Ensembling techniques in solar panel quality classification

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 13, No. 5, October 2023, pp. 5674~5680
ISSN: 2088-8708, DOI: 10.11591/ijece.v13i5.pp5674-5680  5674
Journal homepage: http://ijece.iaescore.com
Ensembling techniques in solar panel quality classification
Trong Hieu Luu1
, Phan Nguyen Ky Phuc2
, Tran Lam3
, Zhi-qiu Yu4
, Van Tinh Lam5
1
College of Engineering, Can Tho University, Can Tho City, Vietnam
2
School of industrial engineering and management, International University-Vietnam National University, Ho Chi Minh City, Vietnam
3
School of Mechanical Engineering, Hanoi University of Science and Technology, Ha Noi City, Vietnam
4
Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan
5
Department of Electrical Engineering, Bac Lieu Vocational College, Bac Lieu, Vietnam
Article Info ABSTRACT
Article history:
Received Nov 15, 2022
Revised Feb 11, 2023
Accepted Mar 9, 2023
Solar panel quality inspection is a time consuming and costly task. This
study tries to develop as reliable method for evaluating the panels quality by
using ensemble technique based on three machine learning models namely
logistic regression, support vector machine and artificial neural network. The
data in this study came from infrared camera which were captured in dark
room. The panels are supplied with direct current (DC) power while the
infrared camera is located perpendicular with panel surface. Dataset is
divided into four classes where each class represent for a level of damage
percentage. The approach is suitable for systems which has limited resources
as well as number of training images which is very popular in reality. Result
shows that the proposed method performs with the accuracy is higher than
90%.
Keywords:
Electroluminescence image
Ensembling
Image processing
Machine learning
Solar panel quality This is an open access article under the CC BY-SA license.
Corresponding Author:
Phan Nguyen Ky Phuc
School of industrial engineering and management, International University-Vietnam National University
Quarter 6, Linh Trung Ward, Thu Duc District, Ho Chi Minh City, Vietnam
Email: pnkphuc@hcmiu.edu.vn
1. INTRODUCTION
To reduce the negative impacts on climate as well as obtain a sustainable development at the energy
field, many researchers around the world are focusing on finding the new energy resources for substituting
traditional ones which often emit large quantities of CO2 to the environment. In tropical countries like
Vietnam where the number of sunshine hours is about 2,000-2,600 per year, solar energy is an ideal and
#promising renewable energy source which is encouraged by government for expansion and investment [1].
Photovoltaic panels are the most important component of a solar energy system. Under ideal conditions, the
lifecycle of panels typically varies from 25 to 30 years [2]. However, this cycle is significantly shortened due
to several environmental factors such as rain, wind, temperature, and radiation from the sun according to [3].
To limit the undesirable effects of these factors on lifespan of solar panels, material quality improvement
[4], [5] is the approach which is studies by most companies. However, this approach is expensive and require
strict conditions for experiment as well as costly facility. Currently to verify quality of panels and detect the
crack before selling them to the market, the solar industry has applied technique, namely resonance ultrasonic
vibration (RUV) to screen them [6]. According to [7], photovoltaic cell is a p-n semiconductor layer which is
similar to ordinary diodes. The characteristics of these cells include absorbing sunlight and producing
infrared electroluminescence (EL) when they are supplied with direct current (DC) power. Infrared light falls
just outside the visible spectrum but can be observed with proper cameras. According to [8]–[10],
electroluminescence image are valuable data which can be used to predict the conditions of panels. However,
in their studies information of experiment preparation was not fully provided and the approaches differed

Int J Elec & Comp Eng ISSN: 2088-8708 
Ensembling techniques in solar panel quality classification (Trong Hieu Luu)
5675
from deep learning technique. Deep learning network was adopted by [11] to detect defected panels based on
infrared EL images. Nevertheless, method for assessing the failure rate per panel as well as data collections
were not fully described in the study.
According to industry practices, the panel is only replaced when its failure rate exceeds a threshold
value. To make the research can be applied into reality, this study provides detailed descriptions of how the
experiment is prepared and how the proposed study can be applied into the reality when the number of
training images is limited. The backbone of our approach is the ensemble technique which is relied on three
different machine learning models, namely logistic regression, support vector machine and artificial neural
network (ANN). The results show that the proposed method still obtained high performance with
electroluminescent images from other data sets. The rest of this study is organized as: section 2 represents the
experimental setup and proposed method. Data set information and results are provided in section 3 while
conclusion are presented in section 4.
2. METHOD
2.1. Experiment setup
Since the solar panels emit infrared light emission when being powered by DC under the condition
of lacking sunlight, the experiment is setup so that all essential factors are ideal to collect the image. The
experiment was conducted in dark room and isolated from sunlight from of outdoor environment. The
infrared camera with the resolution of 1,920×1,080 pixels is located vertically from the top down at a height
of 45 cm. The panel is a monocrystalline photovoltaic (PV) with a maximum capacity of 10 Wp and
dimensions of 44.5×19 cm. During the experiment, the voltage from the DC linear power supply is adjusted
so that its value lies within the range of 18 to 25 V. The electric current is kept stable around the value of
2 A. The experiment setting is shown in Figure 1.
Figure 1. Experiment setup
Figure 2 shows the images which are captured under different conditions. Figure 2(a) is a photo of a
solar panel which was taken with a non-infrared camera. It is obvious that this photo cannot reveal any
information about damage condition of the panel. Using the proposed experiment setting, Figure 2(b)
provides much more details about the current situation of the panels and can be used as training data. During
the capturing photo process, fixtures are also employed to locate and keep each panel at the same position.
Figure 2(c) shows a typical image of a panel with damaged and functioned photovoltaic cells. In this image,
functioned cells are brighter when the damaged cells include black proximities.
Furthermore, the attenuation rates within a week of these panels compared to a brand-new panel
were also collected. Panels were located on the empty fields without any obstacles so that each panel had the
same absorbing light conditions. Based on the attenuation rate, images are labelled as classes A, B, C and D.
Panels belong to class A having the attenuation rate less than 10%, while in class B, this rate is between 10%
and 20%, in class C between 20 and 30%, and the rest belonging to class D.

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 13, No. 5, October 2023: 5674-5680
5676
(a) (b)
(c)
Figure 2. Images of solar cell (a) by non-infrared camera, (b) by infrared camera with direct current power,
and (c) after affine transformation
2.2. Proposed method
To form the feature vector, the process in Figure 3 was applied. In this study, the Hough
transformation approach in [12] is adopted to find the lines then the affine transformation in [13] is adopted
to project the image. As a result, this process returns the feature vector of size 18×6×2 for each image. It is
noted that, when the number of data is limited, sub cells can be merged into one like pooling method so that
the feature vector size can be reduced. Consequently, the number of parameters in each training model can
also be lower which decrease the probability of overfitting.
In this study, three methods include regression analysis [14], [15], support vector machine (SVM)
[16]–[18] and neural network [19]–[22] are employed. In order to increase the accuracy of the classification
process, the bagging technique of voting method is applied at ensembling stage for making final decision
[23]–[25]. The label is decided by majority rule. If three models give three different output results. The label
with the highest average weight is selected. Three models are applied here including logistics regression,
SVM and neural network. The diagram of the proposed classification system is described as Figure 4.
The advantages of logistics regression and SVM methods are their simplicities and explanability.
After being trained, these models can be deployed easily on any hardware without any special requirements
since the number of parameters of these models are not high. Logistic regression and SVM are also highly
explanable approaches, they can provide more insights about how a class is directly related to the feature
vector. Here the neural network approach is also adopted to ensure the model does not ignore some complex
functions.
Figure 3. Feature vector forming process

5677
Figure 4. The proposed classification system
3. RESULTS AND DISCUSSION
The dataset, which is created by our team, includes 4 classes regarding to 4 respective outputs as
described above. The number of samples for each class label in the training set is show in Table 1. Each
image’s resolution is 1,280×960 pixels and is then transformed to a 18×6×2 feature vector. Only 80% in each
class is used for training, 20% is used for validation.
The neural network in this study comprised of one input layer, two fully connected hidden layer, and
one softmax output layer. This study applied gradient descent method for training each network with a
learning rate of approximately 0.001. The batch size is 30 and the number of epochs is 120. In general, the
learning rate must be selected very carefully to avoid the low learning process as well as the divergence.
Table 2 presents the results of the proposed system in the form of a confusion matrix. Finally, the Table 3
will summarize the performance of each model in the proposed system on the different classes of the test set.
Table 1. The characteristic of the training data set
Class Attenuation rate(R) Number of images
A R<10% 225
B 10%<R<20% 225
C 20%<R<30% 225
D The rest 225
Table 2.The confusion result matrix
Actual label A 220 5 - -
B 5 212 8 -
C - 13 203 9
D - - 12 213
A B C D
Predicted label
Table 3. Summary the performance of each model in the proposed system
Class Total Sample Logistic Regression SVM ANN Sensitivity
A 225 90.5% 97.4% 95.4% 91.2%
B 225 89.1% 97% 94.3% 90.5%
C 225 87.7% 95.4% 93.8% 89.2%
D 225 89.8% 95.2% 91.8% 90.8%
In addition, this study also test the proposed model by pooling 4 adjacent sub cells as one.
Consequently, the feature vector has the size of 9×3×2. Tables 4 and 5 present the confusion matrix and
performance summary of each model regarding to this case.
It can be seen that, at both feature vector sizes of 18×6×2 and 9×3×2, the proposed system has very
high sensitivity for all classes. It can be explained by the high sensitivity of each model. Given a class, if all

 ISSN: 2088-8708
5678
models are assumed as independent to other networks and the sensitivity of model i is 𝑝𝑖, with 3 models here
the sensitivity of whole system when using voting process regarding to probability theory is (1).
𝑃(𝑠𝑦𝑠) = 𝑝1𝑝2𝑝3 + 𝑝1𝑝2(1 − 𝑝3) + 𝑝1(1 − 𝑝2)𝑝3 + (1 − 𝑝1)𝑝2𝑝3 (1)
It can be seen that, this probability is much higher than probability of one model. It is obvious that
the proposed system has not consumed many resources to effectively inspect the solar panel quality. Using
the feature vector size of 9×3×2, the system can be easily employed for most applications without strong
hardware requirement.
It can be seen from the both Tables 3 and 5, in both case of the SVM always outperform two other
methods. Furthermore, the synthesis sensitivity also reduced nearly 5% when being compared with using
only SVM. This phenomena occured due to the low sensitivity of the logistic regression approach. Based on
the results, only SVM is enough for the classification task. Furthermore, it is somehow impossible to increase
the sensitivity especially when the attenuation rate values are near the boundary values of a class. In this
study, these values are 10, 20 and 30%. The wrong classifications occur more when pannels have values near
them.
Table 4. The confusion result matrix
Actual label A 222 3 - -
B 2 219 4 -
C - 3 217 5
D - - 7 218
A B C D
Predicted label
Table 5. Summary the performance of each model in the proposed system
Class Total Sample Logistic Regression SVM ANN Sensitivity
A 225 91.5% 98.4% 97.4% 92.4%
B 225 90.1% 97.1% 95.3% 91.5%
C 225 89.7% 96.4% 94.8% 90.7%
D 225 89.8% 95.5% 92.9% 91.1%
4. CONCLUSION
The study presented an approach to evaluate the quality of solar cells through image processing
combined with machine learning technique. In the proposed approach, solar cells’ electroluminescence
images were captured inside a dark room under 18-25 voltage direct currentpower. The infrared camera is
installed perpendicularly to the cell with one meter distance Images are undergone the affine transformation
and feature engineering technique to extract the most effective features. These features together form a
feature vectors which are used as an input to the training network. In this study, we have not used the
available deep learning network which are most common in the modern approaches. There are many
explanations for this choice. The first one is due to the stability of the environment during the taking photo
process. In reality, setting conditions for taking photo is nearly unchanged so that the ground truth of the
image can be identified easily. Secondly, through division of the whole panel into several rectangle areas, the
useful local features are also extracted and utilized in the classification process. When the numbers of
available images are not too many, employing the deep learning network is not an ideal choice since it
requires lot of resources for training as well as easily create the overfitting phenomenon. However, due to
limitations in data collections, this study has not included photovoltaic cells with long usage time, i.e., more
than 1,000 hours. Study these cells will give the inspection more complete answers.
REFERENCES
[1] Q. Nguyen, “Annual report on electric power generation development scenerios in Vietnam,” Green Innovation and Delelopment
Centre of Vietnam, 2019.
[2] P. Rajput, G. N. Tiwari, O. S. Sastry, B. Bora, and V. Sharma, “Degradation of mono-crystalline photovoltaic modules after 22
years of outdoor exposure in the composite climate of India,” Solar Energy, vol. 135, pp. 786–795, Oct. 2016, doi:
10.1016/j.solener.2016.06.047.
[3] V. Sharma and S. S. Chandel, “Performance and degradation analysis for long term reliability of solar photovoltaic systems: A
review,” Renewable and Sustainable Energy Reviews, vol. 27, pp. 753–767, Nov. 2013, doi: 10.1016/j.rser.2013.07.046.
[4] J. Kim, N. Park, J. S. Yun, S. Huang, M. A. Green, and A. W. Y. Ho-Baillie, “An effective method of predicting perovskite solar
cell lifetime-case study on planar CH3NH3PbI3 and HC(NH2)2PbI3 perovskite solar cells and hole transfer materials of spiro-

5679
OMeTAD and PTAA,” Solar Energy Materials and Solar Cells, vol. 162, pp. 41–46, 2017, doi: 10.1016/j.solmat.2016.12.043.
[5] M. Murayama and T. Mori, “Novel tandem cell structure of dye-sensitized solar cell for improvement in photocurrent,” Thin Solid
Films, vol. 516, no. 9, pp. 2716–2722, Mar. 2008, doi: 10.1016/j.tsf.2007.04.076.
[6] W. Dallas, O. Polupan, and S. Ostapenko, “Resonance ultrasonic vibrations for crack detection in photovoltaic silicon wafers,”
Measurement Science and Technology, vol. 18, no. 3, pp. 852–858, Mar. 2007, doi: 10.1088/0957-0233/18/3/038.
[7] J. L. Gray, “The physics of the solar cell,” in Handbook of Photovoltaic Science and Engineering, Chichester, UK: John Wiley &
Sons, Ltd, 2011, pp. 82–129.
[8] K. Drabczyk, G. Kulesza-Matlak, A. Drygała, M. Szindler, and M. Lipiński, “Electroluminescence imaging for determining the
influence of metallization parameters for solar cell metal contacts,” Solar Energy, vol. 126, pp. 14–21, Mar. 2016, doi:
10.1016/j.solener.2015.12.029.
[9] J. S. Fada, M. A. Hossain, J. L. Braid, S. Yang, T. J. Peshek, and R. H. French, “Electroluminescent image processing and cell
degradation type classification via computer vision and statistical learning methodologies,” in 2017 IEEE 44th Photovoltaic
Specialist Conference (PVSC), Jun. 2017, pp. 3456–3461, doi: 10.1109/PVSC.2017.8366291.
[10] A. Bartler, L. Mauch, B. Yang, M. Reuter, and L. Stoicescu, “Automated detection of solar cell defects with deep learning,” in
2018 26th European Signal Processing Conference (EUSIPCO), 2018, pp. 2035–2039, doi: 10.23919/EUSIPCO.2018.8553025.
[11] N. Q. Minh, L. T. M. Châu, N. Đ. Tiến, and L. M. Hiếu, “Detect PV cell defection based on electroluminescence light using deep
learning,” TNU Journal of Science and Technology, vol. 226, no. 11, pp. 117–123, Jul. 2021.
[12] F. Tarsha-Kurdi, T. Landes, and P. Grussenmeyer, “Hough-transform and extended ransac algorithms for automatic detection of
3d building roof planes from lidar data,” in ISPRS Workshop on Laser Scanning 2007 and SilviLaser 2007, 2007, pp. 407–412.
[13] R. Hartley and A. Zisserman, “Projective geometry and transformations of 2D,” Multiple View Geometry in Computer Vision,
2003.
[14] A. Gelman and J. Hill, Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, 2006.
[15] C. Kwak and A. Clayton-Matthews, “Multinomial logistic regression,” Nursing research, vol. 51, no. 6, pp. 404–410, 2002.
[16] G. van den Burg and P. Groenen, “GenSVM: A generalized multiclass support vector machine,” Journal of Machine Learning
Research, vol. 17, pp. 1–42, 2016.
[17] K.-B. Duan and S. S. Keerthi, “Which is the best multiclass SVM method? An empirical study,” in Multiple Classifier Systems,
Springer Berlin Heidelberg, 2005, pp. 278–285.
[18] H. Chih-Wei and L. Chih-Jen, “A comparison of methods for multiclass support vector machines,” IEEE Transactions on Neural
Networks, vol. 13, no. 2, pp. 415–425, Mar. 2002, doi: 10.1109/72.991427.
[19] K. Crammer and Y. Singer, “On the algorithmic implementation of multiclass kernel-based vector machines,” Journal of machine
learning research, vol. 2, pp. 265–292, 2001.
[20] J. J. Hopfield, “Neural networks and physical systems with emergent collective computational abilities,” in Proceedings of the
National Academy of Sciences, Apr. 1982, vol. 79, no. 8, pp. 2554–2558, doi: 10.1073/pnas.79.8.2554.
[21] D. Ciresan, U. Meier, and J. Schmidhuber, “Multi-column deep neural networks for image classification,” in 2012 IEEE
Conference on Computer Vision and Pattern Recognition, Jun. 2012, pp. 3642–3649, doi: 10.1109/CVPR.2012.6248110.
[22] J. Schmidhuber, “Deep learning in neural networks: An overview,” Neural Networks, vol. 61, pp. 85–117, Jan. 2015, doi:
10.1016/j.neunet.2014.09.003.
[23] D. Opitz and R. Maclin, “Popular ensemble methods: an empirical study,” Journal of Artificial Intelligence Research, vol. 11, pp.
169–198, Aug. 1999, doi: 10.1613/jair.614.
[24] R. Polikar, “Ensemble based systems in decision making,” IEEE Circuits and Systems Magazine, vol. 6, no. 3, pp. 21–45, 2006,
doi: 10.1109/MCAS.2006.1688199.
[25] L. Rokach, “Ensemble-based classifiers,” Artificial Intelligence Review, vol. 33, no. 1–2, pp. 1–39, Feb. 2010, doi:
10.1007/s10462-009-9124-7.
BIOGRAPHIES OF AUTHORS
Trong Hieu Luu received his B.E degree in Mechatronic in 2010 and M.S degree
in Automation and Control in 2016 in Can Tho University. He received the Ph.D degree in
Course of Applied Marine and Environmental Studies from Toyko University of Marine
Science and Technology in 2020. He is currently a lecturer in mechatronics in College of
Technology, Can Tho University. His research of intersts include computer vision, AI with
image processing and robotic technology. He can be contacted at email:
luutronghieu@ctu.edu.vn.
Phan Nguyen Ky Phuc received his B.E. degree in Mechanical in 2008 from Ho-
ChiMinh City University of Technology, and his M.S. and Ph.D degrees in Industrial
Management in 2011 and 2013 from National Taiwan University of Science and Technology.
He is currently professor at Industrial Systems Engineering Department of International
University. His research interests include inventory management, ranking fuzzy numbers,
dynamic programming, meta heuristic algorithm. He can be contacted at email:
pnkphuc@hcmiu.edu.com.

 ISSN: 2088-8708
5680
Tran Lam received his B.E. degree in Mechatronics in 2006 from Hanoi
University of Science and Technology, and his M.S. in Mechanical engineering from National
Taiwan University of Science and Technology (Taiwan) in 2011. He received the Ph.D degree
in Mechanical engineering from Hanoi University of Science and Technology in 2019. He is
currently a lecturer in Mechanical engineering in School of Mechanical Engineering, Hanoi
University of Science and Technology. His research of interests includes mechanical
manufacturing, advanced welding processes and mechatronics. He can be contacted at email:
lam.tran@hust.edu.com.
Zhi-qiu Yu received his B.S.B.A in 2004 from Zhejiang Normal University, He
also holds a master’s degree in business administration and a master’s degree in public
administration from Guangxi University and Graduate School of the Chinese Academy of
Social Sciences. He received a Ph.D degree from National Taiwan University of Science and
Technology in 2022. His research interests include production management, system dynamics,
grey system theory. He can be contacted at email: d10201m01@mail.ntust.edu.tw.
Van Tinh Lam graduated M.S of Electrical Engineering in Vinh Long
University of Technlogy and Education in 2022. He currently works in Bac Lieu Vocational
College. His research of interest is about power electronics and images processing. He can be
contacted at email: lvtinhpro@gmail.com.

Ensembling techniques in solar panel quality classification

More Related Content

Similar to Ensembling techniques in solar panel quality classification (20)

More from IJECEIAES (20)

Recently uploaded (20)

Ensembling techniques in solar panel quality classification