Sunteți pe pagina 1din 9

644 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO.

2, APRIL 2015

An Optimal Maximum Power Point Tracking


Algorithm for PV Systems With Climatic
Parameters Estimation
Ali Chikh, Member, IEEE, and Ambrish Chandra, Fellow, IEEE

Abstract—This paper presents a maximum power point track- depending on their complexity, sensors used, convergence,
ing (MPPT) method for photovoltaic (PV) systems with reduced setup, and in further aspects [2]–[8], [22]–[25].
hardware setup. It is realized by calculating the instantaneous The current–voltage (I−V ) and the power–voltage (P −V )
conductance and the junction conductance of the array. The first
one is done using the array voltage and current, whereas the curves of the array are shown in Figs. 1 and 2, respectively,
second one, which is a function of the array junction current, under uniform solar radiations [3].
is estimated using an adaptive neuro-fuzzy (ANFIS) solar cell It is noticed from Fig. 2 that the array maximum power
model. Knowing the difficulties of measuring solar radiation and varies with the solar radiation. Therefore, it is necessary to
cell temperature, since those require two extra sensors that will track constantly these points. The most common methods are
increase the hardware circuitry and measurement noise, an ana-
lytical model is proposed to estimate them with a denoising-based the perturb and observe (P&O), and the incremental conduc-
wavelet algorithm. The proposed MPPT technique helps to reduce tance (InCond). The first method is popular due to its hardware
the hardware setup using only one voltage sensor, while increases simplicity. It functions by creating a perturbation in the volt-
the array power efficiency and MPPT response time. The simula- age, which results in variation in the array power [3], [20].
tion and experimental results are provided to validate the MPPT However, the system oscillates around the MP points, which
algorithm operation as well as the climatic parameters estimation
capabilities. wastes energy while it cannot distinguish between the radiation
variation and the tracker’s perturbation. Many solutions are pre-
Index Terms—Adaptive neuro-fuzzy (ANFIS) solar cell model,
sented based on an adaptive step size or optimization functions
instantaneous conductance, junction conductance, junction cur-
rent, maximum power point tracking (MPPT), photovoltaic (PV) [3]–[5], [9], [21], [26]. The second InCond method has a great
system. accuracy with good flexibility to rapidly varying climatic condi-
tions [24]. Therefore, it is possible to recognize when the MPP
I. I NTRODUCTION is achieved within a given accuracy [7], [22] by computing the
instantaneous and InConds and verifying if they are identical.
N OWADAYS, environmental problems produced by the
use of classical power generation seem to acquire more
attention around the world [1]. The extensive use of these
However, it involves an increased hardware and software com-
plexity. Many improvements are proposed using variable and
types of energies increases continuously the earth temperature modified step size algorithms [8], [22]–[24]. A sliding mode
and warming. Nowadays, renewable energies, especially pho- control has been used to improve tracking accuracy [11] and
tovoltaic (PV) energy, considered as palliative energy resource, InCond using reference voltage and direct duty ratio perturba-
are growing rapidly and are used in several applications, tion [12]. Finally, a model-based (MB) control is used [13].
e.g., battery charging, home energy supply, and pumping sys- Nowadays, intelligent systems are progressively used due to
tems. They have long life, require minimum maintenance, and their human-like capability and ability to adapt and improve
produce no noise or disturbing effects. Therefore, they are their performance [14], [19], [26], [27]. Many neural networks
environmental-friendly. However, they suffer from a relatively (NN)- and fuzzy logic (FL)-based techniques are proposed. An
low conversion efficiency, which makes their optimization artificial NN algorithm is developed with climatic conditions as
necessary. This is done by extracting their maximum power inputs and applied to a boost tracker [28]. Using FL theory, a
for fluctuating climatic environments. Well-known “maximum feed-forward MPPT scheme, a fuzzy controller, and an adaptive
power point tracking (MPPT)” is done through a particular P&O-FL control method are proposed [9], [15], [16], [30].
control of a converter. Various techniques have been proposed The combination of NN and FL seems to be more attractive
since it associates the learning capabilities of the ANN with the
Manuscript received July 23, 2014; revised December 03, 2014; accepted ability of the FL to trait inaccurate data, which makes it suitable
February 07, 2015. Date of publication March 09, 2015; date of current version
March 18, 2015. Paper no. TSTE-00368-2014. for PV applications [17]. A radial-basis NN function (RBFNN)
The authors are with the Department of Electrical Engineering, Ècole de is presented to deliver the reference information to a recurrent
Technologie Supèrieure, Montreal, QC H3C 1K3 Canada (e-mail: achikh@ FNN controller (RFNNC) [31]. Another method is proposed
yahoo.com; ambrish.chandra@etsmtl.ca).
based on a generalized dynamic FNN (GD-FNN) [32]. Finally,
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. adaptive MPPT controllers-based neuro-fuzzy inference system
Digital Object Identifier 10.1109/TSTE.2015.2403845 (ANFIS) is presented [33], [34].

1949-3029 © 2015 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.
CHIKH AND CHANDRA: OPTIMAL MPPT ALGORITHM FOR PV SYSTEMS 645

Fig. 3. Equivalent circuit of a solar cell.

Fig. 1. Array current–voltage curve.

Fig. 4. ANFIS decoupled solar cell model.

The output current of the PV cell is given by [6], [17],


[21], [24]
   
V
I = Iph − Id = Iph − Io exp −1 (1)
Vt

where Iph is the photocurrent, Id is the junction current, Io is


the reverse saturation current, and Vt is the thermal voltage.
Fig. 2. Array power–voltage curve. The photocurrent is a linear function of (G, Tc ) [1], [6],
[17], [21]
It can be noticed from the literature that all the proposed G
Iph = [Isc,ref + ki (Tc − Tc,ref )] = f (G, Tc ) (2)
MPPT methods need information on the array voltage, current, Gref
and environmental parameters, in some cases, which is a diffi-
cult task resulting in an increased and complex hardware with where Isc,ref is the cell short-circuit current in the standard test
high failure probability and noise measurements. conditions (STC), G is the solar radiation, Gref = 1000 W/m2
This paper provides an efficient MPPT method with reduced is the solar radiation in the STC, Tc is the cell temperature,
hardware. It computes the instantaneous and junction array con- Tc,ref = 25◦ C is the cell temperature in the STC, and ki is the
ductances. The first one is done using the array current and coefficient of the short-circuit current.
voltage, whereas the second one uses the array junction cur- The other current involved in (1) is the junction current Id ,
rent, which is estimated using ANFIS cell model presented in which is highly nonlinear [1], [6], [8], [10], [21]
a recent paper of the authors [17]. Still, it requires informa-  3   
tion on the climatic parameters. Hence, it is proposed as an Tc q EG 1 1
Id = Io, ref exp −
analytical model with a denoising-based wavelet algorithm to Tc,ref kA Tc,ref Tc
   
estimate them, which helps reducing the hardware using only qV
× exp − 1 = g (V, Tc ) (3)
one voltage sensor. The simulation and experimental results are kA Tc
provided to validate the proposed MPPT scheme capabilities.
This paper is organized as follows: Section II provides an where Io,ref is the reverse saturation current in the STC, EG
overview of how to estimate the junction current. The pro- is the band gap energy of the cell semiconductor, k is the
posed MPPT scheme is developed and explained in Section III. Boltzmann’s constant, and q is the electron’s electric charge.
Section IV shows in detail the estimation of environmental Finally, the cell current can be modeled as follows:
parameters. The simulation results based on MATLAB/SPS are
presented in Section V and the experimental verifications are I (V, G, Tc ) = f (G, Tc ) − g (V, Tc ) . (4)
discussed in Section VI. Section VII gives a conclusion.
It is called a “decoupled solar cell model” [17]. Due to
the nonlinear dependence of the junction current Id on the
cell voltage and temperature, an adaptive neuro-fuzzy (ANFIS)
II. E STIMATING THE A RRAY J UNCTION C URRENT
algorithm is proposed to estimate it as shown in Fig. 4 [17].
In this paper, the one-diode solar cell model is used to The cell current is then computed by determining separately
represent the PV cell as illustrated in Fig. 3. the photocurrent Iph and the junction current Id .
646 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 2, APRIL 2015

III. P ROPOSED MPPT T ECHNIQUE


The proposed MPPT scheme is built around the solar cell
model shown in Fig. 4 and in (1)–(3). Taking into account the
fact that Iph  Io , three equations have been developed based
on three array operating states as follows.
• In the MPP state:
 
Vm
Im = Iph − Io exp . (5)
Vt
Fig. 5. Proposed MPPT algorithm.
• In the short-circuit state:

Isc = Iph . (6) Finally, after some rearrangements, the power derivative is
• In the open-circuit state: ∂P V
  = I − Id . (15)
Voc ∂V Vt
Iph = Io exp (7)
Vt
We are seeking for the voltage that maximizes the power, so
where Im is the MPP current, Vm is the MPP voltage, and Voc

∂P

Vm
is the open-circuit voltage.

= 0 ⇒ Im − Idm = 0. (16)
By combining (6) and (7), we get ∂V V =Vm Vt
Isc
Io =  . (8) Thus, expression (16) can tell us easily when the PV array is
exp VVoct delivering its maximum power as follows [25]:

By replacing (8) in (5), we get Im Idm


Ym = = = Yd (17)
   Vm Vt
Vm − Voc
Im = Isc 1 − exp . (9)
Vt where Ym is the “instantaneous conductance” and Yd is defined
as the “junction conductance.”
From expression (9), we notice that, at the MPP, the only The proposed MPPT scheme is illustrated in Fig. 5.
unknown parameter is the thermal voltage expressed as The operation of this technique starts by measuring Vpv and
Vm − Voc estimating Ipv using the ANFIS model (Fig. 4). Next, the ther-
Vt =  . (10) mal voltage Vt is computed. However, for the algorithm to
ln IscI−I m
sc operate continuously, the parameters (G, Tc ) must be estimated
Expression (10) can be further simplified using the propor- as explained in the next section.
tionality relationship between the array Isc and Im [2]

Im = K1 Isc (11) IV. E STIMATING S OLAR R ADIATION AND C ELL


T EMPERATURE
where K1 depends on the array under use (0.78 ≺ K1 ≺ 0.92).
By replacing expression (11) in (10), the thermal voltage is Initially, an analytical model for the environmental parame-
ters (G, Tc ) is developed. Then, a wavelet denoising model is
Vm − Voc Vm − Voc proposed as a next step to get filtered out parameters.
Vt = = . (12)
ln (1 − K1 ) K2
To compute the array maximum power, the analytical output
power must be derived from (1) as follows: A. Analytical Model
    For commercial PV modules, manufacturers give only infor-
V
P = IV = Iph V − Io exp − 1 V. (13) mation on: (Im , Vm , Isc , Voc ) in the STC which are subject to
Vt
variations in the climatic conditions (G, Tc ) as [17], [35]–[37]
By deriving the power expression, we get    
        G G
∂P V V V ΔI = α (Tc − Tc,ref ) + − 1 Isc,ref (18)
= Iph − Io exp − 1 + Io exp −1 . Gref Gref
∂V Vt Vt Vt ΔV = −β(Tc − Tcref ) − Rs ΔI (19)
(14)

From expressions (1) and (3), we have where ΔI and ΔV are the corresponding variations in the cell
          current and voltage, α is the cell’s short-circuit current tempera-
V V ture coefficient, β is the cell’s open-circuit voltage temperature
Iph −Io exp −1 = I and Io exp −1 = Id .
Vt Vt coefficient, and Rs is the cell series resistance.
CHIKH AND CHANDRA: OPTIMAL MPPT ALGORITHM FOR PV SYSTEMS 647

The estimation of G and Tc is done by solving simultane-


ously (18) and (19) for G and Tc

Isc, ref + ΔI
G = Gref (20)
Isc, ref − αβ (Rs ΔI + ΔV )

and Fig. 6. Process of discrete wavelet transforms.

1
Tc = Tc,ref − (Rs ΔI + ΔV ) . (21)
β

Finally, the ambient temperature can be estimated as [17]


G
Ta = Tc − (NOCT − Ta,r ) (22)
Gr
where Ta,r = 20◦ C is the nominal ambient temperature,
NOCT is the nominal operating cell temperature, and Gr =
800 W/m2 is the nominal solar radiation.
Expressions (20)–(22) seem to be a nice estimation for the
environmental parameters.
Fig. 7. Overall control system.

B. Proposed Wavelet Denoising Model expressions (23) and (24) as shown in Fig. 6, where (↓ 2) is a
Wavelet theory has been considerably used recently [38]– down sampling with a factor of 2.
[42]. It allows a decomposition of complex information into Fig. 7 shows the overall control system where only one
basic forms at diverse positions and scales [38]. Wavelet trans- voltage sensor is used and the parameters (Ipv , Vt , G, Tc ) are
form is based on small waves positioned in different times, estimated.
which are computed using scaling and translation of a scaling
function; hence, it gives good capability to represent signals
V. S IMULATION R ESULTS AND D ISCUSSION
as compared to FT, which make it nicely applied to signal
denoising [38], [39]. Wavelet analysis is implemented using Simulations were done using a PV array built of two mono-
successive filters where the input at each step is constantly crystalline 80−W modules (Appendix) [43] (Figs. 1 and 2).
divided into two bands, whereas the lower band is furthermore
divided into two other bands till a chosen resolution is reached.
This can be resumed as follows [41], [42]: A. Performance of the Proposed MPPT Algorithm
To examine the performance of the proposed MPPT algo-
aj+1 (k) = h(2 k − n) aj (n) (23) rithm by means of MATLAB/Simulink, we simulated, in a short
n
time period of 0.4 s, three step variations in solar radiation (250,
2
dj+1 (k) = g(2 k − n) aj (n) (24) 500, 750, and 1000 W/m ) with a constant cell temperature
m Tc = 25◦ C. The resulting array operating voltage and current
are shown in Figs. 8 and 9.
where h(k) and g(k) refer, respectively, to unit sampling We can notice that the array voltage and current track sig-
responses of low-pass and high-pass digital filters, and aj and nificantly their corresponding references, which means that
dj are, respectively, the approximations and details coefficients the array is operating at its maximum power. This can be
at level j and k is the index of the coefficient at each level. confirmed by observing its power, which is following precisely
The filters are computed using the two scale equations [41] the corresponding MPP (Fig. 10). The tracking capability of the
−1 algorithm under varying solar radiation is shown in Fig. 11.
N

ϕ(t) = h(k) 2 ϕ (2t − k) (25) It is shown in Fig. 11 that, due to the first radiation value,
k=1 the voltage varies through a certain range before it converges to
−1
its first MP point P1 at Vpv = Vm1 . Then, Vm1 is used as the

N

ψ(t) = g(k) 2 ϕ (2t − k). (26) initial value for the next iteration. This procedure is continued
k=1
to reach P2 , P3 , and P4 when Vpv equals to Vm2 , Vm3 , and Vm4 ,
respectively. Then, the instantaneous and junction conductances
The scaling function ϕ(t) is associated with the low-pass are tested if they satisfy the condition in expression (17), as
filters {h(k), k ∈ Z} and the wavelet function ψ(t) is asso- shown in Figs. 12 and 13.
ciated with the high-pass filters {g(k), k ∈ Z}. The discrete It is seen clearly that the curves of Ypv (Vpv ) and Yd (Vpv )
wavelet transform algorithm performs the decomposition pro- intersect exactly when the MPP is achieved, which confirms
cess by passing the input signal through the filters defined by the validity of the proposed scheme. Moreover, this condition
648 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 2, APRIL 2015

Fig. 12. Proposed MPPT test condition.


Fig. 8. Array voltage Vpv .

Fig. 9. Array current Ipv .

Fig. 13. MPPT condition using Ypv and Yd .

TABLE I
A RRAY P OWER E STIMATION E RROR

Fig. 10. Array power Ppv .

The mean MPPT error through the entire solar radiation


range is found to be around 1.75 %, which is very reason-
able. Moreover, the radiation variations are done through a very
short period of time (t = 0.4 s). Therefore, it can be said that
the algorithm reacts adequately to fast fluctuations in climatic
conditions and operates using only one voltage sensor, which
results in a reduced hardware implementation.

Fig. 11. Array power tracking. B. Performance of the Climatic Parameters Estimator
Measuring the environmental parameters is a difficult task,
can also be verified by confirming that the difference between which yields to noisy and inefficient data. Therefore, using
the two conductances at the MPP condition is equal to zero, expressions (18) and (19), an analytical model is developed to
i.e., [Ypv (Vpv ) − Yd (Vpv )]MPPT = 0 as predicted in expres- estimate them. Also, in order to get usable data, a wavelet-based
sion (17) and illustrated in Fig. 13. The simulation results denoising algorithm is proposed to filter them out where the fil-
presented in Figs. 8–13 show obviously that the developed algo- ters are built based on expressions (23) and (24). The resulted
rithm ensures the maximum power operation of the PV array. radiation approximation coefficients aj | j=1, 2, ..., 7 and detail
Its estimated power follows the maximum power under very coefficients dj | j=1, 2, ..., 7 are shown in Fig. 14.
fast varying solar radiation with a very good precision. The The estimated cell and ambient temperatures are shown in
steady-state error is shown in Table I. Figs. 15 and 16.
CHIKH AND CHANDRA: OPTIMAL MPPT ALGORITHM FOR PV SYSTEMS 649

Fig. 17. Array voltage.

Fig. 14. Wavelet parameters for solar radiation.

Fig. 18. Array current.

Fig. 15. Estimated cell temperature.

Fig. 19. Array power.

to provide a 160−W array (see the Appendix). It is realized


Fig. 16. Estimated ambient temperature. using a dc power supply that operates in the current mode.
By stepping its output current sink, changes in radiation can
The estimated radiation in Fig. 14 exhibits a precise track be emulated following the curves illustrated in Figs. 8 and 10.
of its reference values. The estimated cell temperature is found Due to the limitation in the dSPACE, the time interval used
to be around 25◦ C (Fig. 15). Finally, ambient temperature is in experimental investigations is 40 s, which is small enough
shown in Fig. 16. From the observed curves, it is clear that the to confirm the algorithm response to rapid varying climatic
proposed model gives precise results; therefore, extra sensors to conditions. All the experimental curves are captured using the
measure (G, Tc ) are avoided. dSPACE ControlDesk Developer environment in real time. The
experimental results are presented in Figs. 17–25.

VI. E XPERIMENTAL V ERIFICATION


The experimental setup developed to validate the proposed A. MPPT Controller Operation
MPPT strategy includes a PV emulator, a boost converter con- To highlight the capability of the proposed MPPT scheme,
nected to a resistive load, a gate drive circuitry, a voltage the PV emulator is programmed to make three step changes in
sensor, a current sensor used only to visualize the array cur- its output current corresponding to a variation in solar radiation
2
rent, and an MPPT controller implementation based on DS1104 from 250, 500, 750, to 1000 W/m . During its operation, the
DSP of dSPACE (Fig. 7). A climatic parameters model and algorithm measures the voltage Vpv in order to estimates the
thermal voltage estimator are also used. The PV emulator rep- parameters G and Tc . Then, the photocurrent Iph is computed
resents two 80 W peak power PV panels coupled in series analytically using expression (2), whereas the diode current Id
650 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 2, APRIL 2015

Fig. 20. Instantaneous conductance.

Fig. 24. Solar radiation.

Fig. 21. Junction conductance.

Fig. 25. Cell temperature.

TABLE II
A RRAY P OWER E XPERIMENTAL E RROR
Fig. 22. Condutance difference Ypv − Yd .

The experimental results show a good agreement with the


predicted curves presented in the simulation. Moreover, the
array current, voltage, and power follow the reference values in
the hall PV operating range (Figs. 8–10). Thus, the algorithm
is seen to converge quickly under fast-varying solar radiation.
Fig. 23. MPPT response time due to a change in solar radiation.
On the other hand, the oscillation in the output power can be
reduced by increasing the value of the boost inductor L.
is estimated through the ANFIS algorithm presented in Fig. 4. It is noticed that the emulator current and voltage are very
Next, the thermal voltage Vt is computed using expression (12). close to the MPPT reference values (Figs. 17 and 18). Similarly,
Finally, all these parameters are injected into the MPPT con- its power follows the solar radiation changes almost instantly as
troller to compute the instantaneous and junction conductances, shown in Fig. 19 with an error of 2 %, which is too close to the
so that a corresponding PWM signal with an appropriate duty one predicted by the simulation (Table II).
cycle is generated for controlling the boost converter (Fig. 7). Finally, the MPPT operation can be verified by evaluating
The experimental results of the array voltage, current, and the conductances difference [Ypv − Yd ] under varying solar
power are shown in Figs. 17–19. radiation as shown in Figs. 20–22.
CHIKH AND CHANDRA: OPTIMAL MPPT ALGORITHM FOR PV SYSTEMS 651

It is found that the solar radiation G and the ambient tempera-


ture Ta follow precisely their respective reference step changes,
whereas the cell temperature Tc is constant and equals 25◦ C.
These results are similar to those predicted by the simulation
as shown in Figs. 14–16, which demonstrate that the proposed
estimator is capable to estimate climatic parameters perfectly.
The proposed method has four major advantages over the
techniques reported in the literature: 1) reduced hardware
implementation (only voltage sensor used); 2) fast tracking
Fig. 26. Ambient temperature.
capability; 3) very low error in the tracked power; and 4) finally,
it allows the estimation of the environment parameters.
It is clear from Fig. 22 that Ypv − Yd = 0. This interest-
ing result demonstrates that the MPP condition is reached.
Therefore, the proposed method is valid under fast-varying VII. C ONCLUSION
climatic conditions with high MPPT efficiency and reduced
hardware implementation. This paper develops an optimal MPPT method that has been
All the proposed MPPT techniques in the literature use at tested under varying climatic conditions. It has been shown that
least two sensors [2], [6]–[13], [28], [16], [17], [31], [32], it can be implemented with reduced hardware setup since only
which increases their hardware and noise measurements. On one voltage sensor has been used. The array current, solar radia-
the other hand, the losses in the power extracted from the array tion, and cell temperature are estimated using an ANFIS model
is reported to be equal to 8 % at 1000 W/m2 , where the MPPT and a climatic parameters estimator. The simulation results
implementation need two sensors for measuring the current and supported by experimental verification confirm the effective-
voltage [11], whereas the utilization efficiency varies from 95% ness of the presented method under varying solar radiation.
to 98.5% depending on the radiation with the use of two sen- The experimental results exhibit a mean MPPT efficiency error
sors [12]. Furthermore, an error of 4.46 % in maximum power of 2 % with a response time of 1.7 ms, which ensures that
tracking reported with the use of three sensors for measuring the developed technique can extract accurately the maximum
current, voltage, and solar radiation [28]. However, an impor- power from a PV array with a very low response time, which
tant improvement is pointed out in the proposed method, as it makes its operation optimal with reduced hardware.
needs only the measurement of the array voltage. For the pro-
posed methodology, the mean power loss for the hall PV range
is around 2 %, which an interesting result. A PPENDIX
On the other hand, it is important to verify the response time PV module characteristics [43]: Vm = 17.3 V, Im =
of the proposed methodology, as the papers presented in the 4.6 A, Pm = 80 W, Voc = 21.9 V, Isc = 5 A, α = 0.05 %/K,
literature show diverse results. The extremum seeking (ES) con- β = −0.36 %/K, and NOCT = 38◦ C ± 2◦ C.
troller achieves the MPP with response time of less than 0.1 s
[8]. However, in all the studied cases in [10], it takes 6.62 s
for the proposed particle swarm optimization (PSO) method to R EFERENCES
reach the MPP. In response to a sudden change in radiation, the
experiments in [11] show a typical convergence time of 15 ms [1] E. Koutroulis, K. Kalaitzakis, and N. C. Voulgaris, “Development of a
microcontroller-based, photovoltaic maximum power point tracking con-
where, in [12], with a 1−Vstep size, the algorithm to respond trol system,” IEEE Trans. Power Electron., vol. 16, no. 1, pp. 46–54, Jan.
to a step takes 5 s when using a perturbation frequency of 1 Hz, 2001.
while it takes only 0.5 s with a perturbation frequency of 10 Hz. [2] T. Esram and P. L. Chapman, “Comparison of photovoltaic array max-
imum power point tracking techniques,” IEEE Trans. Energy Convers.,
In [16], the fastest response achieved by the adaptive P&O- vol. 22, no. 2, pp. 439–449, Jun. 2007.
fuzzy algorithm is about 20 ms, followed by the conventional [3] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “Optimization of per-
FL controller (FLC) algorithm, which is around 35 ms and the turb and observe maximum power point tracking method,” IEEE Trans.
Power Electron., vol. 20, no. 4, pp. 963–973, Jul. 2005.
P&O algorithm shows the worst performance with about 40 ms. [4] N. Femia, D. Granozio, G. Petrone, G. Spagnuolo, and M. Vitelli,
The simulation results of the proposed GD-FNN show that the “Predictive & adaptive MPPT perturb and observe method,” IEEE Trans.
MPP can be achieved in 0.01 s [31], whereas the ANFIS MPPT Aerosp. Electron. Syst., vol. 43, no. 3, pp. 934–950, Jul. 2007.
[5] A. Pandey, N. Dasgupta, and A. K. Mukerjee, “High-performance algo-
controller in [32] reaches its steady state at 0.12 s. As compared rithms for drift avoidance and fast tracking in solar MPPT system,” IEEE
to the literature, it can be concluded that the proposed method- Trans. Energy Convers., vol. 23, no. 2, pp. 681–689, Jun. 2008.
ology demonstrates fast response time, which is around 1.7 ms, [6] D. Sera, R. Teodorescu, J. Hantschel, and M. Knoll, “Optimized max-
imum power point tracker for fast-changing environmental conditions,”
as shown in Fig. 23: IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 2629–2637, Jul. 2008.
[7] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, “A technique for
improving P&O MPPT performances of double-stage grid-connected
photovoltaic systems,” IEEE Trans. Ind. Electron., vol. 56, no. 11,
B. Climatic Parameters Estimator Operation pp. 4473–4482, Nov. 2009.
To verify the proposed climatic parameters estimator, exper- [8] S. L. Brunton, C. W. Rowley, S. R. Kulkarni, and C. Clarkson,
“Maximum power point tracking for photovoltaic optimization using
imental results of the solar radiation, cell, and ambient temper- ripple-based extremum seeking control,” IEEE Trans. Power Electron.,
atures are shown in Figs. 24–26. vol. 25, no. 10, pp. 2531–2540, Oct. 2010.
652 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 6, NO. 2, APRIL 2015

[9] F. Zhang, K. Thanapalan, A. Procter, S. Carr, and J. Maddy, “Adaptive [32] N. Khaehintung, P. Sirisuk, and W. Kurutach, “A novel ANFIS controller
hybrid maximum power point tracking method for a photovoltaic sys- for maximum power point tracking in photovoltaic systems,” in Proc. 5th
tem,” IEEE Trans. Energy Convers., vol. 28, no. 2, pp. 353–360, Int. Conf. Power Electron. Drive Syst. (PEDS’03), 2003, pp. 833–836.
Jun. 2013. [33] C. A. Otieno, G. N. Nyakoe, and C. W. Wekesa, “A neural fuzzy based
[10] K. L. Lian, J. H. Jhang, and I. S. Tian, “A maximum power point track- maximum power point tracker for a photovoltaic system,” in Proc. IEEE
ing method based on perturb-and-observe combined with particle swarm Conf. (AFRICON’09), Nairobi, Kenya, Sep. 23–25, 2009, pp. 1–6.
optimization,” IEEE J. Photovoltaics, vol. 4, no. 2, pp. 626–633, Mar. [34] A. Chikh and A. Chandra, “Optimization and control of a photovoltaic
2014. powered water pumping system,” in Proc. IEEE Elect. Power Energy
[11] Y. Levron and D. Shmilovitz, “Maximum power point tracking employing Conf. (EPEC), Montreal, Canada, 2009, pp. 1–6.
sliding mode control,” IEEE Trans. Circuits Syst., vol. 60, no. 3, pp. 724– [35] A. Chikh and A. Chandra, “Voltage and frequency controller for a stand-
732, Mar. 2013. alone PV system with battery storage element,” in Proc. 38th Annu. Conf.
[12] M. A. Elgendy, B. Zahawi, and D. J. Atkinson, “Assessment of the incre- IEEE Ind. Electron. Soc. (IECON’12), Montreal, Canada, Oct. 25–28,
mental conductance maximum power point tracking algorithm,” IEEE 2012, pp. 1172–1177.
Trans. Sustain. Energy, vol. 4, no. 1, pp. 108–117, Jan. 2013. [36] A. Chikh and A. Chandra, “Sizing and power management for a stand-
[13] L. Cristaldi, M. Faifer, M. Rossi, and S. Toscani, “An improved model- alone PV system in cold climate,” in Proc. IEEE PES Transm. Distrib.
based maximum power point tracker for photovoltaic panels,” IEEE Conf. Expo. (T&D), May 7–8, 2012, pp. 1–6.
Trans. Instrum. Meas., vol. 63, no. 1, pp. 63–71, Jan. 2014. [37] H. Liu, G. Liu, and Y. Shen, “A novel harmonics detection method based
[14] J.-S. R. Jang, “ANFIS: Adaptive-network-based fuzzy inference system,” on wavelet algorithm for active power filter,” in Proc. 6th World Congr.
IEEE Trans. Syst. Man Cybern., vol. 23, no. 3, pp. 665–685, May/Jun. Intell. Control Autom., China, Jun. 21–23, 2006, pp. 7617–7621.
1993. [38] S. G. Mihov, R. M. Ivanov, and A. N. Popov, “Denoising speech signals
[15] M. Veerachary, T. Senjyu, and K. Uezato, “Feedforward maximum power by wavelet transform,” Annu. J. Electron., pp. 1–6, 2009, ISSN 1313-
point tracking of PV systems using fuzzy controller,” IEEE Trans. Aerosp. 1842.
Electron. Syst., vol. 38, no. 3, pp. 969–981, Jul. 2002. [39] J. Ning and W. Gao, “Continuous wavelet-based active filter design for
[16] M. A. A. M. Zainuri1, M. A. M. Radzi1, A. C. Soh1, and N. Abd Rahim, harmonic mitigation in hybrid electric vehicles,” in Proc. IEEE Conf. Veh.
“Development of adaptive perturb and observe-fuzzy control maximum Power Propul., Michigan, Sep. 7–11, 2009, pp. 859–865.
power point tracking for photovoltaic boost dc–dc converter,” IET Renew. [40] H. T. Yalazan, T. Sürgevil, and E. Akpinar, “Wavelet transform applica-
Power Gener., vol. 8, no. 2, pp. 183–194, Mar. 2014. tion in active power filter used for slip energy recovery drives,” in Proc.
[17] A. Chikh and A. Chandra, “Adaptive neuro-fuzzy based solar cell model,” Int. Aegean Conf. Elect. Mach. Power Electron., Turkey, Sep. 10–12,
IET Renew. Power Gener., vol. 8, no. 6, pp. 679–686, Aug. 2014. 2007, pp. 398–403.
[18] J.-S. R. Jang, C.-T. Sun, and E. Mizutani, Neuro-Fuzzy and Soft [41] A. H. Ghaemi, H. Askarian Abyaneh, K. Mazlumi, and S. H. H. Sadeghi,
Computing. Englewood Cliffs, NJ, USA: Prentice-Hall, 1997. “Voltage notch indices determination using wavelet transform,” in Proc.
[19] H. P. Desai and H. K. Patel, “Maximum power point algorithm in PV IEEE Powertech, Switzerland, Jul. 1–5, 2007, pp. 80–85.
generation: An overview,” in Proc. 7th Int. Conf. Power Electron. Drive [42] K. H. Kashyap and U. J. Shenoy, “Classification of power system faults
Syst. (PEDS’07), 2007, pp. 624–630. using wavelet transforms and probabilistic neural networks,” in Proc.
[20] C. W. Tan, T. C. Green, and C. A. Hernandez-Aramburo, “An improved Proc. Int. Symp. Circuits Syst. (ISCAS’03), Bangkok, May 25–28, 2003,
maximum power point tracking algorithm with current-mode control pp. 423–426.
for photovoltaic applications,” in Proc. IEEE Int. Conf. Power Electron. [43] [Online]. Available: http://website www.mrsolar.com/content/pdf/Solar
Drives Syst., 2005, pp. 489–494. tech/
[21] Y. Yusof, S. H. Sayuti, M. Abdul Latif, and M. Z. Che Wanik, “Modeling
and simulation of maximum power point tracker for photovoltaic system,”
in Proc. Nat. Power Energy Conf. (PECon), Kuala Lumpur, Malaysia,
2004, pp. 88–93. Ali Chikh (S’09–M’11) received the B.E. degree in
[22] J. Ho Lee, H. Bae, and B. H. Cho, “Advanced incremental conductance electrical engineering from the National Institute of
MPPT algorithm with a variable step size,” in Proc. 12th Int. Power Electricity and Electronics, Boumerdès, Algeria, in
Electron. Motion Control Conf. (EPE-PEMC’06), Portoroz, Slovenia, 1993, and the master’s degree in electrical engineer-
Aug. 30/Sep. 1, 2006, pp. 603–607. ing from the University of Sciences and Technology
[23] B. Liu, S. Duan, F. Liu, and P. Xu, “Analysis and improvement of of Algiers, Algiers, Algeria, in 1998. He is currently
maximum power point tracking algorithm based on incremental con- pursuing the Ph.D. degree at the Electrical Machines
ductance method for photovoltaic array,” in Proc. 7th Int. Conf. Power Laboratory, Department of Electrical Engineering,
Electron. Drive Syst. (PEDS’07), Bangkok, Thailand, Nov. 27–30, 2007, École de Technologie Supérieure (ETS), Montreal,
pp. 637–641. QC, Canada.
[24] Z. Yan, L. Fei, Y. Jinjun, and D. Shanxu, “Study on realizing MPPT by He worked as a Researcher with the Research
improved incremental conductance method with variable step-size,” in Unit of Renewable energies, Algeria. He is working as a Teacher with the
Proc. IEEE 3th Ind. Electron. Appl. Conf., 2008, pp. 547–550. Electrical Machines Laboratory, Department of Electrical Engineering, École
[25] A. Chikh and A. Chandra, “An optimum method for maximum power de Technologie Supérieure (ETS), Montreal, QC, Canada. His research interests
point tracking in photovoltaic systems,” in Proc. IEEE Power Energy Soc. include renewable energy resources, power quality, and active power filters.
Gen. Meeting, Detroit, MI, USA, 2011, pp. 1–6.
[26] K. Samangkool and S. Premrudeepreechacharn, “Maximum power point
tracking using neural networks for grid-connected photovoltaic system,” Ambrish Chandra (S’85–M’86–SM’99–F’14)
in Proc. Int. Conf. Future Power Syst., 2005, 4 pp. received the B.E. degree from the University of
[27] R. Ramaprabha1, B. L. Mathur, and M. Sharanya, “Maximum power Roorkee (presently IIT), Roorkee, India, the M.Tech.
point tracking using GA-optimized artificial neural network for solar PV degree from the Indian Institute of Technology (IIT),
system,” in Proc. 1st Int. Conf. Elect. Energy Syst. (ICEES), India, 2011, New Delhi, India, and the Ph.D. degree from the
pp. 264–268. University of Calgary, Calgary, AB, Canada, in 1977,
[28] B. M. Wilamowski and X. Li, “Fuzzy system based maximum power 1980, and 1987, respectively.
point tracking for PV system,” in Proc. 28th IEEE Annu. Conf. Ind. He worked as a Lecturer and later as a Reader
Electron. Soc. (IECON’02), 2002, pp. 3280–3284. with the University of Roorkee. In 1994, he joined
[29] A. M. Subiyanto and M. A. Hannan, “Maximum power point tracking in the Electrical Engineering Department, École de
grid connected PV system using a novel fuzzy logic controller,” in Proc. Technologie Supérieure (ÉTS), Universié du Québec,
Stud. Conf. Res. Develop. (SCOReD’09), UPM Serdang, Malaysia, Nov. Montréal, QC, Canada, as an Associate Professor and became a Full Professor
16–18, 2009, pp. 349–352. in 1999. Currently, he is the Director of a multidisciplinary graduate program
[30] L. Chun-hua, J. Xu, and Z. Xin-jian, “Study on control strategy for pho- on renewable energy and energy efficiency at ÉTS.
tovoltaic energy systems based on recurrent fuzzy neural networks,” in Prof. Chandra is a Fellow of Institution of Engineering and Technology,
Proc. 5th Int. Conf. Nat. Comput., 2009, pp. 282–286. U.K. (FIET), Fellow of Institution of Electronics and Telecommunication
[31] X. Yang and C. Zeng, “Generalized dynamic fuzzy neural network-based Engineers, India (FIETE), Fellow of Institution of Engineers, India (FIE), and
tracking control of PV,” in Proc. Asia-Pac. Power Energy Eng. Conf. Life Member of Indian Society for Technical Education (ISTE). He is registered
(APPEEC), 2010, pp. 1–4. as a Professional Engineer in the province of Quebec.

S-ar putea să vă placă și