Materials Today: Proceedings 51 (2022) 2131–2138

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

Diagnosis algorithm and detection faults based on fuzzy logic for PV

panel
Marah Bacha ⇑, Amel Terki
LGEB Laboratory, Faculty of Science and Technology, University of Biskra, 07000 Biskra, Algeria

a r t i c l e i n f o a b s t r a c t

Article history: At present, PV systems are among the foremost significant sources of renewable energy. Therefore, it is
Available online 11 January 2022 necessary to diligence and research in detecting and diagnosing the defects that occur, and to identify and
detect failures in PV stations in a short time. So the principle objective of this paper is to present a diag-
Keywords: nosing technique based on detection and classification of photovoltaic panels faults, the analysis have
Diagnosis been achieved by the use of two different algorithms. The first algorithm implements the thresholding
PV panel method and the 2nd one based on Fuzzy logic classifier (sugeno). These two methods have proved to
Faults detection
be able to detect and identify different faults in PV panels accurately and efficiently, our simulation work
Method of thresholding
Fuzzy Logic
was performed using Simulink/Matlab software.
Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the First Maghrebian Con-
ference for Renewable Energies and their applications.

1. Introduction detection method. In Ref [8] the authors applied a method that
combines two algorithms, the ANN and fuzzy logic method for
For industrial PV systems, it is necessary to construct a mentor- detecting short-circuited PV modules and disconnected strings.
ing system that allows detecting, isolating, and even figuring out This paper aims to design a Fuzzy Logic Classifier in Matlab/
any failures that could appear. Therefore, diagnosis of PV systems Simulink software, where the designed model allows to the detec-
is an essential topic of research to enhance production and also tion and classification and localization of eight types of faults
to make the possibility of operating a predictive or a fast corrective occurring in: PV cells, series resistance, shunt resistance, and by-
maintenance after each breakdown issue ensures an improved pass diodes. For that, a model of diagnosis and detection faults will
usage of the installation and less interruptions of the service [1]. be shown. This model simulates both of the normal (healthy) and
There are many faults which affect the performance of the PV the faulty conditions of the PV panel. These operations will be fol-
panel, concerning these faults, environmental condition and panel lowed by the presentation of the two methods are applied for the
degradation are some major causes. The faults that affect the pro- diagnosis of the PV panel. The proposed techniques are based on
duction of the PV panel include diode bypass faults [2], series resis- the analysis of several characteristic quantities of the PV such as
tance augmentation [3] and shading over the panel [4] etc. the power, voltage, and current, our analysis is performed using
Different techniques of fault diagnosis for PV systems were pro- two different Algorithms:
posed in literature. For instance, in Ref. [5,6] the authors suggested
a technique based on artificial neural network (ANN). Two differ-
Algorithm 1 implements the thresholding method for identifies
ent algorithms have been used, the first one is based on threshold- faults which have a different symptoms.
ing method, while the second one is based on an ANN. In Ref. [7] a
Algorithm 2 consists of a fuzzy logic classifier to identify faults
Neuro-Fuzzy classifier (NFC) is used, by extraction the database which have the same combination of symptoms.
from the simulation with Matlab/Simulink software. The method
based on NFC of three types of faults after using the threshold 2. Modeling and simulation

2.1. Photovoltaic module modeling

⇑ Corresponding author.
E-mail addresses: marah.bacha@univ-biskra.dz (M. Bacha), a.terki@univ-biskra. To evaluate the performance of photovoltaic panels under dif-
dz (A. Terki). ferent operating conditions, the single diode model is generally

https://doi.org/10.1016/j.matpr.2021.12.490
2214-7853/Ó 2022 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the First Maghrebian Conference for Renewable Energies and their applications.
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Fig. 1. Solar cell equivalent circuit.

Fig. 2. Schema of photovoltaic module.

considered as the most used model for describe the electrical
behavior of PV cell as shown in Fig. 1 [9].
This model contains Iph current source which represents the Table 1
current generated due to photon interaction. The RS series resistor Electrical characteristics of the SUNTECH PV module.

represents the interconnection between cells. The parallel resis- Electrical characteristics
tance Rsh represents the leakage current. The current generated Pmax: Maximum power 50 W
by the cell is expressed by: Vmp: Voltage at Maximum power 17.4 V
Imp: Current at Maximum power 2.93 A
I ¼ Iph  Id  Ish ð1Þ Voc : Open Circuit Voltage 21.8 V
Isc: Short Circuit Current 3.13 A
The current flowing through the diode Id is given by: The total number of cells connected in series 36
    Number of bypass diodes 2
V þ RS :I
Id ¼ I0 : exp 1 ð2Þ
Vt
(The current flowing through the Rsh is expressed by:

V þ RS :I
Ish ¼ ð3Þ
Rsh
Replacing Eq. (2) and (3) in Eq. ð1Þ, the characteristic equation
becomes:
   
V þ RS :I V þ RS :I
I ¼ Iph  I0 : exp 1  ð4Þ
V t :a Rsh

where V t is defined by:

NS KT
Vt ¼ ð5Þ
q

where:

Iph : Photo-current [A].

Fig. 3. I-V Characteristic of a PV module under STC (25 °C and 1000 W/m2).
I0 : Saturation current [A].
V: Cell voltage [V].
V t : Thermal voltage.
2.3. PV module faults
N S : The total number of cells connected in series.
T: Cell temperature [°K].
Although it is desirable to maintain a regular solar radiation
q: Electron’s charge e = 1.6 * 10–19 C.
over the panel with each cell performing at its Maximum Power
K: Boltzmann constant (1.3854 * 10–23 J°K1).
Point (MPP). In reality, PV panels often experience several abnor-
a: Ideality factor of the junction.
mal conditions that negatively affect their efficiency and the total
output power. In this study, eight faults on PV module are chosen
2.2. PV module characteristics to be achieved as listed in Table 2.

The PV panel used in this work is a SUNTECH PV module, com-

posed of 36 PV cells and two bypass diodes as shown in Fig. 2, with 2.4. Fault Diagnosis PV System
power 50 W of type poly crystalline silicon and its electrical char-
acteristics are presented in Table 1. The simulated diagnosis PV system is composed of two SUN-
Under standard test conditions (STC) (25 °C and 1000 W/m2). TECH PV modules, these modules consist of 36 cells and two
The electrical characteristics I(V) and P(V) curve of PV panel are bypass diodes each. As a first step we simulate the normal PV panel
appeared in Fig. 3 and Fig. 4 using Simulink/Matlab. (healthy) which used as a reference module and the faulty PV panel
2132
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Each fault generates a set of symptoms which are displayed

with a message on the Commande Window illustate the fault type
as shown in Fig. 6. Simulink/Matlab has been used to implement
this configuration illustrated in Fig. 5.

2.4.1. Thresholding method (Algorithm 1)

In this part as shown in Fig. 7, after extraction the three param-
eters (Pmax, Voc, Isc) from each I-V curve (Normal and faulty mod-
ule), the resulting parameters are compared to obtain (Delta Pmax,
Delta Voc, Delta Isc), In the second Step, the obtained parameters
will compared to three relative errors associated to power, voltage
and current, which these errors are related to measurement and
the model errors. From the standard IEC 61724 [10], that indicates
a relative error of 2%, 1%, and 1% while measuring power, voltage,
and current, respectively. The model uncertainty is related to the
industrialization tolerance and sensors noise. The maximum error
due by this uncertainty is calculated, according to [11], by adding a
Fig. 4. P-V Characteristic of a PV module under STC (25 °C and 1000 W/m2). dispersion parameter to the simulation model parameters. The
obtained relative errors associated to power, voltage, and current
are equal to 5%, 3%, and 6%, respectively. The detection of faults
Table 2
is considered effective when these chosen thresholds are exceeded.
Different type of faults chosen for the diagnosis.
After using the threshold method, five groups of faults can be
Symbol Fault type achieved as shown in Table 3:
F1: Shading of one cell in submodule of the panel at 50%. According to this results, the first algorithm cannot discriminate
F2: Shading of one cell in submodule of the panel at 100%. between the faults (F1, F5), and (F2, F7, F8), which have the same
F3: Shading of a cell of the submodule 1 and another of the submodule combination of symptoms. Therefore, to isolate these faults, a very
2 of the panel at 50%.
F4: Shading of a cell of the submodule 1 and another of the submodule
efficient technique of classification is required.
2 of the panel at 100%. The simulation of the previous faults allowed us to obtain dif-
F5: Increase the serie resistors (Rs = 0.09 O) module. ferent curves as shown in Fig. 8, which the outputs of our Simulink
F6: By-pass diode disconnected. model are illustrated in the same figure.
F7: By-pass diode short circuited.
F8: Decrease the shunt resistors (Rp = 0.4 O) module.

The symptom S1: Reduction of maximum power of the PV
module.

The symptom S2: Reduction of Voc of the PV module.
as a tested module for different chosen faults. In the second step,

The symptom S3: Reduction of Isc of the PV module.
for each I-V curve three parameters (Pmax, Voc, Isc) are extracted,
and in the third step, a diagnosis algorithm is used to detect and
classify PV module faults into two groups: 2.4.2. Fuzzy Logic method (Algorithm 2)
To solve this problem A Fuzzy Logic (FL) method will be applied.

Faults with different combination of symptoms. These faults are From Fig. 9, the threshold algorithm block remains the same.
isolated using a signal threshold based approach. Therefore, the modification consists in integrating two diagnosis

Faults with the same combination of symptoms. This type of blocks by Fuzzy Logic Classifier in the system, with (Delta Pmax,
faults are isolated using a Fuzzy Logic Classifier. Delta Voc) as inputs. The 1st FL block works only in the case where
(S1, S2, S3) = (1, 1, 0) and the 2nd block works only in the case
where (S1, S2, S3) = (1, 0, 0). The algorithm used is summarized
in Fig. 11.

Pmax_N
Pmax_N
Isc
Voc_N Voc_N
Voc
Isc_N
Normal PV Module Isc_N
Characteristic plotter 0 0 1
Signature of faults
of the PV Module without fault
Pmax_F
Signature of faults
Pmax_F
Isc
Voc_F
Voc_F
Voc
Isc_F Isc_F
Faulty PV Module
Characteristic plotter
Diagnosis Algorithm
of the PV module with fault
Fig. 5. Diagnosis PV system.

2133
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Fig. 6. Output of fault 5 which is displayed in the command window.

Fig. 7. Diagnosis Model based on Threshold method.

Table 3
The signature of faults after using the threshold method.

Groups Fault type Symptoms [S1,S2,S3]

1 [F1,F5] S1 = 1 S2 = 0 S3 = 0
2 [F2,F7,F8] S1 = 1 S2 = 1 S3 = 0
3 [F3] S1 = 1 S2 = 0 S3 = 1
4 [F4] S1 = 1 S2 = 1 S3 = 1
5 [F6] S1 = 0 S2 = 0 S3 = 0

3.5
Normal
F1
F2
3
F3
S3 F4
2.5 F5
F6
F7
Current [A]

2 F8
S1

1.5

0.5

0
0 5 10 Voltage [V] 15 20 25
S2
Fig. 8. I-V Curves of different type of faults.

To construct a classifier we have to use data mentioned in the Then we have to construct a fuzzy inference base rule IF/THEN,
previous paragraph as inputs. As shown in Fig. 10, in the case 1 the fuzzy rules are chosen to distinguish the defects which have
and 2, fuzzy classifier (FC) starts by fuzzification of these inputs the same indication signature. As presented in Tables 4 and 5, pre-
by using the membership functions. cise bases discriminate between the three faults (in case 1) and the

2134
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Fig. 9. Diagnosis Model after integration of the fuzzy logic.

Fig. 10. Fuzzy Classifier structure and input variables.

two faults (in case 2) have been constructed. Which Table 4 con- 3. Results and discussion
tains 3 rules and Table 5 contains 2 rules.
Finally the values obtained have been defuzzified. This has been This part presents the results of the Simscape based model as
performed by applying the Takagi-Sugeno-Kang type one FL well as the performance of the proposed fault diagnosis technique
method at the output of a FC. Therefore, the outputs membership for the PV module (SUNTECH) system is simulated.
functions are constants. As example, Fig. 12 shows the results given by the technique
used in the case of shading 1 cell at 50%, Fig. 13 shows the results
given in the case of shading 1 cell at 100%.

2135
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Fig. 11. Flowchart of the diagnosis algorithm.

Table 4
Fuzzy Rule base and Defuzzification for Fuzzy Classifier (Case 1).

Rule N° IF THEN Defuzzi-fication

PPM Voc S1 S2 S3
1 PPM_G Vco_P S1 = 0 S2 = 1 S3 = 0 [0 1 0]
2 PPM_M Vco_M S1 = 0 S2 = 1 S3 = 1 [0 1 1]
3 PPM_M Vco_P S1 = 1 S2 = 1 S3 = 0 [1 1 0]

Table 5
Fuzzy Rule base and Defuzzification for Fuzzy Classifier (Case 2).

Rule N° IF THEN Defuzzi-fication

PPM Voc S1 S2 S3
1 PPM_M Vco_P S1 = 1 S2 = 0 S3 = 0 [1 0 0]
2 PPM_G Vco_P S1 = 0 S2 = 0 S3 = 1 [0 0 1]

The different chosen faults are applied in a singular way on the toms send a signal to decision block (shown in Fig. 9) to gives its
faulty PV module, so then the algorithm detects and classifies the accurate location.
fault to different combination of symptoms. The obtained symp-

2136
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Fig. 12. Diagnosis PV system results for F1 fault.

Fig. 13. Diagnosis PV system results for F2 fault.

The results of different fault scenarios for Sugeno FL and thresh- three thresholds associated to power, voltage, and current to
olding method are illustrated in Table 6. For eight different case obtain the values of the outputs for the three symptoms S1, S2,
scenarios have been tested, all different faults have been detected and S3, respectively, which were displayed with the type of fault
except the fault 6, which does not affect in the power generation of applied.
the system. Hence, it can be seen clearly from these obtained After analyzing the simulation results, it can be seen that the
results that these two methods have proved to be able to detect threshold method cannot distinguish all the faults, which leads
and classify and locate different faults in PV panels accurately to apply a more efficient classification technique.
and efficiently. Furthermore, the FL method demonstrated that it is the most
adjusted technique for the diagnosis of PV module. The results
have been attained that all considered faults are detected in a dis-
4. Conclusion
criminable way. As a perspective of this work, we will focus on
improving the diagnostic ability of as many faults as possible and
In this simulation work thresholding and Fuzzy Logic methods
the experimental validation of the technique.
have been used for diagnosis and detection of eight types of faults
in a PV panel. Different attributes (power, voltage and current) of
the normal and the faulty (I-V) characteristics of PV panel have
been compared. The resulting parameters was then compared to
2137
M. Bacha and A. Terki Materials Today: Proceedings 51 (2022) 2131–2138

Table 6
The signatures of each of the symptom for each fault after integration of the fuzzy logic.

CRediT authorship contribution statement [4] S. Sarikh, M. Raoufi, A. Bennouna, B. Ikken, Characteristic curve diagnosis based
on fuzzy classification for a reliable photovoltaic fault monitoring, Sustainable
Energy Technol. Assess. 43 (2021) 100958, https://doi.org/10.1016/j.
Marah Bacha: Conceptualization, Formal analysis, Investiga- seta.2020.100958.
tion, Methodology, Software, Validation, Visualization, Writing – [5] A.A. Djalab, M.M. Rezaoui, L. Mazouz, A. Teta, N. Sabri, Robust Method for
original draft. Amel Terki: Project administration, Supervision, Diagnosis and Detection of Faults in Photovoltaic Systems Using Artificial
Neural Networks, Periodica Polytechnica Electrical Eng. Comput. Sci. 64 (3)
Methodology, Software, Data curation, Writing – review & editing. (2020) 291–302.
[6] W. Chine, A. Mellit, V. Lughi, A. Malek, G. Sulligoi, A. Massi Pavan, A novel fault
Declaration of Competing Interest diagnosis technique for photovoltaic systems based on artificial neural
networks, Renewable Energy 90 (2016) 501–512.
[7] A. Belaout, F. Krim, A. Mellit, Neuro-fuzzy classifier for fault detection and
The authors declare that they have no known competing finan- classification in photovoltaic module, in: 8th International Conference on
cial interests or personal relationships that could have appeared Modelling, Identification and Control (ICMIC), IEEE, 2016, pp. 144–149.
[8] R.G. Vieira, M. Dhimish, F.M. de Araújo, M.I. Guerra, PV Module Fault Detection
to influence the work reported in this paper. Using Combined Artificial Neural Network and Sugeno Fuzzy Logic, Electronics
9 (12) (2020) 2150.
References [9] S. Samara, E. Natsheh, Intelligent PV panels fault diagnosis method based on
NARX network and linguistic fuzzy rule-based systems, Sustainability 12 (5)
(2020) 2011.
[1] T. Andrianajaina, E.J.R. Sambatra, C.B. Andrianirina, T.D. Razafimahefa, N.
[10] International Electrotechnical Commission, Photovoltaic System Performance
Heraud, PV fault detection using the least squares method, in: 2016
Monitoring—Guidelines for Measurement, Data Exchange and Analysis, IEC
International Conference and Exposition on Electrical and Power Engineering
61724 (1998).
(EPE), IEEE, 2016, pp. 846–851.
[11] L. Bun, Détection et Localisation de Défauts pour un Système PV (Faults
[2] M. Dhimish, V. Holmes, B. Mehrdadi, M. Dales, P. Mather, Detecting Defective
Detection and Isolation in a PV System), Université de Grenoble, 2011, PhD
Bypass Diodes in Photovoltaic Modules using Mamdani Fuzzy Logic System,
Thesis.
Global J. Researches Eng.: Electrical Electron. Eng. (2017).
[3] S. Spataru, D. Sera, T. Kerekes, R. Teodorescu, Detection of increased series
losses in PV arrays using Fuzzy Inference Systems, in: 2012 38th IEEE
Photovoltaic Specialists Conference (PVSC), IEEE, 2012, pp. 464–469.

2138