If DP > 0 and Dip < 0, the maxi-
V rms L mum power point can be obtained in decreasing ip. If
LCs2 þ ðK D2 K I2 CÞs DP < 0 and Dip > 0, the maximum power point can be
Um
2 obtained in decreasing ip.
ð1 V rmsUK2 I2 L ðK D2 K I2 CÞÞs þ V rms
Um
K I2
Fig. 4 shows the extremum seeking control block dia-
þ m
¼0 ð14Þ
s þ V rms
Um
K I2
gram in realizing the four distinguished cases for the MPPT
of PV array. Instead of a sinusoidal perturbation (Letting
If: et al., 2012), a disturbance on the PV array current of the
V 2rms K I2 L form:
ðK D2 K I2 CÞ 1 ð15Þ pt
U 2m
DI ¼ asgn sin ð19Þ
(13) can further be approximated as: T
V rms L where a is the magnitude of the disturbance and T is the
LCs2 þ ðK D2 K I2 CÞs þ 1 ¼ 0 ð16Þ
Um duration for either positive or negative half cycle, is added
The undamped natural frequency for (16) is given by: to the control loop to persistently excite the seeking proce-
rffiffiffiffiffiffiffi dure. For the same magnitude of perturbation a, the square
1 wave perturbation converges faster than the sinusoidal per-
xip ¼ ð17Þ turbation because the sum of changes in half cycle of oscil-
LC
lation for a square waveform is larger than a sinusoidal
and the damping ratio for (16) is given by: waveform. In theory, the control loop is stable for any va-
V rms ðK D2 K I2 CÞ lue of Kmppt > 0. As the extremum seeking control involves
fip ¼ ð18Þ the derivative of the power, low-pass filtering of the raw
2Cxip U m
power is required in order to remove the unwanted high
Based on (13), (17), and (18), the dominant characteris- frequency noise in the approximate derivative signal. The
tic of the PV current control loop can easily be set by KD2
and KI2. Notice that Vrms is taken as the nominal rms volt-
age of the supply grid. Fig. 3c shows the connection dia-
gram for the two current controllers and the derivation
of the duty ratio d from the control loops.
Isc = 5A
5
Isc = 4A
4
Current (A)
Isc = 3A
3
Isc = 2A
2
Isc = 1A
1
Isc = 0.5A
0
0 100 200 300 400 500 600 700
Voltage (V)
(a) Simulator output current against output voltage
2500
Vmp = 494V
2000 Imp = 4.46A
Isc = 5A
Vmp = 486V
1500 Imp = 3.55A
Power (W)
Isc = 4A
Vmp = 468V
Imp = 2.67A
1000
Isc = 3A
Vmp = 445V
Imp = 1.79A
500 Isc = 2A
Vmp = 416V
Isc = 1A Imp = 0.87A
Vmp = 380V
IV = 0.5A Imp = 0.44A
0
0 100 200 300 400 500 600 700
Voltage (V)
(b) Simulator output power against output voltage
2500
50% of the available maximum power. Even though the
inverter output current was in phase with the phase voltage,
2000
it composed of a lot of high order harmonics. This clearly
demonstrated that the input stage LC filtering was required
1500
in order to improve the average power extracted from the
PV panel. Even though the limit cycle oscillations can be
1000
removed in slowing down the two current controllers, this
will affect the grid current tracking and the speed of MPPT.
500
The implementation of the extremum seeking control loop
was on the same hardware and also sampling at 10 kHz. As
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 derivative of power with respect to time was required, a linear
phase finite impulse response differentiator with 51 coeffi-
Fig. 7. Relationship between maximum power point and the short cients and a passband up to 100 Hz and stopband starting
circuited current at 25 °C. from 250 Hz had been implemented to obtain the derivative
of power. The time delay introduced by the differentiator
performance of the interface if the filtering capacitor was was s = 0.0025 s. For the amplitude of excitation, a was set
removed. As the reference PV output current was fixed at to roughly 2% of the maximum short-circuited current of
3.55 A and the action of the current controllers were fast, the PV simulator such that it would be large enough to detect
the control system experienced limit cycle oscillations. There the difference and small enough to not disturb the system too
were large variations in the PV output voltage and the PV much. In this example, it was set to a = 0.1 A. For the setting
output current. The limit cycle oscillation also affected the of the duration of excitation, it had to be long enough for the
inverter output current. The extracted power was less than transient to settle down. As the bandwidth of the input cur-
K.M. Tsang et al. / Solar Energy 97 (2013) 285–292 291
6. Conclusions
Voltage (V)
Fig. 8. Tracking performance of the inverter interface. 400
300
rent tracking loop was set to 2.5 Hz, the transient of the cur-
200
rent would die down in around 0.25 s. In this example, the
duration T was set to 0.25 s. As the open circuit voltage 100
Voc of the PV simulator was around 600 V, from (20) the
0
gain of extremum seeking control loop became Kmppt = 53.33. 0 5 10 15 20 25 30 35 40
The maximum power point tracking control was then com- Time (s)
bined with the current tracking controllers to form a com- (a) PV output voltage
plete system. Fig. 10 shows the tracking performance of the 5
whole system. The short-circuited current was set to 5 A 4.5
and the available maximum power was 2203 W. The initial 4
PV output current was zero. Fig. 10a and b showed that 3.5
the PV output voltage gradually reduced to around 478 V
Current (A)
3
and the PV output increased to 4.28 A. Fig. 10c showed that 2.5
the maximum power point was tracked in about 20 s. The 2
1.5
1
0.5
0
0 5 10 15 20 25 30 35 40
Time (s)
(b) PV output current
2500
2000
Power (W)
1500
1000
500
extracted power
maximum power
Channel 1 - PV output voltage 0
0 5 10 15 20 25 30 35 40
Channel 2 – PV output current
Time (s)
Channel 3 – Grid voltage
Channel 4 – Inverter output current (c) Extracted power
Fig. 9. Tracking performance of the inverter interface without filtering Fig. 10. Tracking performance of extremum seeking control and current
capacitor. shaping controllers.
292 K.M. Tsang et al. / Solar Energy 97 (2013) 285–292
15 Gao, Xianwen, Li, Shaowu, Gong, Rongfen, 2013. Maximum power point
tracking control strategies with variable weather parameters for
10 photovoltaic generation systems. Solar Energy 93, 357–367.
Heydari-doostabad, Hamed, Keypour, Reza, Khalghani, Mohammad
5 Reza, Khooban, Mohammad Hassan, 2013. A new approach in MPPT
Current (A)