46
TABLE I: PV PANEL CHARACTERISTICS 0.04
PMAX 5W 0.035
VPM 17.5 V
IPM 0.285 A 0.03
VOC 21.3 V
ISC 0.31 A 0.025
Current [A]
0.02
0.015
0.01
0.005
0
0 2 4 6 8 10 12 14 16 18 20
Voltage [V]
0.5
0.45
0.4
0.35
0.3
Power [W]
0.25
0.15
In order to verify the repeatability of the measurement
system, ten complete characterization of a PV panel has 0.1
been performed. The average IV and the P-V 0.05
characteristics and the relative standard deviation (k = 3) are
reported in Figure 4 and Figure 5. It can be noticed that a 0
0 2 4 6 8 10 12 14 16 18 20
good repeatability of the system has been achieved. At the Voltage [V]
Maximum Power Point (MPP, the point of maximum on the Figure 5. V-P characteristic of PV Panel.
V-P curve) the relative standard deviation is 0.13%.
In order to verify the influence of the pollution on the PV
panel performances, a measurement campaign has been TABLE II: CONDITIONS OF EXPOSITION OF THE PV PANELS
carried out with a sample of 10 PV modules exposed to the
TIME OF CONDITIONS OF EXPOSITION
effect of environmental agents. In particular, PV panels have PANEL
EXPOSURE RAIN
ID
been exposed to different environmental conditions in order [days] INCLINATION
EXPOSITION
to consider the effect of the PV module inclination, the 1 34 HORIZONTAL NO
2 34 HORIZONTAL NO
exposition time and the rain as summarized in Table II. 3 34 HORIZONTAL NO
The PV panel has been characterized by using the 4 24 30 YES
measurement setup above described. In particular the I-V 5 24 30 YES
6 34 HORIZONTAL YES
curve of each module has been measured after the 7 34 HORIZONTAL YES
exposition to the environmental agents and after cleaning. 8 21 HORIZONTAL NO
As shown in [4], the effect of the dust can be quantified 9 21 HORIZONTAL NO
10 21 HORIZONTAL NO
comparing the decrease of the Maximum Power (MP)
before and after cleaning. For this reason the acquired I-V
curves have been analyzed in order to compute the MP
value.
47
3. EXPERIMENTAL RESULTS TABLE IV: CONFIDENCE INTERVALS (CI) FOR CLEAN PANELS
MP Clean (W)
0.49
modules in dirty condition and after cleaning are reported.
0.48
C.I. 95%
C.I. 98%
C.I. 99%
TABLE III: MP FOR USED PANELS
0.47
# MPclean MPdirty
MPDecrease %
PANEL [W] [W] 0.46
1 0.4744 0.4573 3.6
2 0.4713 0.4427 6.1 0.45
3 0.4484 0.4181 6.8
4 0.5062 0.4887 3.5 0.44
1 2 3 4 5 6 7 8 9 10
5 0.4700 0.4541 3.4 Panel Id
6 0.4699 0.4545 3.3
7 0.4680 0.4375 6.5 Figure 6. MP Clean Vs Confidence Interval as evaluated in TABLE IV.
8 0.4738 0.4562 3.7
9 0.4781 0.4663 2.5 In Figure 6 it can be observed that panels 4 and 10
10 0.5052 0.4937 2.3 present a MP value higher than the upper limits of the three
confidence intervals. It means that they could show, during
It can be observed that the power variations from the their operating life, better performance with respect to the
clean condition to dirty, for each panel, are bigger than the other elements of the sample. This situation could lead to
measurement system resolution. The power variation can be wrongly evaluate the state of cleanliness of the surface of
evaluated by means of the MP decrease reported in Table III the panels. This assumption is verified, for each of the
according to the formula: defined confidence interval, by the MP experimental values
assumed by the two panels after 24 and 21 days of exposure,
MPclean MPdirty respectively (Table III). On the other hand, panel 3 shows
MPDecrease % = 100 (1) the worst behavior of the samples both in clean and dirty
MPclean
conditions. Its MP value, evaluated in clean condition, is
lower than the lower confidence limit evaluated at 99%
As expected, all the panels present a reduction of the (Table IV).
generated power. It is important to highlight that the The table of exposure conditions (Table II) and the MP
exposure conditions have been randomly assigned to the decrease evaluation of Table III suggest to divide the dirty
panels. In order to avoid the effects of different ageing panels into four classes (as reported in Table V). We
factor due to the different expositions, the MP evaluations in suppose that the population of each class is normally
clean condition have been performed after the dirty one. distributed. For this reason we consider the mean value of
Starting from the previous statement and assuming that the class experimentally evaluated. The mean values of each
the MP values of the clean panels are normally distributed, class and the differences between class values in clean and
it is possible to evaluate the mean and the standard deviation dirty conditions are reported in Table V, where:
of sample results as 0.4765 W and 0.0058 W, respectively.
Starting from this sample character it is possible to define,
MPdirty MPclean
for different confidence levels, the confidence interval for MPdev = (2)
the mean of the population. In Table IV the Confidence MPclean
Intervals (CI) evaluated at 95%, 98% and 99% have been
reported. It can be noticed that classes 1 and 2 can be correctly
classified as dirty: in fact, their MPdirty values do not belong
to the widest confidence interval (99%) defined for the
population of clean panels.
48
TABLE V: PANEL CLASSIFICATION BASED ON FOUR CLASSES OF DIRTINESS Assuming a significance level 1- = 0.80, we have to
Exposition Condition of MPdirty accept or reject the hypothesis H0 evaluating the following
Class Panel MPdev
days exposition [W] test function:
No rain
1 34 1, 2, 3 0.4393 5.47%
horizontal
Rain x
2 34 6, 7 0.4460 4.89% T= (4)
horizontal s n 1
3 24 4, 5 Rain - tilt 30 0.4714 3.43%
No rain
4 21 8, 9, 10 0.4720 2.81%
horizontal where is the mean value of the clean panel, s is the
estimation of the standard deviation and n denotes the
On the contrary, the MPdirty values of the classes 3 and 4 number of samples of the class.
lead to consider the class panels as belonging to the The test of significance has been performed considering
population of clean panels. each of the four defined classes of dirtiness. Four values,
It is interesting to highlight that the differences between which we indicate with ti (i = 1, ...4), can be than computed
class values in clean and dirty conditions (evaluated by applying (4). The test has been performed comparing the ti
means of the values reported in column MPdev of Table V) values with the critical value tn-1,1- corresponding to the
evaluated for classes 3 and 4 are not negligible. The assigned significance level. The tn-1,1- values are - 1.061 for
classification mistake is mainly due to the presence in these the classes with three elements and - 1.376 for the classes
classes of the panels 4 and 10 which have the best with two elements. Thus, the classes 1 and 2 reject the
performances both in clean condition and in their class of hypothesis H0 while classes 3 and 4 accept the hypothesis
dirtiness. This condition is particular critical for class #3 H0. These results are summarized in Table VI.
because only two panels belong to it.
TABLE VI: TEST FUNCTION VALUES AND CRITICAL VALUES
The MPdev value of class 4 is the lowest among those
reported in Table V (it has been evaluated about 3%) even if Class ti tn-1,1-
its MPdirty is very close to that of class 3: this condition can 1 -3.267 -1.061
be explained considering that the panels belonging to the 2 -3.605 -1.376
class 4 of dirtiness have been exposed for 21 days and they 3 -0.297 -1.376
4 -0.401 -1.061
present better performances in clean condition.
In order to evaluate, in a statistical way, the
consideration reported above, it is possible to define a test of It is important to highlight that class 3 and 4 are considered
significance. The statistical hypothesis is related to the clean on the base of the statistical analysis but, from an
assumption: the panels belonging to the four classes of experimental point of view, they can be considered dirty.
dirtiness are clean. In this way we can implement a test Even if a MP reduction less than 4% cannot be considered a
procedure where each class of dirtiness can be associated to strong condition [13], the previous consideration leads to
a decision to accept or reject the aforementioned assert that we are falling in a second type error.
hypothesis. Considering as random the condition of The second type error can be evaluated by means of the
exposure of the panels, the MP values in clean and in dirty cumulative distribution function related to the hypothesis
conditions can be considered random and independent each test:
other. This condition is mandatory for the test construction.
The test is represented by the hypothesis H0 and the
alternative hypothesis H1 related to the dirty population
(
P Z > (xc x ) s / n 1( )) (5)
49
4. DISCUSSION Measurement, vol.60, no.5, pp.1613-1622, May 2011.
[2] Jeevandoss, C.R.; Kumaravel, M.; Kumar, V.J.; , "A Novel
The analysis described in Section 3 shows that by Measurement Method to Determine the CV Characteristic of
properly choosing a reference PV module it is possible to a Solar Photovoltaic Cell," IEEE Transactions on
Instrumentation and Measurement , vol.60, no.5, pp.1761-
define the actual status of a PV plant by means of a 1767, May 2011.
statistical approach. The aforementioned reference panel can [3] Loredana Cristaldi, Marco Faifer, Marco Rossi and
be chosen on the base of: Ferdinanda Ponci, A Simple Photovoltaic Panel Model:
Statistical characterization of a sample of clean Characterization Procedure and Evaluation of the Role of
panels belonging to the lot used in the plant. In this Environmental Measurements, IEEE Transactions on
way the approach reported in the first part of Instrumentation and Measurement, In press.
section 3 should be used. [4] M. Catelani, L. Ciani, L. Cristaldi, M. Faifer, M. Lazzaroni, P.
Rinaldi, FMECA Technique on Photovoltaic Module, Proc.
Statistical characterization of the PV modules
Of IEEE - International Instrumentation And Measurement
provided by the manufacturer. Technology Conference (I2MTC) Binjiang, Hangzhou,
China May 2011, pp. 1717-1722.
The experimental activity reported in Section 3 shows [5] M. Lazzaroni, L. Cristaldi, L. Peretto, P. Rinaldi and M.
that the proposed approach allows identifying different Catelani, Reliability Engineering: Basic Concepts and
functioning classes related to the presence of powder and to Applications in ICT, Springer, ISBN 978-3-642-20982-6, e-
different environmental agents. ISBN 978-3-642-20983-3, DOI: 10.1007/978-3-642-20983-3,
Therefore, by monitoring the value of MP of the 2011 Springer-Verlag, Berlin Heidelberg.
reference PV panel it is possible to infer, with a given [6] Meyer, E.L.; van Dyk, E.E.; Assessing the reliability and
degradation of photovoltaic module performance parameters,
confidence interval, information about the whole PV plant.
IEEE Transactions on Reliability , vol.53, no.1, pp. 83- 92,
In fact, by means of the statistical hypothesis test it is March 2004.
possible to evaluate the significance of the assumption. This [7] P.R. Mishra, J.C. Joshi, Reliability estimation for
approach allows both cost reduction and lower complexity components of photovoltaic systems, Energy Conversion and
of the control system. The latter is possible thanks to a lower Management, Volume 37, Issue 9, September 1996, Pages
number of installed sensors, cables, etc., that are necessary 1371-1382.
for the monitoring task. [8] Powers, L.; Newmiller, J.; Townsend, T.; , "Measuring and
It is important to highlight that this approach, based on modeling the effect of snow on photovoltaic system
the analysis of the MP values of the reference panel, can be performance," Photovoltaic Specialists Conference (PVSC),
2010 35th IEEE , pp.973-978, 20-25 June 2010.
considered independent by the PV module technology.
[9] El-Shobokshy, M.S.; Mujahid, A.; Zakzouk, A.K.M.; ,
"Effects of dust on the performance of concentrator
5. CONCLUSIONS photovoltaic cells," Solid-State and Electron Devices, IEE
Proceedings, vol.132, no.1, pp.5-8, February 1985
In this work, an experimental study investigating the [10] Hamdy K. Elminir, Ahmed E. Ghitas, R.H. Hamid, F. El-
possibility to implement a simplified monitoring system of Hussainy, M.M. Beheary, Khaled M. Abdel-Moneim, Effect
PV plant based on the analysis of a single PV panel in the of dust on the transparent cover of solar collectors, Energy
plant has been presented. The aim is to infer information Conversion and Management, Volume 47, Issues 18-19,
about the plant status in term of its efficiency by monitoring November 2006, Pages 3192-3203.
[11] Monto Mani, Rohit Pillai, Impact of dust on solar
only one PV panel, here called reference panel, and
photovoltaic (PV) performance: Research status, challenges
applying a statistical method. and recommendations, Renewable and Sustainable Energy
In particular the problem related to efficiency decrease Reviews, Volume 14, Issue 9, December 2010, Pages 3124-
due to the presence of dust and pollution on PV panel 3131.
surface has been taken into account. The effects of pollution [12] Hai Jiang, Lin Lu, Ke Sun, Experimental investigation of the
and environmental agents in terms of MP variation have impact of airborne dust deposition on the performance of solar
been analyzed. The analysis has shown that following the photovoltaic (PV) modules, Atmospheric Environment,
proposed approach it is possible to statistically determine Volume 45, Issue 25, August 2011, Pages 4299-4304.
the cleanness condition of a PV plant, with a given [13] L. Cristaldi, M. Faifer, M. Rossi, M. Catelani, L. Ciani, E.
Dovere, S. Jerace, Economical Evaluation of PV System
confidence interval. This allows one to optimize the
Losses Due to the Dust and Pollution, Proc. of I2MTC '12 -
maintenance activity thus reducing its cost and increasing IEEE Instrumentation and Measurement Technology
the overall system efficiency. Conference, Graz, Austria, 13-16 May 2012 , pp: 614 618.
[14] Norman L. Johnson, Fred C. Leone, Statistic and
REFERENCES Experimental Design: in Engineering and the physical
Sciences, John Wiley & Sons, Inc.,New York, 1964.
[1] Adamo, F.; Attivissimo, F.; Di Nisio, A.; Spadavecchia, M.; ,
"Characterization and Testing of a Tool for Photovoltaic
Panel Modeling," IEEE Transactions on Instrumentation and
50