I. INTRODUCTION
HE MAIN objective of the maximum power point (MPP)
tracking (MPPT) algorithms is to achieve fast and accurate
tracking performance and minimize oscillations due to varying
weather conditions. A comparative study on MPPT techniques
for photovoltaic (PV) power systems is reported in [1] and [2].
Among different MPPT algorithms, much focus has been on
perturb and observe (P&O) [3], hill climbing [4], and incremental
conductance (INC) methods [5]. In [3], a survey of P&O techniques has been presented. It has been shown that the existing
techniques suffer from oscillations, complexity, designer dependency, and more computational effort. In the P&O method, the
operating point oscillates around the MPP giving rise to the
wastage of some amount of available energy. These oscillations
can be minimized by reducing the xed perturbation size, but it
takes relatively more time to reach MPP. The solution to this
conicting situation is to have a variable step size as suggested in
[6]. Although the implementation of these methods are simple, it
is not very accurate and rapid, since the effects of temperature and
irradiation are not taken into consideration. Patel and Agarwal [7]
proposed a variable perturb by adopting four power ranges. In
each range, a specic perturb value is used; hence, the method is
not fully adaptive. Several methods are proposed to address these
issues by considering adaptive perturbation [8], [9].
An INC method is given in [5], [10], and [11], which is based
on the fact that the slope of the PV array power curve is zero at
MPP, negative on the right, and positive on the left of the MPP.
The INC method inherits the same problems as P&O, namely the
tradeoff between the speed and oscillations. In [4], the author
claims that the INC method is prone to failure in case of large
changes in irradiance.
In another method, the MPP current IMPP (or MPP voltage
VMPP ) is continuously monitored with respect to the short circuit
current ISC (or open circuit voltage VOC ) [12][14]. This method
is well known as the fractional short circuit current [(FSCC) or
fractional open circuit voltage] method. Since this method
approximates a constant ratio, its accuracy cannot be guaranteed
under varying weather conditions.
To overcome the above-mentioned drawbacks, several methods have been proposed using articial intelligence (AI)-based
algorithms such as neural network (NN) [15] and fuzzy logic
controller (FLC) [16]. But these methods also have drawbacks
such as the requirement of large data storage and extensive
computation. For instance, NN requires a large amount of data
for training, which is the major constraint. Similarly, FLC
requires extensive computation to deal various stages. Since the
nonlinear characteristics of the solar module should be well
ascertained to create the control rules, the versatility of these
methods is limited. Furthermore, low-cost hardware processors
cannot be used for these applications because the MPP continuously changes with atmospheric conditions in real time.
In this paper, a variable perturbation size adaptive P&O MPPT
algorithm is proposed in order to overcome the drawbacks in the
conventional P&O method. The proposed method consists of
three algorithms namely, current perturbation, adaptive control,
and variable perturbation. The adaptive control algorithm
(ACA) moves the operating point closer to the MPP by multiplying the short circuit current with optimal proportionality
constant. The ACA gets activated only if there is sudden change
in irradiance or sudden change in PV current. In the proposed
method, the short circuit current is estimated instead of measuring, which reduces the losses and saves the additional component
cost required. Further, the proposed algorithm features the tuning
of variable perturbation size in two stages, namely coarse and
ne tuning. In the coarse tuning, the perturbation size is determined based on irradiance level. In the ne tuning, the
1949-3029 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
KOLLIMALLA AND MISHRA: VARIABLE PERTURBATION SIZE ADAPTIVE P&O MPPT ALGORITHM
719
720
(2)
ISC KT
S
Snom
(6)
ksc
opt
opt
low ISC :
in (7) gives
1 ksc
2
low
(8)
opt ISC
1 ksc
2
low
mS:
KT 1 ISC T Tnom
(4)
S
Snom
; Tnom KA
.
where m ISC SnomSnom
(5)
(9)
(10)
Similarly, from (2), the lower operating current limit for a given
irradiance S is dened as
Ilow 2IMPP ISC :
where
(7)
opt
(11)
in (11) gives
low mS:
(12)
(13)
(14)
(15)
(16)
KOLLIMALLA AND MISHRA: VARIABLE PERTURBATION SIZE ADAPTIVE P&O MPPT ALGORITHM
721
Solving (16) using (5), (6), and ksc ksc opt gives S1
1
2ksc opt 1 S0 . Therefore, the incremental limit of S is dened as
SINC S1 S0
21 ksc opt
S0 :
2ksc opt 1
(17)
(18)
Solving (18) using (5), (6), and ksc ksc opt gives S2
2ksc opt 1S0 . Therefore, the decremental limit of S is
dened as
SDEC S2 S0 2ksc
opt
1S0 :
(19)
DYNAMICS
TABLE I
PERTURBATION SIZE
OF
(20)
low mS:
(21)
IC
low
mS:
(22)
From the above equation, it is clear that IC varies linearly with
irradiance. This value sets the coarse tuning of variable
perturbation size.
C. Variable Perturbation Algorithm
The ne perturbation size (IF ) is dened as
jMj
IF m1 m2
Ik
2
(23)
(24)
(25)
(26)
where
(27)
OF
TheproposedMPPTalgorithmisrealizedusingaboostconverter.
The PV panel is connected to the boost converter as shown in Fig. 4.
722
TABLE II
NOMINAL PARAMETERS OF BOOST CONVERTER
It is assumed that the boost converter is operating in continuous current mode (CCM). According to the state space averaging
method [17], the system dynamics are described by the following
equations:
dipv
1 d
vpv
vo
L
dt
L
dvo 1 d
1
ipv
vo
C
RC
dt
(28)
(29)
where ipv , vpv , and vo are input current, input voltage, and output
voltage of boost converter, respectively, and d is duty ratio.
A. Design of Current Control Loop
Fig. 4 shows that vpv and ipv are given to the MPPT controller
that generates the reference inductor current Iref . This reference
current is given to the current control loop. The transfer function
of control to inductor current is given as [18]
Gid s
Vo Cs 2 VRo
i~pv
d~
LCs2 RL s 1 D2
(30)
where v~pv ; i~pv ; v~o , and d~ are small perturbations in vpv ; ipv ; vo ,
and d, respectively.
The transfer function of proportional integral (PI) controller
is given by
Gc s Kp
Ki
:
s
(31)
(32)
Fig. 5. (a) Bode plot and (b) root locus of current control loop.
TABLE III
SPECIFICATIONS OF PV PANEL AT STC
KOLLIMALLA AND MISHRA: VARIABLE PERTURBATION SIZE ADAPTIVE P&O MPPT ALGORITHM
723
Fig. 6. Comparison of proposed algorithm with (a) conventional P&O, (b) adaptive P&O, and (c) variable perturbation size.
724
Fig. 9. Experimental results for m1 0:95, 0.9, and 0.85 at N 2: (a) output
voltage, PV power, voltage, and current; (b) powers; and (c) variable perturbation
size.
KOLLIMALLA AND MISHRA: VARIABLE PERTURBATION SIZE ADAPTIVE P&O MPPT ALGORITHM
725
726
As the operating point oscillates around MPP, VPA gets activated and continuously reduces the I using (23). The CPA
drives the operating point to MPP in less than 60 ms, as shown in
Fig. 11(c).
At t2 instant, the irradiance is suddenly decreased from 535 to
410 W/m2 . According to (19), SDEC 117 W/m2 . But the
irradiance is decreased by 125 410 535 W/m2 , which is
less than SDEC . Therefore, ACA gets activated and sets the Iref
to 1.678 A and I to 0.0415 A using (9) and (22), respectively.
The CPA drives the operating point to MPP in less than 40 ms, as
shown in Fig. 11(d).
At t3 instant, the irradiance is suddenly increased from 412
to 537 W/m2 . According to (17), SINC 116 W/m2 . But
the irradiance is increased by 125 537 412 W/m2 ,
which is greater than SINC . Therefore, ACA gets activated
and sets the Iref to 2.198 A and I to 0.0543 A using (9) and
(22), respectively. The CPA drives the operating point to MPP
in less than 70 ms as shown in Fig. 11(e).
At t4 instant, the boost converter is brought back to inactive
mode. Therefore, the PV panel is directly connected to the load.
At t5 instant, algorithm 1 is activated with VP&O 0:1 V. At
t6 instant, the irradiance is suddenly decreased to 418 W/m2 . At
t7 instant, the irradiance is suddenly increased to 515 W/m2 .
The results are shown in Fig. 11(b).
From Fig. 11(c)(e), it is observed that the steady-state
response time to reach MPP (tss ) for proposed algorithm is
60, 40, and 70 ms, respectively, whereas for algorithm 1 is 350,
180, and 160 ms, respectively.
B. Comparison of Proposed Algorithm With Current Based
P&O Algorithm (Algorithm 2) [21]
Initially, the boost converter is operated in inactive mode.
At t1 instant, the proposed algorithm is activated. At this
instant, vpv , ipv , S, and Ilow are 25 V, 1 A, 525 W=m2 , and
1.88 A, respectively. As ipv is less than Ilow , ACA gets
activated and sets Iref to 2.14 A and I to 0.0531 A. The
current perturbation algorithm takes these values as inputs
and drives the operating point to MPP, as shown in
Fig. 12(b).
At t2 instant, the irradiance is suddenly decreased from 525 to
374 W/m2 . According to (19), SDEC 115 W/m2 . But the
irradiance is decreased by 151 374 525 W/m2 , which is
less than SDEC . Therefore, ACA gets activated and sets the Iref
to 1.531 A and I to 0.0378 A.
At t3 instant, the irradiance is suddenly increased from 374 to
523 W/m2 . According to (17), SINC 105 W/m2 . But the
irradiance is increased by 149 523 374 W/m2 , which is
greater than SINC . Therefore, ACA gets activated and sets the
Iref to 2.141 A and I to 0.0529 A.
At t4 instant, the boost converter is brought back to inactive
mode. At t5 instant, algorithm 2 is activated with IP&O
0:01 A. At t6 instant, the irradiance is suddenly decreased to
375 W/m2 . At t7 instant, the irradiance is suddenly increased
to 535 W/m2 .
From Fig. 12(c)(e), it is observed that tss for proposed
algorithm is 80, 60, and 70 ms, respectively, whereas for algorithm 2 is 750, 190, and 340 ms, respectively.
Fig. 12. Experimental results of proposed algorithm and algorithm 2: (a) irradiance; (b) output voltage, PV power, voltage, and current; and (c)(e) powers.
KOLLIMALLA AND MISHRA: VARIABLE PERTURBATION SIZE ADAPTIVE P&O MPPT ALGORITHM
727
Fig. 13. Experimental results of proposed algorithm and algorithm 3: (a) irradiance; (b) output voltage, PV power, voltage, and current; and (c)(e) powers.
728
Mahesh Kumar Mishra (S00M02SM10) received the B.Tech. degree from the College of Technology, Pantnagar, Uttarakhand, India, in 1991,
the M.E. degree from the University of Roorkee,
Roorkee, Uttarakhand, India, in 1993, and the Ph.
D. degree in electrical engineering from the Indian
Institute of Technology Kanpur, Kanpur, Uttar
Pradesh, India, in 2002.
He has teaching and research experience of about
20 years. For about 10 years, he was with the Electrical
Engineering Department, Visvesvaraya National Institute of Technology, Nagpur, India. Currently, he is a Professor with the
Electrical Engineering Department, Indian Institute of Technology Madras,
Chennai, Tamil Nadu, India. His interests are in the areas of power distribution
systems, power electronic applications in microgrid, and renewable energy
systems.
Dr. Mahesh is a life member of the Indian Society of Technical Education
(ISTE).