This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

A Novel MPPT Technique based on an Image of

PV Modules
Yousef Mahmoud, Member, IEEE, and Ehab El-Saadany, Senior Member, IEEE

advantage of integrating a single solar panel (50–400 W) into

Abstract—PV modules operating under partially shaded the grid and thus allows for independent maximum power
conditions exhibit multiple peaks in their output power curves, point tracking, which improves the harvested energy.
which cause the majority of the maximum power point tracking However, the problem of the appearance of multiple peaks
(MPPT) techniques to become trapped in a local power peak.
still exists because the PV module itself is susceptible to
This unfortunately leads to additional energy losses that could
otherwise be harvested if the global maximum power peak partial shading.
(GMPP) were correctly tracked. The available MPPT methods Numerous MPPT techniques have been developed in the
that are able to track the GMPP require periodic scanning of the literature for tracking the GMPP of partially shaded PV
PV curve, which disturbs the operation of the system and causes modules. In [7], a short circuit pulse-based MPPT is proposed
energy losses. A new MPPT technique is proposed in this paper but requires measuring the short circuit current of the PV
that is distinguished by its ability to find the GMPP without the
system periodically. A similar method, though based on open
need for periodic curve scanning. The proposed method utilizes
the mathematical model of the PV module, as well as the circuit voltage measurement, is available in [8]. Both methods
irradiances received by its PV cells, to analytically calculate the cause momentary energy losses due to the periodic
location of the GMPP. The required irradiances are innovatively disconnection of the PV system for a short circuit current and
estimated using an image of the PV module captured by an open circuit voltage measurements [4].
optical camera. The proposed method is also combined with the The line search algorithm with Fibonacci sequence,
P&O method to compensate for errors in the model or irradiance
presented in [9], uses the Fibonacci sequence to track the
estimation. Experimental verifications are conducted to validate
the effectiveness of the proposed MPPT method under various GMPP under partially shaded conditions. The major drawback
shading scenarios. of this method is the possibility of missing the GMPP under
Index Terms—Maximum power point trackers, Modeling, some shading conditions [10, 11]. The load-line MPPT
partial shading, perturb and observe (P&O) algorithm, approach proposed in [12] uses the computed load line to find
photovoltaic systems, solar Energy. the vicinity of the GMPP. However, the GMPP cannot be
obtained if it lies on the left side of the load line [11]. In [13],
I. INTRODUCTION an instantaneous operating power optimization approach is

T he main source of energy losses in PV systems is caused

by partial shading, which occurs when the PV units in a
PV system do not all receive the same irradiance [1, 2]. Unlike
proposed. It relies on the linear relationships between the
optimal current and max power as well as the optimal current
and short circuit current to estimate the GMPP. However, all
PV systems receiving homogeneous conditions in which a the voltage factors of all the maximum power peaks (MPPs)
single power peak exists in the output power curve, partially have to be previously known, which means that the method is
shaded PV systems may exhibit several power peaks [2, 3]. system-dependent [11].
Therefore, tracking the GMPP in partially shaded systems can A new observation about the shape of power-voltage curves
be challenging because the conventional MPPT algorithms of partially shaded PV systems is utilized to build an effective
(such as P&O, hill climbing, incremental conductance, fuzzy and fast MPPT method in [14]. However, the comprehensive
logic, etc.) can get trapped in a local power peak and thus miss study in [15] demonstrated that the observation is not correct
the true GMPP [4]. In addition to the energy losses caused by for all shading scenarios, and therefore the method cannot
missing the GMPP, partial shading causes extra losses due to guarantee finding the GMPP.
power mismatching between series-connected PV units The Dividing Rectangles Technique proposed in [16]
receiving unequal irradiance levels [5]. Unfortunately, this employs a Lipschitz condition and advanced mathematical
results in significant overall energy losses that may exceed tools to track the GMPP. However, if the initial point in the
70%, as in [6]. algorithm is not chosen properly, the controller may get
An effective solution which has gained popularity for trapped in a local power point [4]. Another method is
reducing the mismatch energy losses is using micro inverters proposed in [17] which relies on a constant power operation to
instead of a single centralized inverter. This approach has the track the GMPP. Unlike the majority of the methods which
track the power peak by incrementing the current or voltage,
Y. Mahmoud is with the Department of Electrical and Computer Engineering, this method directly increments power using a constant power
Worcester Polytechnic Institute, Worcester, 01607, USA (e-mail:
[email protected]). electronic converter. The deficiency of this method is the slow
E. El-Saadany (currently on leave with the PI, UAE) is with the Department tracking due to the small power perturbation that is required to
of Electrical and Computer Engineering, University of Waterloo, Waterloo, detect small power differences between power peaks [18].
ON N2L 3G1, Canada.
Biological optimization algorithms such as Artificial Bee

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

Colony [19] and particle swarm [11, 20, 21] have been
extensively utilized for maximum point tracking. However,
they suffer from slow settling time due to the need to scan the
power curve [10, 19]. Moreover, they rely heavily on the Incident
precision of a few parameters [22] and cause oscillations in Irradiance Reflected Irradiance
output power before reaching the GMPP [19, 23]. in the direction of the
Camera
In addition to the highlighted drawbacks in the
aforementioned techniques, the majority require a periodic
scanning of the power curve. Unfortunately, this has several
drawbacks [18]. For instance, it causes a reduction in the
extracted energy because it forces the PV system to operate on
PV Module
non-maximum power points. Also, it disturbs the tracking
during normal operation (homogeneous conditions) of the
system, causing further energy losses.
A different solution is developed in [24], in which a thermal
camera is employed to find the GMPP. The thermal camera
provides the temperatures of the PV cells, after which Fig. 1. The proposed system used to estimate the incident irradiance.
empirical approximations are used to locate the GMPP. This
method could suffer from poor accuracy, however, due to the
Reflected Irradiance in
approximations used. Moreover, the expensive cost of thermal the Direction of the
Incident
cameras hinders the method’s practical implementation. Image
Camera
Irradiance
Camera’s Reflectance of PV
In this paper, an innovative approach based on an optical Response Function Cell
camera is proposed to track the GMPP in partially shaded PV
Fig. 2. A flowchart illustrating the proposed irradiance estimation method.
systems. The image of a PV module, captured by an optical
camera, is utilized to estimate the incident irradiance levels
of photo-sensors sensitive to light. Each photo-sensor is a
received by its cells. The irradiance values are then inputted to
small rectangular box that converts light into voltage. Digital
the PV mathematical model to calculate the GMPP
images are represented by a matrix of numbers, called pixels,
analytically. Because any inaccuracy in the estimated incident
which relay the intensity of light received by the
irradiances would directly affect the accuracy of the
corresponding photosensitive sensors. A digital image I(x, y)
determined GMPP, the well-known iterative P&O is also
denotes the image brightness at pixel (x, y) and encodes the
utilized to correct any imprecision in the determined GMPP.
intensity recorded by the photo-sensors of the array
The paper is organized as follows: Section II develops the
contributing to that pixel. It is an integer usually in the range
proposed irradiance estimation method. Section III adjusts the
[0-255] (typically, 0 is black and 255 is white) [25].
estimated irradiances to the elevation and azimuth angles of
The relation between the received intensity of light at a
the sun. Section IV presents a global power peak estimator.
photo-sensor (x, y) and the corresponding brightness of a pixel
Finally, a developed prototype in Section V is used to
I(x, y) is not linear. There is usually a nonlinear mapping
experimentally validate the proposed MPPT.
(called the camera response function) that determines how
radiance in the scene becomes pixel values in the image.
II. PROPOSED IRRADIANCE ESTIMATION METHOD
In the literature, there are several methods to determine the
In the proposed irradiance estimation approach, the camera response function of a camera. One effective method is
continuously takes images of the PV module’s surface, as presented in [26], which determines response function using a
shown in Fig. 1. The captured images are then processed to set of images taken for the same scene at different exposure
extract the incident irradiance on the cells of the PV module. durations. The exposure duration is the time during which the
As shown in Fig. 2, this requires knowledge of two relations: aperture of the camera opens while capturing an image.
first, the camera’s response function, which is the relation Using that algorithm, the response function of a Canon
between the reflected irradiance in the direction of the camera Power Shot A620, which is the camera used in this paper, is
and the intensity of the corresponding image; and second, the estimated. Four images for an arbitrary scene are captured at
reflectance of the PV cell, which is the relation between the different exposure durations, as shown in Fig. 3, and are
incident irradiance on a PV cell and the reflected irradiance in inputted to the algorithm to estimate the response function.
the direction of the camera. The following describes the pre- The resulting response function is depicted in Fig. 4. As can
characterization of both relations. be seen, it contains a mapping for all pixel values from 0 to
A. Camera’s Response Function 255 to a corresponded irradiance. It is important to note that
the corresponding irradiance (i.e., y-axis in Fig. 4,) is
Optical cameras consist of an array of X×Y rectangular grid
determined up to an unknown scale factor K. However, as will
be shown later, this will not pose a problem.

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

(a) (b)

(c) (d)

Fig. 3. Four images taken by Canon Power Shot A620 at different exposure
durations.
Fig. 5 The reflectance map of a monocrystalline PV cell.

100

80
( scale factor = K )
Scaled Irradiance

60

40

20 (a) (b) (c)

0
0 50 100 150 200 250 300
Pixel Intensity

Fig. 4. The response function for Canon Power Shot A620.

(d) (e) (f)

B. Reflectance of PV Cell
The reflectance of a PV cell refers to how much of the Fig. 6 Images of the PV cell under various irradiance levels
incident irradiance is reflected in the direction of the camera.
It is the ratio between the reflected irradiance in the direction of the PV cell are extracted, they are used to estimate the
of the camera and the incident irradiance on the PV cell. It is incident irradiances from the captured images. In this section,
entirely a surface property and is not dependent on the captured images of the PV cell at six different levels of
incident light. irradiance, depicted in Fig. 6, are used with the two extracted
Measuring the reflectance of a PV cell requires capturing an relations from previous subsections to find the incident
image of the cell at a known level of incident irradiance. The irradiance. An illustrative diagram for the experimental setup
ratio between the reflected irradiance in the direction of the is depicted in Fig. 7. As seen, it consists of a lamp (lighting
camera (which is the outcome of processing the captured source), a PV cell, a camera, and a current meter for
image through the camera’s response function shown in Fig. estimating the irradiance through the short circuit current.
4) and the known incident irradiance is the reflectance of the Each pixel in each image is first mapped through the
PV cell. camera response function shown in Fig. 4. Then, the mean of
First, a monocrystalline PV cell is captured by a Canon the resulting array, which represents the scaled irradiance
Power Shot A620 camera under a known level of irradiance reflected to the camera, is divided by the mean of the scaled
(462 W/m2). Each pixel in the captured image is then mapped reflectance array shown in Fig. 5 to find the incident
through the response function shown in Fig. 4. The ratio irradiance on the PV cell. Because both the irradiance
between the resulting array of numbers to the known reflected to the camera and the reflectance is scaled to the
irradiance is depicted in Fig. 5, which is the reflectance of the same factor K, the division will cancel the unknown K factor
PV cell. Because the used response function is scaled to factor (Incident Irradiance = K × Reflected Irradiance / K ×
K, the resulting reflectance is also scaled to factor K. The two Reflectance). Table I summarizes the estimated irradiance
high reflectivity lines in the reflectance map represent the two levels using the proposed approach and compares them to the
metallic current collecting grids in PV cells. measured irradiances. The percentage deviation between the
estimated and measured irradiances is also shown in the table.
C. Experimental Verification It is important to mention that the STC irradiance (Go=1000
Once both the camera response function and the reflectance W/m2) is used as the base for the percentage.

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

Fig. 7 An illustrating diagram for the experimental setup.

III. ADJUSTMENT TO LIGHTING SOURCE ANGLES

In this section, it is shown that the reflectance of PV cells Fig. 8 The Geniophotometer constructed by the authors.
changes with variations in lighting source angles in both
elevation and azimuth angles. Because the sun’s location in
the sky changes during the day, it is important to investigate
the effect of the sun angle on the proposed irradiance
estimation method. The authors built a Geniophotometer for
this purpose, shown in Fig. 8, which is a device used to allow
positioning the lighting source (a lamp) at various angles in
the semi-sphere surrounding the PV cell. The azimuth angle of
the lighting source is varied from 0° to 180° with a step of
10°; at each step, the elevation angle is also varied between 0°
and 180°. At each angle, an image is captured, the irradiance
is measured, and the mean reflectance is calculated. The
reflectance at all the angles is plotted in Fig. 9. The data in the
figure can be saved in a look-up table to be used for irradiance
estimations at different lighting source angles. The sun
elevation and azimuth angles at any time of the day and on
any day of the year can be calculated via the equations
provided in [27, 28]. It should be noted that the estimated Fig. 9 The reflectance at various elevation and azimuth angles.
angles are with respect to PV panels placed horizontally, and
thus the angles should be adjusted if the panels are inclined The irradiances were then estimated from the captured images,
with an angle. Once the angles are obtained, they are used to taking into consideration the different angles of the sun at time
find the reflectance of the PV cell using Fig. 9, after which the the image was captured. The resulting estimated irradiances
irradiance is estimated as described in the previous section. were compared to the measured irradiances in Table II. The
To experimentally test the effectiveness of the proposed sun angles for the images are also provided in the table. As
irradiance estimation method at different angles, the can be seen, the table reveals good accuracy for the proposed
experiment conducted in the previous section is repeated at method.
various times of the day. Ten images were captured at
different irradiance levels at different times. TABLE II
COMPARISON BETWEEN THE MEASURED AND ESTIMATED IRRADIANCES OF
THE PV CELL UNDER VARIOUS ANGLES
TABLE I
COMPARISON BETWEEN THE MEASURED AND ESTIMATED IRRADIANCES OF Measured Elevation Azimuth Estimated Error %
THE PV CELLS SHOWN IN FIG. 6 (W/m2) Angle Angle (W/m2)
675.3 40.20 10.9 683.7 0.84
Image # Measured Estimated Error % 701.4 39.60 9.5 705 0.36
(W/m2) (W/m2) 536 25.20 15.0 516.9 1.90
(a) 410 423.8 1.38 217.2 23.60 17.0 255.9 3.90
(b) 368 387.2 1.92 448 22.70 18.0 410.7 3.70
(c) 310 339.4 2.94 369.5 21.90 19.0 350.9 1.90
(d) 230 256.2 2.62 232 21.00 20.0 258.2 2.60
(e) 134 165.6 3.16 289.2 21.00 20.0 280 0.92
(f) 63 83.2 2.02 213.3 13.20 28.4 160.0 5.33
84.7 13.20 28.4 100.4 1.57

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

IV. MODEL-BASED GLOBAL POWER PEAK ESTIMATOR Substituting (1) and (7) in (6) results in
𝑑𝑑𝑑𝑑 𝑅𝑅𝑠𝑠ℎ .𝐼𝐼.�𝐼𝐼𝑝𝑝ℎ +𝐼𝐼𝑠𝑠 −𝐼𝐼�.Lambert(𝑊𝑊)
As discussed in the introduction, the estimated irradiances = 𝑅𝑅𝑠𝑠ℎ . �𝐼𝐼𝑝𝑝ℎ + 𝐼𝐼𝑠𝑠 � − −
𝑑𝑑𝑑𝑑 1+Lambert(𝑊𝑊)
will be used for the MPPT by utilizing the PV mathematical
2𝐼𝐼. (𝑅𝑅𝑠𝑠 + 𝑅𝑅𝑠𝑠ℎ ) − 𝑎𝑎. Lambert(𝑊𝑊) (8)
model. In this section, a part of the general global power peak
Substituting (8) in (5) yields
estimator, proposed in [29] by the same authors, will be
𝑛𝑛 𝑛𝑛𝑠𝑠 𝑅𝑅𝑠𝑠ℎ .�𝐼𝐼𝑝𝑝ℎ +𝐼𝐼𝑠𝑠 −𝐼𝐼�Lambert(𝑊𝑊𝑖𝑖 )𝐼𝐼
reviewed in the context of this paper. ∑𝑗𝑗=1
𝑑𝑑
𝑍𝑍𝑗𝑗 �∑𝑖𝑖=1 �𝑅𝑅𝑠𝑠ℎ �𝐼𝐼𝑝𝑝ℎ𝑖𝑖 + 𝐼𝐼𝑠𝑠𝑖𝑖 � − 𝑖𝑖 𝑖𝑖
−
��1+Lambert(𝑊𝑊𝑖𝑖 )��
A. Global Power Peak Estimator
The adopted mathematical model in the proposed global 2𝐼𝐼. (𝑅𝑅𝑠𝑠 + 𝑅𝑅𝑠𝑠ℎ ) − 𝑎𝑎𝑖𝑖 . Lambert(𝑊𝑊𝑖𝑖 ) �� = 0 (9)
𝑗𝑗
power peak estimator is the Lambert PV circuit model [30].
The Lambert based model was chosen in this paper because it Equation (9) can be easily solved by any numerical method
adopts a non-transcendental equation which reduces the such as the well-known Newton Raphson method. Solving (9)
computational effort. The relationship between the output for a PV module produces the value of the MPP current
current I and terminal voltage V of a PV module in this model corresponding to one of the power peaks. Because multiple
is represented as provided in [30, 31] power peaks may exist, equation (9) should be solved several
times at various initial guesses to determine the MPP currents.
𝑉𝑉 = �𝐼𝐼𝑝𝑝ℎ + 𝐼𝐼𝑠𝑠 �. 𝑅𝑅𝑠𝑠ℎ − (𝑅𝑅𝑠𝑠 + 𝑅𝑅𝑠𝑠ℎ ). 𝐼𝐼 − 𝑎𝑎. Lambert(𝑊𝑊) (1)
The maximum number of power peaks for a PV module is
equal to the number of its bypass diodes. Therefore, three
𝑅𝑅𝑠𝑠ℎ .�𝐼𝐼𝑝𝑝ℎ +𝐼𝐼𝑠𝑠 −𝐼𝐼�
� � initial guesses are usually needed, one for each bypass diode.
𝑅𝑅𝑠𝑠ℎ .𝐼𝐼𝑠𝑠 𝑎𝑎
𝑊𝑊 = e (2) An initial guess for a bypass diode is best chosen to be 90% of
𝑎𝑎
the photon current of the PV cell receiving the lowest
where Iph, Is, Rs, Rsh and Ns are the photon current, saturation irradiance level in a group of cells connected to the bypass
current, series resistant, shunt resistant, and the number of diode. The integer Zj for a group is equal to one if the photon
series PV cells in a PV module, respectively, and the constants current of the group is higher than the used initial guess.
K and q are the Boltzman constant and electron charge, Otherwise, it is equal to zero.
respectively. The parameter a is equal to NsKTA/q, where T is The power peak at a determined MPP current, resulting
the PV module temperature and A is the ideality factor of the from solving (9), can be calculated by summing the powers
diode. generated from all the series-connected PV cells operating at
The photon and saturation currents are calculated as [32] the MPP current using
𝐼𝐼𝑝𝑝ℎ = 𝐺𝐺. (𝐼𝐼𝑠𝑠𝑠𝑠 + α. ∆𝑇𝑇) (3) ∑𝑗𝑗=1
𝑑𝑑 𝑛𝑛
𝑃𝑃𝑗𝑗 = ∑𝑗𝑗=1
𝑑𝑑 𝑛𝑛 𝑛𝑛𝑠𝑠
�𝑍𝑍𝑗𝑗 �𝐼𝐼 × ∑𝑖𝑖=1 𝑉𝑉𝑖𝑖 �𝑗𝑗 � (10)
𝑞𝑞(𝑉𝑉𝑜𝑜𝑜𝑜+𝛽𝛽.∆𝑇𝑇) 𝑗𝑗
𝑉𝑉𝑜𝑜𝑜𝑜 +𝛽𝛽.∆𝑇𝑇
𝐼𝐼𝑠𝑠 = [(𝐼𝐼𝑠𝑠𝑠𝑠 + α. ∆𝑇𝑇) − ]/[𝑒𝑒 𝑁𝑁𝑠𝑠 𝐾𝐾𝐾𝐾𝐾𝐾 − 1] (4) Substituting (1) in (10), equation (10) becomes
𝑅𝑅𝑠𝑠ℎ
where G is the incident irradiance and ∆T is the difference 𝑛𝑛
∑𝑗𝑗=1
𝑑𝑑
𝑃𝑃𝑗𝑗 = ∑𝑗𝑗=1
𝑑𝑑 𝑛𝑛 𝑛𝑛𝑠𝑠
�𝑍𝑍𝑗𝑗 . ∑𝑖𝑖=1 ��𝐼𝐼𝑝𝑝ℎ𝑖𝑖 + 𝐼𝐼𝑠𝑠 𝑖𝑖 �. 𝑅𝑅𝑠𝑠ℎ . 𝐼𝐼 − (𝑅𝑅𝑠𝑠 +
between the PV cell temperature T and the temperature at
standard test conditions (25 C°). The parameters α and β are 𝑅𝑅𝑠𝑠ℎ ). 𝐼𝐼 2 − 𝑎𝑎𝑖𝑖 . 𝐼𝐼. Lambert(𝑊𝑊𝑖𝑖 )� − (1 − 𝑍𝑍𝑗𝑗 ). 𝑛𝑛𝑠𝑠 . 𝐼𝐼. 𝑉𝑉𝑑𝑑 � (11)
𝑗𝑗
the current and voltage temperature coefficients. Relying on equations (9) and (11), which determine the
PV modules are composed of series-connected PV cells and MPP currents and powers, respectively, an MPPT method is
bypass diodes. Three bypass diodes are usually used in developed that is able to find the GMPP of partially shaded
commercial PV modules. The bypass diodes cause the PV modules. The proposed MPPT utilizes the aforementioned
appearance of multiple peaks in the output power curve. The equations to analytically find the GMPP and its voltage. Since
value of the MPP currents at the power peaks can be there might be an error in the estimated irradiances which
calculated by equating the summation of the derivatives of the could affect the accuracy of the determined GMPP, the well-
power-current curves for all series PV cells to zero known P&O MPPT method will be utilized to compensate.
𝑛𝑛 𝑛𝑛𝑠𝑠 𝑑𝑑𝑃𝑃𝑖𝑖
∑𝑗𝑗=1
𝑑𝑑
𝑍𝑍𝑗𝑗 . �∑𝑖𝑖=1 � =0 (5) A flowchart for the proposed MPPT method is shown in
𝑑𝑑𝑑𝑑 𝑗𝑗
Fig. 10. It starts by calculating the photon and saturation
where nd and ns refer to the numbers of bypass diodes and PV
currents for all the PV cells using (3) and (4). The initial
cells connected to each bypass diode, respectively. Here, the
guesses are then estimated. As mentioned earlier, they can be
variable Zj is an integer modeling the effect of a bypass diode.
best considered as 90% of the photon currents of the PV cells
It is equal to zero when the corresponding bypass diode is in
receiving the lowest irradiance levels in each group of PV
the conducting state; otherwise, it is equal to one. The
cells connected with a bypass diode. Next, equations (9) and
derivative of the power-current curve of a PV unit dP/dI is
(13) are solved at the determined initial guesses. This gives the
expressed as
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑 values of MPP powers and currents. The maximum power
= 𝑉𝑉 + 𝐼𝐼. (6) peak among the resulting peaks is the GMPP, and its
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
corresponding voltage is the GMPP voltage. Because there
where dV/dI is the voltage derivative with respect to current
could be errors in the irradiance estimation and the
and is expressed as
𝑑𝑑𝑑𝑑 𝑅𝑅𝑠𝑠ℎ .�𝐼𝐼𝑝𝑝ℎ +𝐼𝐼𝑠𝑠 −𝐼𝐼�.Lambert(𝑊𝑊)
mathematical model, the determined GMPP voltage may not
= −(𝑅𝑅𝑠𝑠 + 𝑅𝑅𝑠𝑠ℎ ) − (7) be exactly accurate. Therefore, the error in the estimated
𝑑𝑑𝑑𝑑 1+Lambert(𝑊𝑊)

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

Irradiances and Temperatures of the PV

Cells

PV Cell operating at
1 KW/m2 and 50°
Calculate photon and saturation currents
Iph , Is for each PV cell using (3) and (4) PV Cell operating at
0.5 KW/m2 and 35°
PV Cell operating at
Find the minimum photon current Iph_min 0.2 KW/m2 and 20°
(the initial guess ) for each group of series
PV units connected with a bypass diode

Solve (9) and (13) for each (a) (b)

initial guess

Select the maximum peak

Pm and its corresponded
current Im

P&O MPPT
Find the MPP Voltage Vm

Vm
++
Vp
(c) (d)

Fig. 11 The different shading patterns of the PV module under study

Global MPP
Voltage
100
Fig. 10 The flowchart of the proposed MPPT method (a)
80 (b)
GMPP will be compensated by the P&O method combined (c)
Power (W)

60
with the proposed method. In other words, the proposed (d)
MPPT will find an approximated GMPP in the vicinity of 40
GMPP and then the P&O will converge to the exact GMPP.
This means that if dirt or moisture on the camera affects the 20
accuracy of the irradiance estimation, this will not affect the
0
correct operation of the proposed MPPT method. The resulting 10 20 30 40
overall voltage is the final GMPP voltage, which will be PV Voltage (V)
supplied to the inverter. Fig. 12 Power-voltage curves of the PV module under shading patterns shown
Finally, it is important to note that the presented MPPT in Fig. 9.
method requires the irradiances and temperatures of the PV
modules. A PV module of the type “JAM5(l)-72-155” is
cells. The estimation of the irradiances is already discussed in
used for the simulation under various shading patterns, as
the previous section. The temperatures can be determined
shown in Fig. 11. The simulation starts with the shading
through the thermal model of PV modules presented in [33].
pattern (a) for 20 s, then (b) and (c), and finally (d), each with
The difference between the atmospheric and PV cell
the same time period. The power curves for the PV module
temperatures is directly related to the irradiance G received by
under the various scenarios are also depicted in Fig. 12.
the PV cell as
The proposed MPPT method is used to extract the maximum
∆𝑇𝑇 = 𝐾𝐾𝑒𝑒 . 𝐺𝐺 (12)
power of the PV module under the different shading patterns.
where ∆T is the difference between the atmospheric
An intentional 10% error in the irradiance, temperature and
temperature To and PV cell temperature T. Based on this
PV model is inserted to check the effectiveness of the method
relation, knowing the temperature of a reference cell will be
under any inaccuracy in the estimated irradiance and
sufficient to estimate the temperatures of the rest of the cells.
temperature or in the PV model. The resulted extracted power
This means that measuring the temperature of one PV cell in
is plotted in Fig. 13. As can be seen, the proposed method was
the PV module can determine the temperatures of the rest of
able to extract the global power peaks in all the four scenarios.
the cells. This is represented in the following equation, where
Initially, it extracted the maximum power (around 30W) from
Tco and Gco are the temperature and irradiance of the reference
the PV module operating at the first shading profile. Then, it
PV cell, respectively
jumped from the peak of the black curve in Fig. 10 to the
𝑇𝑇 = 𝑇𝑇𝑐𝑐𝑐𝑐 + 𝐾𝐾𝑒𝑒 . (𝐺𝐺 − 𝐺𝐺𝑐𝑐𝑐𝑐 ) (13)
global peak of the blue curve (around 90W). It then jumped
A. Simulation and Validation back again to the global power of the red curve (around 80W).
This section verifies the effectiveness of the proposed Because the voltage at the global peak of the blue curve
MPPT to accurately find the GMPP for partially shaded PV

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

100

80
Power (W)

60

40

20
(a) (b)
0 Fig. 14 Solar PV array test platform under: (a) unshaded and (b) partially
0 20 40 60 80
shaded conditions.
Time (s)
Fig. 13 The resulted power waveform of the proposed MPPT
2.5 Measured Data
generates low power from the red curve, there is a small drop Estimated MPP
2
in the generated power. Finally, it extracted the global power

Power (W)
of the green curve (around 90W). This demonstrates that the 1.5

proposed method guarantees tracking the GMPP under various 1

shading patterns.
0.5

V. EXPERIMENTAL VERIFICATION OF THE OVERALL 0

0 1 2 3 4 5
IMAGE-BASED MPPT TECHNIQUE
. Voltage (V)

Although the accuracy of the irradiance estimation method (a)

was already experimentally verified, the overall MPPT
method combining the irradiance estimation and the global 0.6
Measured Data
peak estimator is not yet validated. This section aims to Estimated MPP
0.5
validate the proposed method, including irradiance estimation,
Power (W)

0.4
temperature estimation and global peak estimator. A prototype
0.3
composed of eight PV cells, connected with two bypass
0.2
diodes, and a camera is developed to test the effectiveness of
0.1
the proposed technique experimentally under two different
shading patterns as shown in Fig. 14. The electrical 0
0 1 2 3 4
characteristics of the entire PV module are as follows: Isc= Voltage (V)
2.95 A, Im= 2.73 A, Voc=4.96 Vm = 4 V. (b)
Fig. 15 Measured PV curves and estimated GMPP in both shading scenarios.
Before running the irradiance estimation algorithm, the
boundaries of the PV cells in an image must be determined so
that the part of the image representing the PV cells is Then, the reflectance of the PV cells, presented in Fig. 7,
processed while the portion of the image representing was used to estimate the incident irradiances on the different
surrounding background is neglect. The simplest approach, solar cells of the prototype. After estimating the irradiances,
followed in this paper, is the manual calibration during the the temperatures of the solar cells were determined using (13).
setup of the system. An image of the PV module is captured The global peak estimator was then used to find the GMPP.
and then displayed using "imshow" function in MATLAB The resulting GMPPs at shading scenarios were plotted in Fig.
through which the coordinates of each PV cell are manually 15, along with the measured power curves. The measured
determined and then recorded. This can be easily conducted GMPPs are 3.27 V and 1.01 V in the first and second
by clicking on the corners of the PV cells appearing in the scenarios, respectively, while the estimated GMPPs are 3.6V
image to view their pixel coordinates. Once these boundaries and 0.83V. As noticed, the estimated GMPPs lie in the
are determined, they will be used so that the algorithm measured GMPP vicinity and are very close to the exact
considers only the pixels inside the determined coordinates GMPP. The small error can be attributed to the imprecision of
and neglects the rest pixels. It is also worth mentioning that the PV model and irradiance estimation. However, the error
the two metallic lines in PV cells can be kept in the image as can be safely compensated by the P&O method, as illustrated
they have been found to have negligible effect on the accuracy in the previous section. The time needed for the proposed
of the proposed approach. MPPT to find the GMPP, including the time of the image
First, the captured images were mapped through the camera processing as well as the time of the peak estimator equals to
response function, shown in Fig. 4, to find the irradiance 0.1 s, which is less than the required time in the competitive
reflected in the direction of the camera. methods available in the literature (optimization based
methods) which is about 1-2 s using the same computing
device (MATLAB) and the same computer [20].

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

It is worth mentioning that the computation of the algorithm

and interpreting the data from the sensors can be conducted on Normal Operation
the same DSP. Finally, it must be noted that the camera should 150
Under Random
Shading
be placed appropriately to avoid shading the PV module as it
might result in shading the module in some scenarios.

Power (W)
100

VI. ECONOMIC FEASIBILITY OF THE PROPOSED METHOD

Although it was mentioned that a camera would be needed 50
for each micro inverter in the proposed method, it has not yet
been discussed whether this arrangement is economically 0
feasible. In this section, the cost of energy savings gained by 0 5 10 15 20 25 30
Voltage (V)
using the proposed method is quantified and compared to the
Fig. 16 Power-voltage curves of a PV module under normal operating and
cost of the camera and other hardware requirements. random shading conditions.
It might be useful here to review the two sources of power
losses in optimization-based MPPT methods, as these are In addition to the needed camera in the proposed approach,
eliminated in the proposed method. The first is the power loss it requires a camera interfacing hardware, mounting
during the periodic scanning time in which the PV module is arrangement and a temperature sensor. The average cost of the
forced to work on non-maximum power points. The second DSP camera module, which consists of a camera and an
occurs when shading scenario changes between two interfacing hardware to the DSP, is 25$. Mounting hardware
successive scans. In this case, the PV module will be working for a camera could be less than 10$. Temperature sensors are
on the previous GMPP, not the true GMPP, until the next available with less than 5$. This means that the added
scan. This means the PV module will not generate its hardware costs approximately 40$. Adding 10% additional
maximum power after the change in shading scenario until the cost for any extra hardware results in a total cost of 44$.
subsequent scan. These two sources of power losses are Considering this cost and the cost of energy savings, the
eliminated in the proposed method, as it does not require a payback period for the required hardware of the proposed
heuristic scanning and can find the GMPP analytically after method is less than a year and a half.
changes in the shading scenario.
The energy losses due to these two sources in the optimization VII. CONCLUSION
based methods will now be quantified and their cost will be
This paper proposed an innovative MPPT method that was
compared to that of the camera. The P-V curve of the PV
able to track the GMPP of partially shaded PV systems. The
module used in the comparison under a random shading
proposed method utilized the images captured by an optical
scenario is plotted in Fig. 16. In the comparison, it is supposed
camera to estimate the incident irradiances, and then used the
that the periodic scanning in the optimization-based methods
estimated irradiances to find the GMPP analytically via the
occurs every 15 minutes for 2-second durations (typically used
global power peak estimator, which was developed using the
assumptions in the available methods). It is assumed that there
mathematical model of PV modules. The proposed global
would be an abrupt irradiance change every one hour
power peak estimator was also combined with the P&O
throughout a day having 10 hours of illumination (which is
method to correct any inaccuracy in the determined GMPP,
felt a reasonable assumption in cloudy days). An abrupt
which can exist due to any imprecision in the irradiance
change in the irradiance is also expected to occur
estimation. The accuracy of the proposed global peak
approximately in the midst of two successive curve scans
estimator was verified using MATLAB simulation. The
(after 5 minutes of the scanning). Therefore, the system would
proposed estimator was able to track the exact GMPP, even
work on the local point for the rest ten minutes until the next
under an intentional 10% error in the estimated irradiances,
scanning every time the system experience abrupt change in
temperatures, or PV model.
the irradiance.
The proposed irradiance estimation method utilized two
The losses during the scanning time can be calculated by
relations (which must be pre-characterized) to determine the
multiplying the scanning duration (2 sec) by the number of
incident irradiances: the camera response function, and the
scans in a day (40 scans) and fifty percent of the GMPP (165
solar cell reflectance. These two relations were first
W), assuming that the scanning causes a 50% energy loss.
determined in this paper for the camera and solar cell, and
This results in 1.83 W.h. of energy per day. Similarly, the
then utilized to find the incident irradiances. The effectiveness
power losses occurring between the successive scans due to
of the proposed irradiance estimation method was
the changes in shading scenarios can be calculated by
experimentally verified. A further experimental validation was
multiplying the difference between the GMPP power (105 W)
also carried out to verify the effectiveness of the whole image-
and the local peak power (19 W) by the number of times a
based MPPT method under various shading scenarios.
change in shading scenario occurs in a day (10 times) and its
duration (10 min), which is equal to 143.3 W.h. per day. The
price of this energy is $34 per year, calculating the Canadian
feed in tariff at 0.642 $/KWh.

0885-8969 (c) 2016 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.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TEC.2016.2629514, IEEE
Transactions on Energy Conversion
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

REFERENCES [20] K. Ishaque and Z. Salam, "A Deterministic Particle Swarm

Optimization Maximum Power Point Tracker for Photovoltaic System
[1] J. D. Bastidas, E. Franco, G. Petrone, C. A. Ramos-Paja, and G. Under Partial Shading Condition," IEEE Transactions on Industrial
Spagnuolo, "A model of photovoltaic fields in mismatching conditions Electronics, vol. 60, pp. 3195-3206, 2013.
featuring an improved calculation speed," Electric Power Systems [21] S. Roy Chowdhury and H. Saha, "Maximum power point tracking of
Research, vol. 96, pp. 81-90, 3// 2013. partially shaded solar photovoltaic arrays," Solar Energy Materials and
[2] M. Seyedmahmoudian, S. Mekhilef, R. Rahmani, R. Yusof, and E. Renani, Solar Cells, vol. 94, pp. 1441-1447, 9// 2010.
"Analytical Modeling of Partially Shaded Photovoltaic Systems," Energies, [22] T. Kok Soon and S. Mekhilef, "Modified Incremental Conductance
vol. 6, p. 128, 2013. Algorithm for Photovoltaic System Under Partial Shading Conditions and
[3] Y. Mahmoud, M. Abdelwahed, and E. F. El-Saadany, "An Enhanced Load Variation," IEEE Transactions on Industrial Electronics, vol. 61, pp.
MPPT Method Combining Model-Based and Heuristic Techniques," IEEE 5384-5392, 2014.
Transactions on Sustainable Energy, vol. 7, pp. 576-585, 2016. [23] K. Sundareswaran, S. Peddapati, and S. Palani, "MPPT of PV Systems
[4] B. N. Alajmi, K. H. Ahmed, S. J. Finney, and B. W. Williams, "A Under Partial Shaded Conditions Through a Colony of Flashing Fireflies,"
Maximum Power Point Tracking Technique for Partially Shaded IEEE Trans. on Energy Convers., vol. 29, pp. 463-472, 2014.
Photovoltaic Systems in Microgrids," IEEE Transactions on Industrial [24] Y. Hu, W. Cao, J. Wu, B. Ji, and D. Holliday, "Thermography-Based
Electronics, vol. 60, pp. 1596-1606, 2013. Virtual MPPT Scheme for Improving PV Energy Efficiency Under Partial
[5] Ma, x, A. ki, and S. Valkealahti, "Power Losses in Long String and Shading Conditions," IEEE Transactions on Power Electronics, vol. 29,
Parallel-Connected Short Strings of Series-Connected Silicon-Based pp. 5667-5672, 2014.
Photovoltaic Modules Due to Partial Shading Conditions," IEEE [25] E. Trucco and A. Verri, Introductory Techniques for 3-D Computer
Transactions on Energy Conversion, vol. 27, pp. 173-183, 2012. Vision: Prentice Hall PTR, 1998.
[6] G. Petrone, G. Spagnuolo, R. Teodorescu, M. Veerachary, and M. Vitelli, [26] D. Paul E. and M. Jitendra, "Recovering High Dynamic Range
"Reliability Issues in Photovoltaic Power Processing Systems," IEEE Radiance Maps from Photographs," SIGGRAPH, 1997.
Transactions on Industrial Electronics, vol. 55, pp. 2569-2580, 2008. [27] Available: http://www.iesmith.net/tools/solarcalc.html
[7] T. Noguchi, S. Togashi, and R. Nakamoto, "Short-current pulse-based [28] K. Jazayeri, S. Uysal, and M. Jazayeri, "MATLAB/simulink based
maximum-power-point tracking method for multiple photovoltaic-and- simulation of solar incidence angle and the sun's position in the sky with
converter module system," IEEE Transactions on Industrial Electronics, respect to observation points on the Earth," in Renewable Energy Research
vol. 49, pp. 217-223, 2002. and Applications (ICRERA), 2013 International Conference on, 2013, pp.
[8] Y. Chih-Yu, H. Chun-Yu, F. Fu-Kuei, and C. Ke-Horng, "Highly Efficient 173-177.
Analog Maximum Power Point Tracking (AMPPT) in a Photovoltaic [29] Y. Mahmoud and E. F. El-Saadany, "Fast Power-Peaks Estimator for
System," IEEE Transactions on Circuits and Systems I: Regular Papers, Partially Shaded PV Systems," Energy Conversion, IEEE Transactions on,
vol. 59, pp. 1546-1556, 2012. vol. PP, pp. 1-12, 2015.
[9] M. Miyatake, T. Inada, I. Hiratsuka, Z. Hongyan, H. Otsuka, and M. [30] M. L. Orozco-Gutierrez, J. M. Ramirez-Scarpetta, G. Spagnuolo, and C.
Nakano, "Control characteristics of a fibonacci-search-based maximum A. Ramos-Paja, "A technique for mismatched PV array simulation,"
power point tracker when a photovoltaic array is partially shaded," in Renewable Energy, vol. 55, pp. 417-427, 7// 2013.
Power Electronics and Motion Control Conference, 2004. IPEMC 2004. [31] G. Petrone, G. Spagnuolo, and M. Vitelli, "Analytical model of
The 4th International, 2004, pp. 816-821 Vol.2. mismatched photovoltaic fields by means of Lambert W-function," Solar
[10] M. Miyatake, M. Veerachary, F. Toriumi, N. Fujii, and H. Ko, Energy Materials and Solar Cells, vol. 91, pp. 1652-1657, 11/6/ 2007.
"Maximum Power Point Tracking of Multiple Photovoltaic Arrays: A PSO [32] Y. Mahmoud and E. El-Saadany, "Accuracy Improvement of the Ideal
Approach," IEEE Transactions on Aerospace and Electronic Systems, vol. PV Model," IEEE Transactions on Sustainable Energy, vol. 6, pp. 909-911,
47, pp. 367-380, 2011. 2015.
[11] Y.-H. Liu, S.-C. Huang, J.-W. Huang, and W.-C. Liang, "A Particle [33] L. Cristaldi, M. Faifer, M. Rossi, and S. Toscani, "An Improved Model-
Swarm Optimization-Based Maximum Power Point Tracking Algorithm Based Maximum Power Point Tracker for Photovoltaic Panels," IEEE
for PV Systems Operating Under Partially Shaded Conditions," IEEE Transactions on Instrumentation and Measurement, vol. 63, pp. 63-71,
Transactions on Energy Conversion and Management, vol. 27, pp. 1027- 2014.
1035, 2012. Yousef Mahmoud (S’13) received the B.Sc. and
[12] K. Kobayashi, I. Takano, and Y. Sawada, "A study of a two stage M.Sc. degrees in electrical power engineering from
maximum power point tracking control of a photovoltaic system under Albalqa'a Applied University, Amman, Jordan, in
partially shaded insolation conditions," Solar Energy Materials and Solar 2009, and from Masdar Institute of Science and
Cells, vol. 90, pp. 2975-2988, 11/23/ 2006. Technology, United Arab Emirates in 2012. He
[13] G. Carannante, C. Fraddanno, M. Pagano, and L. Piegari, obtained his Ph.D. degree in Electrical and
"Experimental Performance of MPPT Algorithm for Photovoltaic Sources Computer Engineering from University of
Subject to Inhomogeneous Insolation," IEEE Transactions on Industrial Waterloo, Canada in 2016.
Electronics, vol. 56, pp. 4374-4380, 2009. He is currently an assisatnt professor at
[14] H. Patel and V. Agarwal, "Maximum Power Point Tracking Scheme for Worcester Polytechnic Institute, USA. His research
PV Systems Operating Under Partially Shaded Conditions," IEEE interests are mainly in control and operation of
Transactions on Industrial Electronics, vol. 55, pp. 1689-1698, 2008. grid connected photovoltaic power systems including PV modeling,
[15] Y. Liu, "Advance control of photovoltaic converters," Ph.D., Dept. Maximum Power Point Tracking, and Partial shading impact.
Eng., Univ. Leicester, 2009.
[16] T. L. Nguyen and K.-S. Low, "A Global Maximum Power Point Ehab F. El-Saadany (SM’05) was born in Cairo,
Tracking Scheme Employing DIRECT Search Algorithm for Photovoltaic Egypt, on 1964. He received the B.Sc. and M.Sc.
Systems," IEEE Transactions on Industrial Electronics, vol. 57, pp. 3456-
degrees in electrical engineering from Ain Shams
3467, 2010.
University, Cairo, Egypt, in 1986 and 1990,
[17] E. Koutroulis and F. Blaabjerg, "A New Technique for Tracking the
Global Maximum Power Point of PV Arrays Operating Under Partial- respectively, and the Ph.D. degree in electrical
Shading Conditions," IEEE Journal of Photovoltaics, vol. 2, pp. 184-190, engineering from the University of Waterloo,
2012. Waterloo, ON, Canada, in 1998.
[18] M. Boztepe, F. Guinjoan, G. Velasco-Quesada, S. Silvestre, A. Currently, he is a Professor with the Department
Chouder, and E. Karatepe, "Global MPPT Scheme for Photovoltaic String of Electrical and Computer Engineering,
Inverters Based on Restricted Voltage Window Search Algorithm," IEEE University of Waterloo, Waterloo, ON, Canada.
Transactions on Industrial Electronics, vol. 61, pp. 3302-3312, 2014. His research interests include smart grids operation and control, power
[19] K. Sundareswaran, P. Sankar, P. S. R. Nayak, S. P. Simon, and S. quality, distributed generation, power electronics, digital signal processing
Palani, "Enhanced Energy Output From a PV System Under Partial Shaded applications to power systems, and mechatronics.
Conditions Through Artificial Bee Colony," IEEE Transactions on
Sustainable Energy, vol. 6, pp. 198-209, 2015.

0885-8969 (c) 2016 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.