You are on page 1of 11

Available online at www.sciencedirect.

com

ScienceDirect
Solar Energy 132 (2016) 617–627
www.elsevier.com/locate/solener

Jumping maximum power point tracking method for PV array


under partially shaded conditions
Yie-Tone Chen ⇑, Yi-Cheng Jhang, Tsung-Hsiu Kuo, Ruey-Hsun Liang, Chung-Wen Hung
Department of Electrical Engineering, National Yunlin University of Science and Technology, Douliou, Yunlin 64002, Taiwan, ROC

Received 11 November 2015; received in revised form 8 March 2016; accepted 9 March 2016
Available online 16 April 2016

Communicated by: Associate Editor Mario A Medina

Abstract

When the shaded condition occurs in the solar array, its P–V curve will produce many power peaks. The conventional maximum
power point tracking (MPPT) method may fall into the local MPP (LMPP). However, time is wasted a lot by the full-range searching
method, although the method can find out the global MPP (GMPP). The equal-power jumping GMPPT method for the series solar array
is proposed in this paper. The tracking time between the power peaks can be reduced. The auto-scaling variable step-size (ASVSS) MPPT
method is then associated with the jumping method to further improve the tracking performance. When the next peak power is lower, the
tracking of this lower peak can be skipped to shorten the tracking time. Moreover, compared with the value obtained from the datasheet,
the real open-circuit voltage of the solar array can be directly achieved and applied in the execution of the proposed tracking algorithm.
Therefore, the proposed method is more adaptive to the environmental variation.
Different shaded situations are simulated. The outcomes with the equal-power and non-equal-power methods are also given. Exper-
iments are finally conducted to verify the validity of the proposed method even under the partially shaded conditions.
Ó 2016 Elsevier Ltd. All rights reserved.

Keywords: Global maximum power point tracking; Partially shading; PV arrays in series

1. Introduction have a quick response when the irradiation changes. How-


ever, these methods are only for the solar system with a sin-
Because the output characteristic curve of the PV array gle maximum power point. If the solar system exists
is nonlinear in solar power systems, the MPPT technique is multiple local peak power points, it will be necessary to
necessary to be applied. There are a lot of methods that evaluate the tracking methods of multiple power points
have been implemented, such as perturb-and- observe to achieve the fast and stable tracking performance.
method (Femia et al., 2005,2007; Piegari and Rizzo, PV arrays are susceptible to cell damage and may be
2010), incremental conductance method (Safari and shaded by clouds or leaves in the outdoor. Under these
Mekhilef, 2010; Xuesong et al., 2010), climbing method conditions, the P–V curves of the PV arrays can appear
(Xiao and Dunford, 2004), and single sensor method multiple local peak power points from the original single
(Jiang and Abu Qahouq, 2012). The improvement and maximum power point. At these local peak power points,
combination of these methods are further studied to reduce the highest point of all is called the Global maximum
the tracking time, improve the steady-state stability, and power point (GMPP), and the other points are called the
local maximum power points (LMPP). The conventional
⇑ Corresponding author. MPPT methods could fall into the LMPP so that the

http://dx.doi.org/10.1016/j.solener.2016.03.020
0038-092X/Ó 2016 Elsevier Ltd. All rights reserved.
618 Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627

GMPP of the PV arrays cannot be tracked accurately than by using duty cycle. When the number of PV modules
under partially shaded conditions. In order to improve is changed, the setting of parameters will also need to be
the disadvantages of the conventional MPPT methods, changed.
the research in Kazmi et al. (2009) proposed a global scan- The paper in Tey and Mekhilef (2014) proposed a
ning method to scan the P–V curve entirely and store each MPPT method without the need to detect the output volt-
LMPP. Then the duty cycle of GMPP is output after com- age of the converter even under the partially shaded condi-
paring all local maximum power points. This method is tions. It has the advantage of fast response during the load
feasible, but takes quite a long time. and irradiation changes. But this method will fall into
The paper in Kobayashi et al. (2006) proposed a two- LMPP in some cases and needs to know the open-circuit
stage tracking method. When the shading occurs, the first voltage of the PV module in advance. The voltage-
step is that the operating points will jump onto the load line window GMPP method is presented in Boztepe et al.
Rpm (Vpm/Ipm), where Vpm is 0.8 times of the open-circuit (2014). However, the open-circuit voltage and short-
voltage and Ipm is 0.9 times of the short-circuit current. circuit current also need to be known in advance.
The second step is that using the incremental conductance The above methods need to know the open-circuit volt-
method to track the MPP. This method applies to the situ- age, and (or) short-circuit current of PV arrays. They have
ation where the GMPP is on the right place of load line good results in fixed or specific systems, but the parameters
(Rpv > Rpm). However, if the GMPP is on the left place need to be changed after replacing the PV system. There-
of load line (Rpv < Rpm), it will track the LMPP. Further- fore, this paper proposed the equal-power jumping MPPT
more, this method needs extra circuits to detect the open- method to reduce the tracking time between peak powers
circuit voltage and short-circuit current and causes the and combined with the auto-scaling variable step-size
complexity of the circuit. The paper in Ji et al. (2011) also (ASVSS) MPPT method (Chen Y. T. et al., 2014) to
proposed the concept of load line, and it counts the series quickly track each local peak power. It becomes a novel
and parallel number of PV arrays when the operating two-stage MPPT tracking algorithm and does not need
points jump onto the load line. This approach improves to know the open-circuit voltage of the PV module in
the tracking performance, but still exists the situation of advance. Because the proposed algorithm starts from zero
the mistake tracking. The open-circuit voltage and short- duty cycle to track, the real open-circuit voltage of PV
circuit current of the solar array are also necessary to be arrays can be directly obtained in the execution of the
known. Although the parameters of the open-circuit volt- tracking algorithm. This method can be used in different
age and short-circuit current can be supplied directly by series PV module systems. And if the next peak power is
the datasheet of PV module, the parameters still need to smaller than the present one, it will not track the small
be reset when changing other types of PV module and they peak power by the proposed equal-power judgement
may vary in the real applications. method to reduce the tracking time. The straightforward
Particle Swarm Optimization (PSO) algorithms are parts and simple characteristics of the full-range searching
of artificial intelligence algorithms. PSO has the advantage method are reserved in the proposed method; however,
of the ability to adapt to the dynamic environment. It is the tracking time of the proposed method is reduced
good for nonlinear curve of PV arrays under partially greatly.
shaded conditions. The papers in Ishaque et al. (2012), In this paper, the P–V curves with three peak powers
Miyatake et al. (2011), Lian et al. (2014), Liu et al. will be used to simulate and describe the tracking situations
(2012), Chao (2014) use this method to achieve the maxi- of the GMPP in different peak powers. Then, through the
mum power point tracking of PV arrays under partially experiments, the tracking accuracy and stability of the pro-
shaded conditions. The duty cycle is generally used as the posed method in practical conditions can be validated.
particle and several duty cycles are generated randomly,
then we can pass messages between particles to find the glo- 2. Characteristic of solar cell
bal optimal solution. More particles get higher accuracy,
but take a longer time. Moreover, the parameters must The short-circuit current and open-circuit voltage of the
be set by the researcher, and it will fall into the local opti- solar cell will be affected by the temperature and irradia-
mal solution if the parameters are not set rightly. The paper tion. Therefore, the maximum power point of the solar cell
in Miyatake et al. (2011) pointed out that the large inertia will be different when the environment is changed. This
weight (x) is not easy to make the algorithm fall into the paper uses the modules of MSX-60 (MSX-60 and
local optimal solution in the initial stage and little inertia MSX-64 datasheet, 1997) manufactured by Solarex. This
weight can accelerate the convergence and improve the solar module is in series of 36 solar cells. Its open-circuit
problem of the parameter setting. Then the paper in voltage is 21.1 V, short-circuit current is 3.84 A, and rating
Ishaque and Salam (2013) pointed out that the LMPP power is 60 W (at 25 °C and 1000 W/m2). The current in
nearly locates at 80% of the open-circuit voltage each module is the same when the solar arrays are without
(n  0.8  Voc_module, n is the number of PV modules and shading. When a module in series arrays is shaded, the
Voc_module is the open-circuit voltage of a single PV mod- current generation of this module will reduce. Because
ule). The tracking speed is faster by using Voc as particle the current generations of other modules without shading
Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627 619

keep the original values, this could cause the overheating to converter can be represented by Eqs. (1) and (2). The for-
damage the shading module in series arrays. Therefore, mula for impedance matching can then be derived as Eq.
each solar module is in parallel with a bypass diode to pre- (3). By controlling the duty cycle D of the converter, the
vent the possible damage when the generating currents of system can complete the GMPPT even the solar array is
no shading modules flow fully through the shaded module. under partially shaded conditions.
Because of bypass diodes, the P–V curve of the solar array
V in ¼ V out ð1  DÞ ð1Þ
will appear multiple MPPs under the partially shaded con-
ditions. Fig. 1 shows two solar modules in series to form I out
I in ¼ ð2Þ
the solar array and one of the solar modules is shaded. ð1  DÞ
There are two peaks on the curve of power-versus- Eq. (1) is divided by Eq. (2), then:
voltage and the stepped waveform is presented on the curve
of current-versus-voltage. However, even if the curve exists Rin ¼ RLoad ð1  DÞ2 ð3Þ
multiple MPPs, the effective use of the maximum power
where Rin ,V in =I in and RLoad ,V out =I out . Eq. (3) can be rep-
output is still required. Therefore, this paper proposes a
resented by the symbols of the solar system.
tracking method for tandem solar arrays to track the
GMPP under partially shaded conditions. V PV 2
¼ RLoad ð1  DÞ ð4Þ
The conventional MPPT methods could fall into the I PV
LMPP under partially shaded conditions; therefore, they If Eq. (4) is represented by Eq. (5), RLoad can be calcu-
are unable to accurately find out the GMPP. This results lated from VPV, IPV, and D. When the algorithm detects the
in less effective use of solar power. The paper in Ji et al. load variation, the change in the load can be estimated by
(2011) proposed a global searching method to scan the Eq. (5). If Eq. (4) is represented by Eq. (6), the duty cycle
P–V curve entirely and store each LMPP and its duty cycle, can be calculated by VPV, IPV, and RLoad. If the values of
and then the duty cycle of GMPP is output by comparing VPV and IPV that the proposed algorithm desires to jump
all of LMPPs. This method is feasible but takes quite a long to are known, the corresponding duty cycle can be calcu-
time; therefore, the new tracking method is proposed to lated by Eq. (6).
shorten the tracking time.
According to references (Patel and Agarwal, 2008; V PV 1
RLoad ¼ ð5Þ
Veerasamy et al., 2014; Nasir and Zia, 2014; Chen K. I PV ð1  DÞ2
et al., 2014; Boztepe et al., 2014), if the solar array consists rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
V PV
of n solar modules in series, and when the n solar modules D¼1 ð6Þ
I PV  RLoad
occur the shading in different situations individually, it will
generate n peaks on the curve of power-versus-voltage and
each of the peaks will lie around in the voltage position of 3. Proposed method
n  0.8  Voc_module.
The paper in Tey and Mekhilef (2014) is through the 3.1. Description of the proposed method
relationship between the input (Vin, Iin) and output (Vout,
Iout) of the DC–DC converter to derive the formula for Fig. 2 shows the diagram of the tracking root. The pro-
impedance matching. The relationship of the Boost posed method starts to track from the duty cycle being
zero. Therefore, the real open-circuit voltage Voc_array of
the solar array can also be detected at first time. Then,
the open-circuit voltage Voc_module of a single module can
I pv
be obtained after dividing the open-circuit voltage of the
Bypass
solar array with the series number of modules. The ASVSS
Diode MPPT method is then used to track the local MPP. When
point P1 is tracked, the voltage, duty cycle, and power of
the solar array at this point are recorded. Afterward, the
duty cycle is set to be 0.95 to jump to point a. The aim is
V pv to obtain the current value I2 at point a as shown in
Fig. 2(b). According to the power value at P1 gotten in
the above step and Eq. (7), the reference voltage Vref can
be derived. This voltage value is substituted into Eq. (6)
to calculate the duty cycle, and then the algorithm can
Shading jump to point b according to this duty cycle. Point b is
Module the so-called ‘‘equivalent power point” of P1. Therefore,
the tracking time between point a and point b is reduced.
Fig. 1. The P–V and I–V curves of the solar series array under the After tracking point b, the algorithm continues to track
partially shaded condition. with the ASVSS MPPT method.
620 Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627

further tracked. If the equal power point does not exist,


P3
60 P2 jump
tracking the algorithm will directly jump to the next peak to skip
d the track of P3 to reduce the tracking time. The local
50 c
tracking
MPP occurs about at the neighborhood of n  0.8 
40 Voc_module. This characteristic is here used to detect current
Power (W)

b P1
jump
tracking
I3 in the above process. Even its value is variable; the cur-
30
rent value of I3 at the neighborhood of P3 is still very close
20 jump
as shown in Fig. 2(b). The simulated results show that the
current values only vary between 2.1 A and 2.2 A when
10 a
Vref Vref1 VOC_compute
Duty 0
the voltages vary between 0.76  Voc_module and 0.84 
Duty 0.95
Voc_module. Therefore, no matter 0.76  Voc_module or
10 20 30 40 50 60 0.84  Voc_module is used, current I3 can still be detected suc-
Voltage (V) cessfully and does not affect the following tracking process.
(a) The most right peak in the PV curve has been tracked at
the start of the algorithm. To avoid tracking this peak
repeatedly, the algorithm must judge whether the left peak
4 I2
P2 just next to the most right peak has been tracked. As shown
a b
in Fig. 2(a), when peak P3 has been tracked, the next peak
3 is peak P1. The difference between these two peaks is about
0.8  Voc_module. So, if the voltage difference of VP1 (the
Current (A)

I3 P3
2 most right peak) and VPn (the local peak) is between 0.8
d
and 1.2 times of 0.8  Voc_module, the left peak just next
1 I1 P1 to the most right peak (here, it is P3) will be judged to have
Vp2 VOC_compute
×VOC_module been tracked. Then, the algorithm stops the action. All the
recorded peak powers are then compared to output the
10 20 30 40 50 60 duty cycle of the GMPP to complete the GMPP tracking
Voltage (V) of the solar array under the partially shaded conditions.
(b) If the next peak is lower than the present peak, as shown
Fig. 2. The diagram of the tracking root (a) curve of power-versus-voltage
in Fig. 3, the power of peak P3 is lower than that of peak P2;
(b) curve of current-versus-voltage. the peak P3 will not be tracked. The equivalent power
method is adopted to check whether to continue the next
P PV
V ref ¼ ð7Þ peak track. When the peak at P2 has been tracked, the next
I PV step is to detect the current at the next peak. The method
has been discussed in the above description. The algorithm
When we track point P2, the voltage, duty cycle, and
calculates Voc_compute first, and then the duty cycle is derived
power at this point are also recorded. From MSX-60 and
by Eq. (6) to detect current I3. However, when current I3 has
MSX-64 datasheet (1997), Veerasamy et al. (2014), Nasir
been detected, I3 and Voc_compute are not substituted into
and Zia, (2014), Chen Y.T et al., (2014), Boztepe et al.,
Eq. (6) first; instead I3 is substituted into Eq. (7) to get a ref-
(2014), one can know that the maximum power point of
erence voltage with the power at P2. This reference voltage
each peak is located at about the voltage value of
Vref1 is substituted into Eq. (6) to get the corresponding
n  0.8  Voc_module.
duty cycle at this reference voltage. Then, the tracking is
As shown in Eq. (8), the voltage Voc_compute near the next
MPP peak P3 can be known after adding 0.8  Voc_module to
the value of VP2. This voltage near the next peak P3 can be
60
substituted into Eq. (6) to calculate the duty cycle to detect P2
current I3 at the next peak P3. 50
jump
P3
V oc compute ¼ V Pn þ 0:8  V oc module n ¼ 2; 3; 4 . . . ð8Þ 40 d
jump
P1
Power (W)

After detecting current I3, one can obtain a new refer- 30 e


ence voltage Vref1 with Eq. (7) (V ref1 ¼ PI32 ). By substituting
20 ×Voc_module
this new reference value as a new VPV into Eq. (6), the cor-
responding duty cycle at this new reference voltage can be 10
Vref1 Voc_compute
obtained. Then, the tracking will move to the equal-power
test point c to see whether the equal power point exists. If
10 20 30 40 50 60
the equal power point exists, the algorithm will jump to
Voltage (V)
point d. Therefore, the tracking time between P2 and d
can be reduced. Afterward, the local MPP P3 will be Fig. 3. The diagram when the power at P3 is lower than the power at P2.
Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627 621

moved to the test point of the equal power. The power Pref1 2 (block 9–10). The current with duty cycle being 0.95
at this reference voltage is compared to see whether the can be obtained. According to Eqs. (7) and (6), we can cal-
equal power point exists. When the power at P2 is greater culate the reference voltage and the duty cycle, and the
than Pref1, the equal power point does not exist. If the equal equivalent power point of P1 can be obtained. Then, a is
power point exists, the algorithm will come back to the orig- set to be zero and the algorithm comes back to block 2
inal flow to continue the tracking. If the equal power point (block 11–12). When a = 0, the MPPT is conducted (block
does not exist, Voc_compute will add the voltage of the single 3–5). The values of the power, duty cycle, and solar voltage
module to jump to the next peak at e point to detect the cur- at this peak are recorded. Then, a is set to be 3 (block 13) to
rent. The same equal power method is used once again to execute the judgement of the final condition (block 14–15).
see whether the equal power point exists. If the equal power If the condition is not satisfied, the tracking won’t be com-
point exists, the algorithm will continue its tracking or the plete, and the algorithm will continue to jump to the next
algorithm will jump to the next peak. peak to track. According to Eq. (8), Voc_compute is calcu-
The flowchart of the proposed method is shown in lated and substituted into Eq. (6) to calculate the duty cycle
Fig. 4. The program initiates the parameters first. The to detect the current of the next peak and a is set to be 4
power array, duty cycle array, and solar voltage array are (block 16). When a = 4 is satisfied, the detected current
claimed to access the value of each peak (block 1). The value and Voc_compute will be substituted into Eq. (6) once
voltage and current of the solar array are then detected again to make the detected current value more accurate
(block 2). When a = 0, the maximum power tracking is and a is set to be 5 (block 17–18). When a = 5 is satisfied,
conducted (block 3–5). Because the starting point is when the current value obtained in the last step is substituted into
the duty cycle is zero, the most right peak of the P–V curve Eq. (7) to calculate the reference voltage and duty cycle.
is tracked when the condition of block 5 is satisfied. The Then, a is set to be 6 (block 19–20). When a = 6, the equal
values of the power, duty cycle, and solar voltage at this power point is judged whether it exists or not (block
peak are recorded into the arrays (block 6). When j = 1 is 21–22). The present power is compared with that of the
satisfied, then a = 1 (block 7–8). The duty cycle is then previous peak. If they have equal power, it will mean the
set to be 0.95 and a is set to be 2 to come back to block equal power point exists. If the power of the previous peak

Start

a 0, j 0
1
P[n], D[n], V [n]
PV
Sense Voc_array
Voc_module ( Voc_array / n )

Sense
2
Vpv, Ipv
4
3 Y ASVSS
a 0
10 MPPT
N
Duty 0.95 Y 9 N
a 1
a 2 dI/dV I PV /VPV 0.02
12
N
11 5
V = PPV [j]/I PV
ref Y a 2 Y
D = 1 - sqrt(Vref/(IPV * ZLoad))
N P[j] PPV
a=0 Y 14
a 3 D[j] Duty 6
N V [j] V
17 PV PV
Y Irradiation & load
a 4 j
variation subroutine
N Y N 19 N Y

15 a 5 j 1
Vmpp VPV_now
V PV [1] V Pn N 7 29
0 .8 1 .2 Y 21 Impp IPV_now
0 .8Voc _ module N
a 6
Y
N 13 a 3 a 1 8 Y dV 0 & dI 0 30
25 27 MAXP
Call MAXP or
a 7
dV 0 & dI 0
Y 28
26 N Max=0,j=0
Duty D[e] Call irradiation & load
variation subroutine
a 7 dV 0 & dI 0 31
N
16 22 or Duty Duty
V oc _ compute =V Pn 0 .8V oc_ module N dV 0 & dI 0 N
Is 34 P[j] Max j++
D= 1-sqrt (V oc_compute / ( I PV *Z load ))
equal power? Y
a= 4
ASVSS MPPT 32
20 Y Y
24 Max P[j] Y
18 23 e j j n
V = PPV [j]/I PV Measure Vpv and Ipv to calculate
ref V Pn V oc _compute j
D = 1 - sqrt(Voc _compute /(IPV * Zout )) D = 1 - sqrt(Voc _compute /(IPV * Zout )) new Rload by equation(5) ,and
D = 1 - sqrt(Vref/(IPV * ZLoad)) a=0 a=3 substitute Vmpp,Impp and new 33
a=5 N
a 6 Rload into equation(6) to calculate
the Duty return

Fig. 4. The flowchart of the proposed algorithm.


622 Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627

is greater than that of the present one, the equal power


point will not exist. If the equal power point exists, the
algorithm will track from the voltage of Voc_compute and a
will be zero to continue the tracking algorithm with the
ASVSS MPPT method (block 23). If the equal power point
does not exist, VPn will be set to be Voc_compute and a will be
3 (block 24). Then, the objective is to judge whether the ter-
mination condition is satisfied. If the condition is not satis-
fied, the algorithm will continue to track. If the termination
condition is satisfied, the tracks of all peaks will be com-
plete. The algorithm will enter the comparison of all values
in the power array to obtain the GMP (block 25). The duty
cycle of the GMP is output and a is set to be 7 (block 26).
When a = 7, the algorithm enters the subroutine of the
judgement of the load variation and irradiation change
(block 27–28). The voltage and current of the present
MPP are given to Vmpp and Impp (block 29) and the algo-
rithm enters the judgement of the load variation (block
30). The load increment causes the voltage increment and
current decrement of the solar array and the load decre-
ment causes the voltage decrement and current increment
of the solar array. When the condition is satisfied, the dis-
turbed voltage and current of the solar array are measured
and substituted into Eq. (5) to calculate the new load value.
The new load value and the Vmpp and Impp obtained a while
ago are substituted into Eq. (6) to derive the corresponding
duty cycle to make the system come back to the MPP. If
the load variation condition is not satisfied, the irradiation
will be judged to see whether it is changed. If the irradia-
tion is changed, the ASVSS MPPT method will be used
to track again (block 31–32). If the irradiation is not chan-
ged, the duty cycle will be output (block 34) and the algo-
rithm will come back to block 30 to judge whether the load
or irradiation is changed continuously.

3.2. Mathematical analysis of the proposed method

The mathematical analysis of the proposed method is


discussed here. The related calculations are with the help
of the mathematical tool of Mathcad. The result in the fol-
lowing Fig. 5(a) will be used to match with the illustration.
According to Fig. 2, the duty cycle is first set to be zero to
detect the Voc,array in the beginning. It is 56.5 V (as shown
in Fig. 5(a)). Then, the program enters the MPPT tracking
to track P1 (in P1, D = 0.254, Ip1 = 0.81 A, Vp1 = 47.7 V,
Pp1 = 38.65 W). The current shown in Fig. 2(b) can be
expressed as follows.
8    
>
> I  I exp qV PV
 1 if I PV > I 3 ðaÞ
>
>
ph1 S AkT n
<
1
   
qV PV
I PV ¼ I ph2  I S exp AkT  1 if I 1 < I PV < I 3 ðbÞ
>
>  
n2
 
>
>
: I  I exp qV PV  1 if I < I ðcÞ
ph3 S AkT n3 PV 1

ð9Þ Fig. 5. The simulation results of the solar array under the partially shaded
condition (a) curve of power-versus-voltage (b) full-range searching
where Iph1,2,3 are the photocurrent, IS is the saturation method (c) linear function method (d) modified INC method (e) proposed
current, k is the Boltzmann constant, A is the ideality method.
Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627 623

factor, q is the elementary charge, and n1,2,3 are the num- 4. Simulation results
bers of the solar cells of the solar module.
Afterward, the program jumps to the point where the This paper uses the simulation software Powersim to
duty cycle is 0.95 to detect the current of I2 (block 10). simulate the solar generation system. The boost converter
Therefore, Ipv = Isc = I2 = 3.87 A. Then, we use Eq. (7) to is the main structure, and the parameters in it are that
obtain Vref (block 12), and use Eq. (6) to obtain the duty the input capacitance (Cin) is 470 uF, inductance (L) is
cycle at point b. 5 mH, output capacitance (Co) is 47 uF, and switching
frequency is 20 kHz. The solar module is the MSX-60
V ref ¼ PI 21 ¼ 38:65
3:87
¼ 10 ðVÞ (MSX-60 and MSX-64 datasheet, 1997). Three modules
qffiffiffiffiffiffiffiffiffiffiffiffi
of MSX-60 are in series to form the solar array.
D ¼ 1  3:87100 10
¼ 0:84
To simulate the real situation, the temperature of the
solar array is set to be 55 °C. Therefore, the open-circuit
The algorithm then enters the MPPT tracking of P2 to
voltage of the solar array is 56.5 V, short-circuit current
obtain the parameters at P2 (D = 0.795, Ip2 = 3.55 A,
is 3.84 A, and rated power is 157.3 W. Fig. 5(a) shows the
Vp2 = 14.76 V, Pp2 = 52.44 W). With Eq. (8), the algorithm
power-versus-voltage curve of the solar array under the
jumps to point d (block 16).
shaded situation. The GMPP is located in the middle of
56:5 the three peaks. Fig. 5(b) shows the results with the full-
V PV ¼ 14:76 þ 0:8  ¼ 29:83 ðVÞ
3 range searching method. It takes about 8.67 s to track.
The tracking results with the linear function method and
Substituting the voltage at this point and the current at modified INC method are shown in Fig. 5(c) and (d). The
P2 into Eq. (6), one can get a new duty cycle. track with the linear function method is fault and the time
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi with the modified INC method is also longer. As for the
29:83
D¼1 ¼ 0:71 result of the proposed method, it is shown in Fig. 5(e). It
3:55  100 takes 0.73 s to track the most right peak. At the interval
With Eqs. (4) and (9a), a current at this duty cycle can be between 0.73 and 0.83 s, the power is about zero. This
obtained. interval is used to achieve the current. According to Eqs.
( (7) and (6), we can calculate the equal power point that
V PV
I PV
¼ ð1  0:71Þ2  100 ¼ 8:41 is wanted to jump to. During the interval between 0.83
I PV ¼ 3:872  8  107 ðexpð0:82  V PV Þ  1Þ and 1.52 s, the most left peak is tracked. At this peak,
the stop condition is not satisfied. Then according to Eq.
I PV ¼ 2:2 ðAÞ
(8) and (6), the duty cycle is calculated to get the current
Replacing the current value of point P2 in the above at the next peak during the time between 1.52 and 1.62 s.
equation with this current value (block 18) into Eq. (6), During the interval between 1.62 and 1.72 s, the duty cycle
one can recalculate the duty cycle. is calculated again to get the more accurate current value.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Then, Eqs. (6) and (7) are used again to calculate a new
29:83
D¼1 ¼ 0:632 duty cycle to judge whether the equal power point of the
2:2  100 most left peak exists during 1.72–1.82 s. The second peak
With Eqs. (4) and (9b), the current I3 at this duty cycle has the equal power point. The algorithm will come back
can be obtained. to the ASVSS MPPT method during 1.82–2.12 s. When
( the maximum power at this peak is tracked, the stop con-
V PV 2
I PV
¼ ð1  0:632Þ  100 ¼ 13:54
dition is satisfied. The power of the three peaks is com-
I PV ¼ 1:972  8  107 ðexpð0:41  V PV Þ  1Þ pared and the duty cycle of the GMPP is output to
I PV ¼ 1:95 ðAÞ complete the GMPPT. The tracking time of 6.5 s is short-
ened. The other two simulation results (when the GMPP is
With Eq. (7) to calculate the voltage Vref1 at point c on the right and left of the three peaks) are similar and they
(block 20), then one can get the duty cycle at point c with are summarized in Table 1. Table 1 also shows the compar-
Eq. (6). ison of different tracking methods. From the results, the
V ref1 ¼ PI 32 ¼ 52:44 ¼ 26:89 ðVÞ tracking time of the proposed method is comparable and
1:95
qffiffiffiffiffiffiffiffiffiffiffiffi even the tracking time is less.
D ¼ 1  1:95100 26:89
¼ 0:63 Fig. 6(a) shows the P–V curve of the solar array with
four power peaks. The result without the equal power
After checking the equal power point, the algorithm method (Fig. 6(b)) is compared with that with the equal
enters the MPPT tracking to track point P3 (D = 0.597, power method (Fig. 6(c)). Without the equal power
Ip3 = 1.89 A, Vp3 = 31.03 V, Pp3 = 58.68 W). method, the tracking will take 3.13 s; however, it will take
Then the power at points P1, P2, and P3 is compared to 1.83 s to track with the equal power method. Therefore, if
achieve the GMPP. Therefore, the proposed algorithm is the situation of no equal power point exists, the proposed
verified with the help of the Mathcad. method can save the tracking time of the lower peak power
624 Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627

Table 1
Comparison of different tracking methods.
Situation Method
Linear function method Full-range searching Modified INC method Proposed method
(Ji et al., 2011) (Kazmi et al., 2009) (Tey and Mekhilef, 2014)
Power(time) Power(time) Power(time) Power(time)
Max power at left 52.44 W(2.64 s) 52.44 W(8.57 s) 52.16 W(3.87 s) 52.44 W(2.16 s)
Max power at middle FAULT 58.66 W(8.67 s) 58.58 W(4.06 s) 58.66 W(2.14 s)
Max power at right FAULT 112.64 W(6.19 s) 112.60 W(4.23 s) 112.64 W(2.87 s)
Need Voc & Isc in No need Voc & Isc in Need Voc in advance No need Voc & Isc in advance
advance advance

Rload decreased Rload increased

VPV decreased VPV increased

IPV increased IPV decreased

Fig. 7. The simulation results of the load variation. (100 X ? 50 X at 5 s


and 50 X ? 100 X at 7 s).

At 7 s, the load is changed from 50 X to 100 X. The pro-


posed algorithm can still adapt to the load variation. The
simulated systems with three and four PV panels in series
are shown in Figs. 5 and 6. However, the proposed method
can still be applied and extended even in the system with
more panels in series.

5. Experimental results

Fig. 6. The situation of the solar array with four peaks (a) curve of the The experimental system is like that shown in Fig. 8.
power-versus-voltage (b) without the equal power judgement (c) with the The component values of the boost converter are like those
equal power judgement. used in the simulations. The DSP TMS320LF2407 is used
to handle the digital signals and output the PWM signal to
the boost converter. The solar array is composed of MSX-
to speed up the tracking process. Fig. 7 shows the simula- 60s. The 18 solar cells in MSX-60 are in series, and then
tion of the load variation. The load is changed from 100 X they are paralleled with one bypass diode. Therefore, one
to 50 X at 5 s. The algorithm detects that the voltage is MSX-60 consists of two series sections and the open circuit
decreased and the current is increased. The voltage and voltage of one section is 10.55 V.
current of the solar array under variation are measured In the experiment, one real solar panel is divided into
and the new load value is calculated. The duty cycle of two sections, so two real panels create the effect of four
the GMPP after variation is then derived to make the sections in series. Two MSX-60s are used to become 4
system come back to the GMPP. sections; therefore, the experimental system is equivalent
Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627 625

10.55 L D
V

MSX-60
10.55
V

Cin S Co Load
10.55
V

MSX-60
10.55
V

Vpv Ipv

Driver
TMS320LF2407
s
Fig. 8. The test solar generation system.

Fig. 9. The situation of the GMPP being in the middle peak (a) measured
Fig. 10. The situation of the GMPP being in the right peak (a) measured
curve of the power versus voltage (PROVA) (b) experimental results with
curve of the power versus voltage (PROVA) (b) experimental results with
the proposed method.
the proposed method.

to four panels in series. However, the open-circuit voltage Fig. 9(a) gives the result of the GMPP being in the mid-
of the solar array is 42.2 V. To achieve three peaks in the dle peak with PROVA instrument. The voltage of the
four series solar sections, only two sections are needed to GMPP is 22.04 V, the current is 2.073 A and the power is
be shaded and the shaded degrees have to be different. 45.70 W. Fig. 9(b) shows the experimental results with the
The PROVA 210 solar test instrument is used to measure proposed method. The tracking process can be known by
the P–V curve of the solar array under the partially shaded observing the change of the voltage. The voltage is finally
condition. The experimental results will be compared with stable at 22 V. As for the measurement results of the
the results with PROVA to verify the accuracy of the pro- GMPP being in the right peak, they are shown in
posed method. The three situations that the GMPP is Fig. 10. Fig. 11 gives the related results of the GMPP being
located in the middle, right and left peaks will be in the left peak. Because the power of the middle peak is
investigated. lower than that of the left peak, the algorithm will
626 Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627

the tracking. This paper proposed the jumping method


according to the characteristic of each peak power point
of the solar array being about at n  0.8  Voc_module plus
the equal power method to reduce the tracking time greatly.
From the simulation results, the time taken by the pro-
posed method is less than that by the full-range searching
method. The results with and without using the equal
power method are also compared. The tracking perfor-
mance is better when the equal power method is used in
the GMPP tracking algorithm.
The experimental results validate the advantages of the
proposed method. Under the partially shaded conditions,
the proposed method will not fall into the local MPP and
the tracking time can be reduced. With the equal power
method, the tracking process can be skipped to shorten
the tracking time if the power of the following peak is
lower. The solar system with the proposed tracking method
is also verified to have a good performance even the load is
changed.

References

Boztepe, M., Guinjoan, F., Velasco-Quesada, G., Silvestre, S., Chouder,


A., Karatepe, E., 2014. Global MPPT scheme for photovoltaic string
Fig. 11. The situation of the GMPP being in the left peak (a) measured inverters based on restricted voltage window search algorithm. IEEE
curve of the power versus voltage (PROVA) (b) experimental results with Trans. Industr. Electron. 61, 3302–3312.
the proposed method. Chao, K.H., 2014. An extension theory-based maximum power tracker
using a particle swarm optimization algorithm. Energy Convers.
Manage. 86, 435–442.
Chen, K., Tian, S., Cheng, Y., Bai, L., 2014. An Improved MPPT
controller for photovoltaic system under partial shading condition.
IEEE Trans. Sust. Energy, 978–985.
Chen, Y.T., Lai, Z.H., Liang, R.H., 2014. A novel auto-scaling variable
step-size MPPT method for a PV system. Sol. Energy 102, 247–256.
Femia, N., Granozio, D., Petrone, G., Vitelli, M., 2007. Predictive &
adaptive MPPT Perturb and Observe method. IEEE Trans. Aerosp.
Electron. Syst. 43, 934–950.
Femia, N., Petrone, G., Spagnuolo, G., Vitelli, M., 2005. Optimization of
Perturb and Observe maximum power point tracking method. IEEE
Trans. Power Electron. 20, 963–973.
Ishaque, K., Salam, Z., 2013. A deterministic particle swarm optimization
maximum power point tracker for photovoltaic system under partial
Fig. 12. The experimental results of the load variation (100 X ? 50 X and shading condition. IEEE Trans. Industrial Electron. 60, 3195–3206.
50 X ? 100 X). Ishaque, K., Salam, Z., Amjad, M., Mekhilef, S., 2012. An Improved
Particle Swarm Optimization (PSO)–Based MPPT for PV With
investigate no equal power point, and the tracking of the Reduced Steady-State Oscillation. IEEE Trans. Power Electron. 27,
3627–3638.
middle peak will be skipped to shorten the tracking time. Ji, Y.H., Jung, D.Y., Kim, J.G., Kim, J.H., Lee, T.W., Won, C.Y., 2011.
Fig. 12 shows the experimental results of the solar system A Real Maximum Power Point Tracking Method for Mismatching
under the load variation. From the results, the system Compensation in PV Array Under Partially Shaded Conditions. IEEE
can be seen to come back to the GMPP even under the con- Trans. Power Electron. 26, 1001–1009.
dition of the load change. Jiang, Y., Abu Qahouq, J.A., 2012. Single-sensor multi-channel maximum
power point tracking controller for photovoltaic solar systems. IET
Power Electronics. 5, 1581–1592.
6. Conclusion Kazmi, S., Goto, H., Ichinokura, O., Guo, H.J., 2009. An improved and
very efficient MPPT controller for PV systems subjected to rapidly
The solar series array will produce many power peaks varying atmospheric conditions and partial shading. Power Engineer-
when it is under the partially shaded condition. The conven- ing Conference. AUPEC 2009. Australasian Universities. 27-30.
Springer, pp. 1–6.
tional MPP tracking method may fall into the local MPP, so Kobayashi, K., Takano, I., Sawada, Y., 2006. A study of a two stage
it cannot track the GMPP exactly. Although the full-range maximum power point tracking control of a photovoltaic system under
searching method has been proposed to improve the con- partially shaded insolation conditions. Sol. Energy Mater. Sol. Cells 90
ventional MPP method, it needs much time to complete (18/19), 2975–2988.
Y.-T. Chen et al. / Solar Energy 132 (2016) 617–627 627

Lian, K.L., Jhang, J.H., Tian, I.S., 2014. A Maximum Power Point Piegari, L., Rizzo, R., 2010. Adaptive perturb and observe algorithm for
Tracking Method Based on Perturb-and-Observe Combined With photovoltaic maximum power point tracking. IET Renew. Power
Particle Swarm Optimization. Photovoltaics, IEEE Journal. 4, 626– Gener. 4, 317–328.
633. Safari, A., Mekhilef, S., 2010. Simulation and hardware implementation
Liu, Y.H., Huang, S.C., Huang, J.W., Liang, W.C., 2012. A Particle of incremental conductance MPPT with direct control method using
Swarm Optimization-Based Maximum Power Point Tracking Algo- Cuk converter. IEEE Trans. Ind. Electron. 58, 1154–1161.
rithm for PV Systems Operating Under Partially Shaded Conditions. Tey, K.S., Mekhilef, S., 2014. Modified incremental conductance algo-
IEEE Trans. Energy Conversion. 27, 1027–1035. rithm for photovoltaic system under partial shading conditions and
Miyatake, M., Veerachary, M., Toriumi, F., Fujii, N., Ko, H., 2011. load variation. IEEE Trans. Ind. Electron. 61, 5384–5392.
Maximum Power Point Tracking of Multiple Photovoltaic Arrays: A Veerasamy, B., Kitagawa, W., Takeshita, T., 2014. MPPT method for PV
PSO Approach. IEEE Trans. Aerospace and Electronic Systems. 47, modules using current control-based partial shading detection. Inter-
367–380. national Conference on Renewable Energy Research and Application
MSX-60 and MSX-64 Photovoltaic Modules datasheet, 1997. SOLAREX. (ICRERA), pp. 359–364.
Nasir, M., Zia, M.F., 2014. Global maximum power point tracking Xiao, W., Dunford, W.G., 2004. A modified adaptive Hill Climbing
algorithm for photovoltaic systems under partial shading conditions. MPPT method for photovoltaic systems. 35th Annual IEEE Power
The 16th International Power Electronics and Motion Control Electronics Specialists Conference, pp. 1957–1963.
Conference and Exposition (PEMC), pp. 667–672. Xuesong, Z., Daichun, S., Youjie, M., Deshu, C., 2010. The simulation
Patel, H., Agarwal, V., 2008. Maximum Power Point Tracking Scheme for and design for MPPT of PV system based on incremental conductance
PV Systems Operating Under Partially Shaded Conditions. IEEE method. Information Engineering (ICIE). 2010 WASE International
Trans. Industrial Electronics. 55, 1689–1698. Conference on. 14–15, pp. 314–317.

You might also like