IEEE PEDS 2011, Singapore, 5 - 8 December 2011

Gradient Approximation Based Maximum Power

Point Tracking for PV Grid-Connected System
Ahmed. G. Abo-Khalil1, Tarek Ahmed2
Electrical Engineering Department, Faculty of Engineering
Assiut University, Egypt,
1
E-mail: a_galal@aun.edu.eg

Abstract-- The photovoltaic (PV) generators have a

nonlinear V-I characteristics and maximum power points
which vary with the illumination level and temperature.
Using maximum power point tracker (MPPT) with the
intermediate converter can increase the system efficiency by
matching the PV systems to the loads. This paper presents a
maximum power point tracker based on Gradient
Approximation (GA) and a control scheme for and a control
scheme for a single-phase inverter connected to the utility
grid. GA provides an adaptive nature for the system
performance. Also the GA provides excellent features such as
fast response, good performance and the calculation of the
maximum power point condition is independent of the PV
characteristics. The single-phase dc/dc converter is designed
to control the power flow to the maximum value. The
simulation results have verified the validity of the proposed
estimation algorithm.
I.

INTRODUCTION

Photovoltaic generation is being increasingly

important as a renewable source since it offers many
advantages such as incurring no fuel costs, no pollution,
little maintenance, and emitting no noise among other
alternatives. The PV modules still have relatively low
conversion efficiency. Therefore, the control of maximum
power point tracking for the solar array is essential in a PV
system.
The amount of power generated by a PV depends on
the operating voltage of the array. The maximum power
operating point changes with insolation level and
temperature. Its V-I and V-P characteristic curves specify
a unique operating point at which the maximum power is
generated. At the MPP, the PV system operates at its
highest efficiency [1]. Therefore, the tracking control of
the MPP is a complicated process. To overcome this
problem, different tracking control strategies have been
proposed, which are the perturbation and observation
P&O [2], the incremental conductance [3], the parasitic
capacitance [4], the constant voltage [5], the neural
network [6], and the fuzzy logic control [7]. In the
literature [8], grid-connected PV inverters have been
proposed, but they merely deliver the real power to the
utility and fixed loads. Recently, researchers made efforts
in developing parallel PV inverters which can deliver the
real power and reduce harmonic currents caused by
nonlinear loads [9].

Generally, there are several methods which are

commonly used to determine the maximum power point.
With P&O algorithm the operating voltage V is perturbed
with every MPPT cycle. As soon as the MPP is reached, V
oscillates around the ideal operating voltage Vmpp . This
causes a power loss which depends on the step width of a
single perturbation. If the step width is large, the MPPT
algorithm will respond quickly to sudden changes in the
operating conditions with increasing losses under stable or
slowly changing conditions. If the step width is very small
the losses under stable or slowly changing conditions will
be reduced, but the system will only able to respond very
slowly to rapid changes in the temperature or illumination
levels. The value of the ideal step width is systemdependent and needs to be determined experimentally.
Incremental conductance algorithm has an advantage over
P&O method in that it can determine when the MPPT
reaches the MPP, while the output power in P&O method
oscillates around the MPP. Also, incremental conductance
can track rapidly changing irradiance conditions with
higher accuracy than P&O method.
In this paper, a novel maximum power point tracker of
photovoltaic systems is proposed, which is based on the
Gradient Approximation theory. The effectiveness of the
proposed algorithm has been verified by simulation results
and shows that the estimation error is less than 1 %.
II.

MATHEMATICAL MODEL

Solar cells are devices that convert photons into

electrical potential in a PN junction, of which equivalent
circuit is shown in Fig. 1. Due to the complex physical
phenomena inside the solar cell, manufacturers usually
present a family of operating curves (V-I) as shown in Fig.
2. These characteristics are obtained by measuring the
array volt-ampere for a different illumination values. From
these characteristics, the optimum voltage or current,
corresponding to the maximum power point, can be
determined. It is clearly seen in Fig. 2 that the current
increases as the irradiance levels increase. The maximum
power point increases with a steep positive slope
proportional to the illumination.
The main parameters which influence the illumination
levels on a surface at a fixed tilt on earth are the daily and

978-1-4577-0001-9/11/$26.00 2011 IEEE

104

seasonal solar path, the presence of clouds, mist, smog and

dust between the surface and the sunlight, and the shade of
the object positioned such that the illumination level is
reduced, etc.
The equation of the PV output current I is
expressed as a function of the array voltage V

I = I sc - I o {e

q (V + IRs )
KTk

- 1} - (V + IRs )/Rsh

(1)

where V and I represent the PV output voltage and

current, respectively; R s and R sh are the series and
shunt resistance of the cell (in Fig. 1); q is the electronic
charge; Isc is the light-generated current; Io is the reverse
Saturation current; K is the Boltzman constant, and Tk
is the temperature in K .
Equation (1) can be written in another form as [10]

I =I sc {1 K 1 [e K 2V 1]} - (V + IRs )/Rsh

m

where the coefficient K1 ,

Fig. 2 V-I and P-I characteristics at constant temperature

( 25 o C )

(2)

K 2 and m are defined as

K1 = 0.01175,
K 2 = K 4 /(Voc ) m ,

Fig. 3. Power circuit for PV system.

K 4 = ln(( K1 + 1) / K1 ) ,

K 3 =ln[( I sc (1 +K1 ) I mpp ) / K1I sc ],

m = ln( K 3 / K 4 ) / ln(Vmpp / Voc )
and Impp is the current at maximum output power,
the voltage at maximum power,
current and

Vmpp is

I sc is the short circuit

Voc is the open circuit voltage of the array.

Equation (2) is only applicable at one particular

operating condition of illumination G and cell
temperature Tc .The parameter variations can be calculated
by measuring the variation of the short-circuit current and
the open-circuit voltage in these conditions using the
parameters at the normal illumination and cell temperature.
Equation (2) is used for the I-V and P-V characteristics for
various illumination and fixed temperature ( 25 [ o C ] ) in
Fig. 2.
III.

SYSTEM DESCRIPTION AND CONTROL

A conventional two-stage energy conversion system

is connected between the PV array and the electrical
power system as shown in Fig. 3. A boost converter is
used to increase the PV voltage for the inverter circuit and
it also plays a role of an intermediate circuit for tracking
the maximum power point.
The inverter circuit converts the direct current to the
alternating current which flows into the utility or local
loads. The inverter controller has two main functions. One
is to synchronize the output current with the grid voltage,

Fig. 4. Control circuit for the PV system.

which means the power factor is equal to unity. The other

is that it controls the dc link voltage. To achieve these two
goals, the inner current control loop and the outer voltage
control loop are used as shown in Fig 4.
The main advantages of the boost converter are higher
efficiency & reduced component count and it converts the
unregulated voltage into desired regulated voltage by
varying the duty cycle at high switching frequency
lowering the size and cost of energy storage components.
When a boost converter operates in continuous mode,
the current through the inductor (IL) never falls to zero.
The output voltage can be calculated as follows, in the
case of an ideal converter (i.e using components with an
ideal behavior) operating in steady conditions. During the
On-state, the switch S is closed, which makes the input
voltage appear across the inductor, which causes a change
in current (IL) flowing through the inductor during a time
period (t) as shown in Fig. 5(a). During the Off-state, the
switch S is open, so the inductor current flows through the
load as shown in Fig. 5(b).
Applying Kirchhoffs rules around the loops and
rearranging terms yields an intuitive result:

105

If DPpv < 0 ,

then D * =D d

(5)

If DPpv > 0 ,

then D = D + d

(6)

If DPpv 0 ,

Fig. 5. Equivalent circuit of the DC/DC boost converter during

switching.

then D = D
*

(7)

The disadvantage of this method includes empirical setting

for d . To overcome this disadvantage, an adaptive duty
ratio is introduced by the gradient approximation (GA)
method proposes. It doesnt require parameter setting,
sensors, and characteristics of the photovoltaic.
GA is a stochastic optimization method, which will be
used for extracting maximum power from the PV systems
in this paper. GA involves the process for attaining the
roots of nonlinear equation and will be explained in this
section.
One may use Newtons method to solve a nonlinear
equation f ( x) = 0 :

xk +1 =xk f ' ( xk ) 1 f ( xk )
1
Vo
=
Vi 1 D

(3)

From the above expression it can be seen that the output

voltage is always higher than the input voltage (as the duty
cycle goes from 0 to 1), and that it increases with D,
theoretically to infinity as D approaches 1.
If we considered the operation of the boost converter only
during the on-off states we can estimate the converter
current without a need to a current sensor. If D is set to
zero for time longer than the switching frequency -which
is the open circuit condition- the output voltage in this
case is equal to the input voltage and the inductor current
is zero. On the other hand if the duty ratio is set to 1
which is the short circuit condition- the output voltage will
be zero.

(8)

where k is the iteration index. Alternatively, it was

found that many of the optimization problems are
formulated by Lagrangian, denoted by L, for solving the
unknown h. The necessary condition for obtaining the
optimality is as follows:

g ( )

L( )
=0

(9)

Solving g ( ) = 0 in Eq. (9) is the same as solving

f ( x) = 0 .
In GA, the variables are updated by Eq. (9), which is
similar to Eq. (8) based on Newtons method.

k +1 =k ak g k ()

(10)

where

k +1

vector of the estimated unknowns in the (k+1)-th

iteration.
IV.

GRADIENT APPROXIMATION

In the case of a PV array connected to a power

converter, perturbing the duty ratio of power converter
perturbs the PV array current and consequently perturbs
the PV array voltage. Hill-climbing and P&O methods are
different ways to envision the same fundamental method.
Incrementing (decrementing) the voltage increases
(decreases) the power when operating on the left of the
MPP and decreases (increases) the power when on the
right of the MPP.
Therefore, if there is an increase in power, the
subsequent perturbation should be kept the same to reach
the MPP and if there is a decrease in power, the
perturbation should be reversed. The process is repeated
periodically until the MPP is reached.
The PV power variation is defined as [11]:
(4)
DPpv =Ppv (n) Ppv (n 1)
where Ppv (n) and Ppv (n 1) are the power output at
time n and n -1, respectively. The Dc-DC converter
duty-ratio is determined according to the following rules:

vector of the estimated unknowns in the k-th

iteration.
ak
a multiplier for decreasing updating in the kth
iteration.
g k
estimated gradient of L in the k-th iteration.
The variable

ak in (10) plays the role of f ' ( xk ) 1 in

(8).
The simultaneous perturbation stochastic process is used
to avoid local optimums. Furthermore,
y k( +) y k( )

2c k k 1

g k =
( +)

()
yk yk
2c

k kp

(11)

where

ck

{ ck }is a sequence of positive scalars where

106

Fig. 7. Photovoltaic MPPT, irradiation, current and voltage.

Fig. 8. Photovoltaic MPPT, irradiation, power and DC-link voltage.

Fig. 6. Flowchart for MPPT algorithm.

ck approaches zero.
k

k R p is a vector of p mutually independent

mean-zero random variables

The

{ k1 , k 2 , ...... kp }

yk( ) in Eq. (11) is calculated by Eqs. (12) and (13)

y k( + ) =L( +ck k ) + k( + )
y k( ) =L( ck k ) + k( )

and

(12)

(13)

E k( +) k( )0 , 1 ,..., k , k = 0 for k ,

(14)

where { ck } is a sequence of positive scalars such that

ck

approaches zero. The symbols

This duty cycle will be sent to the DC/DC converter

through Pulse Width Modulation controller, as shown in
Fig. 4.
Figure 6 shows the flowchart of the proposed method
for maximum power point tracking, which is described as:

k(+ ) and k( ) represent

measurement noise terms.

For the MPPT problem, the Lagrangian may be related to
the control signal for tuning the duty cycle of the boost
converter. GA is especially useful when the Lagrangian
can be estimated with two measurements or two estimated
values. The MPPT problem fulfills this main feature of
GA.
Because Ppv (n) is expected to be maximized and GA is
used for minimization, the lagrangian is Ppv (n) in GA.
This control variable is the duty cycle of the DC/DC
converter and is adapted as Eq. (15) in kth iteration at time
n:
(15)
Dk +1 Dk + a k c k

Step 1 : Initialization and selection of the initial

guess and coefficients a, c, A, , and
Practically and theoretically valid values for ,
and are 0.602 and 0.101 respectively.
a
c
ak =
, ck =

( A + k + 1)
( k + 1)
Step 2 : Generation of simultaneous perturbation
vector by Monte Carlo a random perturbation
vector.
Step 3 : Loss function measurements

L(k

c+k
k ), L ( k ck k )

Step 4 : Gradient Approximation as in Eq. 10

Step 5 : Update

Step 6 : Iteration and Termination.

V.

k to a new value k +1

SIMULATION RESULTS

Solar illumination varies quickly with time. It means

that the maximum power point moves to another curve
quickly. Such a case needs a MPPT controller capable of
reaching the MPP as quickly as possible in order to reduce
the output power oscillations and the system power loss.
Comprehensive simulation studies were made to
investigate not only the performance of the boost
converter controlled by GA algorithm but also the inverter

107

VI.

CONCLUSIONS

This paper presented two-stage energy conversion

system to connect the PV array to the grid utility and local
loads. Gradient Approximation method is used to extract
the maximum power point tracker by controlling the boost
converter in order to adjust the duty ratio to maintain the
maximum power condition. It also presented the inverter
control scheme to deliver the converter output power to
the utility grid at unity power. The simulation results have
shown that the advantage of this system is fast response,
good transient performance, and high accuracy.
REFERENCES
Fig. 9 Dc-link control.

[1]
[2]
[3]
[4]

[5]
[6]
Fig. 10 Instantaneous waveforms of grid voltage and current

[7]

[8]

[9]
[10]
Fig. 11 Frequency spectrum of grid current.

controller performance of connecting the PV system to the

grid utility at a unity power factor. The PV array is
simulated using (1) and (2) as described in section II. The
GA MPPT performance over a wide range irradiation is
shown in Fig. 7 and 8. The PV output voltage, current, and
power change according to the illumination variation. It is
noticed that the fast dynamic performance at step variation
of illumination level is obtained. It is obvious that the GA
estimation for the irradiation level is very high that lead to
high accuracy maximum power point tracking. The DC
link voltage is well controlled as shown in Fig. 9. The
unity power factor control is also well performed as shown
in Fig. 10. The current spectrum in Fig. 11 shows that the
grid current includes very small components of high
frequency harmonics.

[11]

M. Bodur and M. Ermis, Maximum Power Point Tracking for

Low Power Photovoltaic Solar Panels, IEEE Electrotechnical
Conference Proc., vol. 2, pp. 758-761, 1994.
H. Sugimoto and H. Dong, A New Scheme For Maximum
Photovoltaic Power Tracking Control, IEEE Power Conversion
Conference Proc., vol. 2, pp. 691-696, 1997.
C. Won, et all, A New Maximum Power Point Tracker of
Photovoltaic Arrays using Fuzzy Controller, IEEE PESC Proc.,
vol. 1, pp. 396403, 1994.
K. H. Hussein, I. Muta, T. Hoshino, and M. Osakada, Maximum
Photovoltaic Power Tracking: an Algorithm for Rapidly Changing
Atmospheric Conditions, IEE Proc. on Generation,
Transmission, and Distribution, vol. 142, no. 1, pp. 59-64, 1995.
Y. Hsiao and C. Chen, Maximum Power Tracking for
Photovoltaic Power System, IEEE IAS Conf. Proc, vol. 2, pp.
10351040, 2002.
T Senjyu and K. Uezato, Maximum Power Point Tracker Using
Fuzzy Control for Photovoltaic Arrays, IEEE ICIT proc., pp. 143
147, 1994.
A. Al-Amoudi and L. Zhang, Application of Radial Basis
Function Networks for Solar-Array Modeling and Maximum
Power-Point Prediction, IEE Proc. on Generation, Transmission
and Distribution, vol. 147, no. 5, pp. 310316, 2000.
D. M. Baker, V. G. Agelidis, and C. V. Nayar, A New Zero
Average Current Error Control Algorithm for Inverter,
Australian Universities Power Engineering Conference (AUPEC)
proc., vol. 1, pp. 67-72, 1997.
M. Armstrong, et al, Low Order Harmonic Cancellation Scheme
for Multiple PV Grid-Connected Inverters, EPE proc., vol. 1, pp.
1-10, 2003.
D. C. Lee, et. al, Maximum Power Point Tracking for
Photovoltaic System Using Fuzzy Logic Controller, KIPE Power
Electronics Annual Conference Proc., vol. 2, pp. 503-507, 2003.
Y. Y. Hong, S. D. Lu and C. S. Chiou, MPPT for PM wind
generator using gradient approximation, Elsevier, Energy
Conversion and Management, pp. 82-89, 2009.

108