Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

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

Fault detection and diagnosis methods for photovoltaic systems: A review T

a,b,⁎ c d
A. Mellit , G.M. Tina , S.A. Kalogirou
a
Renewable Energy Laboratory, Faculty of Sciences and Technology, University of Jijel, P.O. Box .98, Jijel 18000, Algeria
b
The Abdus Salam International Centre for Theoretical Physics (ICTP), Starada Costiera, 11-34151 Trieste, Italy
c
DIEEI: Department of Electric, Electronic and Computer Engineering, University of Catania, v.le A. Doria 6, 95125 Catania, Italy
d
Department of Mechanical Engineering and Materials Science and Engineering, Cyprus University of Technology, P.O. Box 50329, Limassol 3603, Cyprus

A R T I C LE I N FO A B S T R A C T

Keywords: Faults in any components (modules, connection lines, converters, inverters, etc.) of photovoltaic (PV) systems
Photovoltaic system (stand-alone, grid-connected or hybrid PV systems) can seriously affect the efficiency, energy yield as well as the
Monitoring system security and reliability of the entire PV plant, if not detected and corrected quickly. In addition, if some faults
Fault detection persist (e.g. arc fault, ground fault and line-to-line fault) they can lead to risk of fire. Fault detection and di-
Fault diagnosis
agnosis (FDD) methods are indispensable for the system reliability, operation at high efficiency, and safety of the
PV plant. In this paper, the types and causes of PV systems (PVS) failures are presented, then different methods
proposed in literature for FDD of PVS are reviewed and discussed; particularly faults occurring in PV arrays
(PVA). Special attention is paid to methods that can accurately detect, localise and classify possible faults oc-
curring in a PVA. The advantages and limits of FDD methods in terms of feasibility, complexity, cost-effec-
tiveness and generalisation capability for large-scale integration are highlighted. Based on the reviewed papers,
challenges and recommendations for future research direction are also provided.

1. Introduction
There has been an increased attention to the photovoltaic (PV) en-
ergy systems during the last decade owing to the many advantages that
these systems have such as: it is a worldwide available energy source, it
is pollution free, it has noiseless operation, it is modular and easy to
install, it is a reliable method of energy conversion, and it is able to be
installed and/or integrated in the buildings. As a result, the number and
size of PV systems (PVS) have increased rapidly all over the word. The
PV market grew by 75 GW in the year 2016, while the total capacity has
reached 303 GW around the globe [1]. With reference to IRENA (In-
ternational Renewable Energy Agency) [2] the price of photovoltaic
modules dropped by 80% between 2009 and 2015, and the actual cost
is less than 1 USD/Wp.
Photovoltaic systems are subject to different variety of failures that
can involve all PVS components (modules, cabling, protections, con-
verters and inverters), mainly due to the external operating conditions.
Faults in PVS are caused by: shading effects, module soiling, inverter
failure, and mismatch due to variation in manufacturing or aging of PV
modules (PVM). The main catastrophic failures in PV arrays (PVA) are:
the line-to-line (LLF), ground (GF) and arc (AF) faults [3]. An analysis
of some important failure modes associated to PV modules (PVM),
Balance of System (BOS) and PVA has been given in [4]. Faults in PVS
may cause a huge amount of energy loss as well as risk of fires. For
example, Fig. 1 shows the number of incidents related to fires of various
magnitudes that involved PVS installations in Italy. The Italian data
analysis [5] showed that the number of fires peaked in 2012 following
the first wave of installations. Some recommendations for preventing
the fire hazards in PVS are reported in [5]. Guidelines for the mitigation
of electrical faults that may result in a fire are also given in [6,7].
To ensure reliable and safe operation of PV installations, monitoring
and fault diagnosis (MFD) systems should accompany these installa-
tions to timely detect and solve problems. Addressing these issues,
numerous monitoring and fault diagnosis methods have been studied in
literature, which vary in rapidity, complexity and sensors requirements,
and the capability for the identification of a large number of faults
[8,9].
Fault detection and diagnosis (FDD) for grid-connected photovoltaic
(GGPV) plants, is a fundamental task to protect the components of PVS
(modules, batteries and inverters), particularly PVM, from damage and
to eliminate possible fire risks [6,10]. The main task of fault detection
(FDe), in PVS, consists of comparing the difference between the mea-
sured and calculated parameters with reference values, in order to
verify the occurrence of any fault, while the fault diagnosis (FDi)
method aims to identify the type of faults and localise the faults based
on a priori knowledge or search techniques [11]. Fault localisation

⁎
Corresponding author at: Renewable Energy Laboratory, Faculty of Sciences and Technology, University of Jijel, P.O. Box .98, Jijel 18000, Algeria.
E-mail address: amellit@ictp.it (A. Mellit).

https://doi.org/10.1016/j.rser.2018.03.062
Received 10 February 2017; Received in revised form 14 February 2018; Accepted 17 March 2018
1364-0321/ © 2018 Elsevier Ltd. All rights reserved.
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Nomenclature ASIC Application Specific Integrated Circuit

API Application programmer interface
Terminology MPP Maximum Power Point
MS Monitoring System
ABC-DE Artificial Bee Colony- Deferential Evolution MFD Monitoring and Fault Diagnosis system
AC Arc Fault OCPD OverCurrent Protection Device
AIT Artificial Intelligence Technique PID Potential Induced Degradation
ANN Artificial Neural Network PLA Power Losses Analysis
ANOVA ANalysis Of VAriance PR Power ratio
BBN Bayesian Belief Network PV PhotoVoltaic
BIPV Building Integrated PV PVA PV Array
BkD Blocking Diode PVAF PVA Fault
BOS Balance of System PVM PV Module
BpD Bypass Diode PVMF PVM Fault
DF Diode Fault PVP PV Plant
DS Diagnosis System PVS PV System
EC Eaton Corporation PVStr PV String
ECM Earth Capacitance Measurement RF Radio Frequency;
ETNN Extension Theory with Neural Networks SAPVS Stand-Alone PVS
ENN Extension Neural Networks SPA Signal Processing Approach
EVA Ethylene Vinyl Acetate SPRT Sequential Probability Ratio Test
FDD Fault Detection & Diagnosis SNL Sandia National Laboratories;
FDe Fault Detection SS-PVA Small-Scale PV Array
FDi Fault Diagnosis SS-PVP Small-Scale PV Plant
FDM Fault Detection Method SS-PVS Small-Scale PV System
FFT Fast Fourier Transform STC Standard Test Conditions
FL Fuzzy Logic SVM Support Vector Machine
GA Genetic Algorithm TDR Time-Domain Reflectometry;
GBSSL Graph-Based Semi-Supervised Learning TSKFR Takagi Sugeno Kahn Fuzzy Rule
GS Grid connected VR Voltage Ratio
GCPVS Grid-Connected PV System
GF Ground Fault Symbol
GFD Ground Fault Detectors
GSM Global System Mobile; FF Fill Factor
HS Hot Spot Im Current at MPP
I-VCA I-V Characteristic Analysis Isc Short circuit current
JB Junction Box Rs Series resistance
JBF JB Fault Rsh Shunt resistance
K-NN k-Nearest Neighbour Vm Voltage at MPP
LAPART LAterally Primed Adaptive Resonance Theory Voc Open circuit voltage
LLF Line to Line Fault Ta Air temperature
LS-PVP Large-Scale PV Plant Tc Cell temperature
MCD Minimum Covariance Determinant G Solar irradiance
FPGA Field Programmable Gate Array

remain a big challenge, particularly in large scale PV plants [12]. The
effectiveness of the monitoring and detection systems is also strictly
Fires peaked
related to the architecture of the PVP particularly for small scale PV
plants (SS-PVP) with distributed maximum power point tracking
(DMPPT) [13]. A review on the application of non-electrical methods
(e.g. infrared, thermal imagining) for FDi of PVS is presented in
[14,15]. The most common techniques on image analysis can detect
and localise faults, but they have been applied and verified only for SS-
PVP. A brief review on fault detection and monitoring systems is pub-

lished recently in [16], in which the authors addressed the major PVS
failures.
This paper aims to review the current state of fault detection and
diagnosis (FDD) for PVS based on electrical methods. Different fault
types are reported in this paper by presenting for the concerned ele-
ments (cell, module, string and array), the cause as well as the effects.
The FDD methods presented are discussed in terms of complexity;
ability to detect, identify and locate faults; the response time, and the
2003 2004 2005 2006 20007 2008 2009 2010 2011 2012 2013 2014 generalisation capability so as to be able to be applied for a variety of
Months PV plants. Special attention is given to the FDD methods that can detect
Fig. 1. Fires related to PVS installations, courtesy of Italian National Fire Corp,
and classify accurately the faults occurring on PVA (DC side). Finally,
Statistical Service (INFCS). the advantages and limits or the various methods are presented and

2
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

discussed. Challenges, recommendations and future directions are also
given in this review paper.
This paper is organized as follows: Section 2 discusses the possible
types of faults that can occur in a DC and AC sides of PVS. The main
causes and effects are also presented in this section. In Section 3, FDD
methods for PVS, proposed in literature, are reviewed and discussed.
Finally, some challenges, recommendations and future trends are pro-
vided in Section 4.
2. Types of faults
A PVM failure is usually relevant to the system warranty when it
occurs under conditions which the module normally experiences [17].
Any type of fault that occurs in PVS leads automatically to unexpected
safety hazards, reduced efficiency, power availability, systems relia-
bility and safety. Different kind of defects in PVM are reported and
discussed in [4,18–25]. These include discolouration, cracking, snail
tracks, antireflection coating damage, bubbles, soiling, busbar oxida-
tion and corrosion, and split encapsulation over cells and inter-
connections, back sheet adhesion loss, etc. Some failure modes detec-
tion strategies are discussed in [25], the investigated failure modes are:
encapsulation, module corrosion, cells cracking, and PV inverter.
Generally, faults in PVM can be classified into two main categories:
permanent and temporally. Permanent faults are, for example: dela-
mination, bubbles, yellowing of cells, scratches and burnt cells. So, this
category of faults can be eliminated simply by replacing the faulty
modules. While, temporal faults are basically due to partial shading
effects, dust accumulation (soiling), dirt on PVM, and snow that can be
removed by operators without replacing the faulty PVM. In addition,
the cause of the fault could be external or internal, and both may lead to
a decrease in the output power, efficiency and reliability of the PVS.
The main faults that may occur in a PVS are summarized in Table 1. The
faults include: hotspot, bypass and blocking diodes faults, faults in a
junction box, faults in a PVM, PVA, arc, ground and line-to-line faults.
2.1. Hot spot (HS) fault
Hot Spot (HS) can be caused when some cells in a PV string/array
have different I-V curves [22], i.e., there are variations in I-V the
characteristics of PVMs, high resistance or cold solder points due to
manufacturing processes [22]. In addition, such characteristic may be
affected by soiling and dust accumulation [26–29] (Fig. 2a), degrada-
tion of the cells, incomplete edge isolation [30] by transparent module
materials or by the manufacture's tolerance and the non-uniform in-
solation. The partial shadow effect (Fig. 2b) can be considered as a
particular case of the mismatch fault. The HS phenomenon (Fig. 2c) can
result when the bypass diode of the shaded cells is damaged/dis-
connected, thus its current decreases and its voltage becomes negative,
so the shaded cells consumes power from other non-shaded cells instead
of generating it [31], and if this phenomenon persists the affected solar
cells will be damaged [32] (Fig. 2d). Some methods for detecting HS are
reported in [33], but many techniques for quick detection of HS in a
PVM are based on infrared measurements.
2.2. Diodes fault (DF)
Bypass and blocking diodes (BBD) play a very important role in
maintaining the performance of PVS. The bypass diodes (BpD) are used

Table 1
Summary of different type of faults with connected information: affected components, causes and effects.
Type of fault Affected components Causes Effects

Hot spot (HS) Cells/Module External causes – Damage of solar cells.

– Soiling, dust, snow and shadow – Open circuits (OC)
– PVMs of different classes or technology – Reduce efficiency and reliability
Internal causes
– Fragmentation of cells
– Current mismatch between cells
– High resistance or “cold” solder points
– Aging and degradation of cells
Diode faults (DF) Bypass diode (BpD) or – Partially shaded cells (PS) – Damage diodes
blocking diode (BkD) – Overheating – Short circuited diode, open
circuited or Shunted diode
Junction box fault (JBF) junction box (JB) – Fretting corrosion, – Damage and risk of fire
– Loosen and oxidation – Reduce efficiency and reliability
PV module fault (PVMF) and PVM External causes – Shunted module
PV array fault (PVAF) – Glass breakage of frameless PVMs caused by the clamps – Short circuit (SC) within a
– Connector failure (disconnection) module
– Isolated from ground – SC between PVMs
– Encapsulation – Leakage currents within a PVM
– Wiring mistake at install fault – Damage PVM
Internal causes – Reduce efficiency and reliability
– Corrosion of solar cells – Reduced output power
– Manufacturing defects
– Delaminated, bubbles, yellowing, scratches and burnt cells
Ground fault (GF) PVA or PV String (PVStr) – Insulation failure of cables – Risk of fire
– Incidental short circuit between normal conductor and ground
– GF within PVMs cable insulation damage during the installation
– GF within the PVMs (e.g., degraded sealant and water ingress)
– Insulation damage of cables.
– Accidental short circuit inside the PV combiner box.
Arc fault (AC) PVStr – A short break is created in a conductor – Damage of PVStr
– Two conductors of widely differing voltage are placed near one another – Risk of fire
– Degradation in solder joints, wiring or connections inside the insulation
damage due to mechanical damage, aging, or wild life junction box,
loosening of screws.
Line-to-line fault (LLF) PVA – An unintentional low impedance current path between two points. – Damage PVMs and conductors.
– Insulation failure of cables – Risk of fire
– Incidental short circuit between current carrying conductors
– LLFs within the DC junction box.

3
A. Mellit et al.
Fig. 2. HS fault: causes (a) shading, (b) soiling and dust accumulation; (c) HS damaged on solar cells [19]; (d) detected HS phenomena on a PVM using infrared
equipment.
for reverse voltage protection; while blocking diodes (BkD) are used for
reverse current protection (See Fig. 3a and b). The electrical faults as-
sociated to these diodes are: short circuited diode and open circuited
diode. These faults may occur when a PV module/array is partially
shaded for a long period [34,35]. BpD is a key element for safe system
operation [17], however, BkD in series with PVM will stop Over Current
Protection Devices (OCPD) to operate correctly [36]. The reverse cur-
rent under LLF will be cut off by BkD and the system fails. Fig. 3c shows
brunt diodes in case of fault, by X-ray inspection. To avoid this type of
fault, both the BpD and BkD should be chosen carefully and tested
adequately.
2.3. Junction box fault (JBF)
Junction Box (JB) reliability is one of the most crucial issue for PVM
during verification testing and operation in the field [37]. Fretting
corrosion occurred in JB may lead to a quick increase in contact re-
sistance [38] (See Fig. 4). An electric arc between the contact leads to
wearing out and melting of the JB. This would finally damage the
modules and the whole string, causing the PVS owner further damages
due to loss of energy production. Some actions and suggestions to avoid
JB reliability risk are given in [37].
Fig. 3. Bypass and blocking diodes: a) schematic block, b) junction boxes, c) brunt diodes, X-rays inspection.
4
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
2.4. PV module fault (PVMF)
The faults on the PVM can occur when the array is isolated from the
ground, due to corrosion, delamination of the PVM, leakage currents
within a module and manufacturing defects which may lead to shunted
module and short circuit within a module [19,21,39]. Generally, faults
in PVM may cause electrical shock hazard and fire risk. Some PVM
defects are illustrated in Fig. 5. For example Fig. 5a shows a broken
glass in a PVM, while Fig. 5b, c and d show respectively the oxidation,
delamination bubbles defects that can appear on PVM.
2.5. Ground fault (GF)
A ground fault (GF) in PVA can be considered as an accidental
electrical short circuit involving ground and one or more normally
designated current-carrying conductors [40]. GF in PVA often draw
people's safety concerns because they may generate DC arcs at the fault
point on the GF path. If the fault is not removed properly, the DC arcs
could sustain and cause a fire hazard [6,41].
Identifying ground faults is a significant problem in ungrounded
PVS because such earth faults do not provide sufficient fault currents
for their detection and location during system operation [42]. GF is the
most common fault in PVS and may be caused by the following reasons
[41]: incidental short circuit between normal conductor and ground,
insulation failure of cables, and GF within PVM. They lead to risk of fire
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Fig. 4. a) Fretting corrosion occurred in the JBs and b) brunt bypass diodes JB.

and the overall effect is shown in Fig. 6. Some recommendations for the
isolation monitoring of Ground Fault Detectors (GFD) on residential
and utility-scale PVS have been proposed in [43].

2.6. Arc faults (AF)

An Arc Fault (AF) is the unintended flow of current through air or

another dielectric. AF are generally divided into two categories [44]:

– Series AF: arc from discontinuity in electrical conductor

– Parallel AF: electrical discharge between conductors with different
potentials.
An AF Detector (AFD) should be included in each system. There are
two approaches to detect an AF; the first one is based on the measured
value of the DC current in a conductor, it consists of adding a small
impedance in series with the circuit and measure the resultant voltage.
While, the second is based on the measured value of the AC current in a
conductor, this approach is relatively easy, due to the oscillatory nature
of an AC current, a transformer may be used as the sensing element.
More details about both approaches can be found in [45]. Parallel and
series AF must be de-energized to protect PVS from fires [44]. With
reference to NEC 2011, Article 690.11 [45] a PV system higher than
80 V penetrating a local or a utility network is suggested to integrate an
AF circuit interrupter (AFCI) device as a protection measure.
2.7. Line-to-Line fault (LLF)
A line to line fault (LLF) is an accidental low-resistance connection
between two points of different potential in an electrical network or
system. In PVS, a LLF is usually defined as a short-circuit fault among
PVM or array cables with different potential [46]. LLF in PVA may be
caused by: insulation failure of cables, incidental short circuit between
current carrying conductors, low insulation between string connectors
Fig. 6. A blind spot in a listed inverter's fuse-based ground-fault protection
scheme resulted in this damage. The fire occurred prior to the wide-scale de-
ployment of advanced GD.
in DC string box and mechanical damage. To protect the PVA from LL
incidents, many companies have developed protecting devices.
3. Fault detection and diagnosis methods
Monitoring systems (MS) are crucial for controlling, supervising and
performing fault detection of photovoltaic plants, so many systems have
been recently proposed aiming to perform a real-time monitoring of PV
plants (PVP); in this context the common reference documents are the
standard IEC 61724 [47], titled: Photovoltaic system performance
monitoring – Guidelines for measurement, data exchange and analysis,
and the guidelines of the European Joint Research Centre in Ispra, Italy
[48,49]. The required accuracies and check procedures for data quality
are detailed in [47–49]. Careful considerations should be given to the
purpose behind the monitoring before developing a specific project.
The philosophy should be to measure only those variables that are
necessary using the minimum acquisition rate required to get

Fig. 5. Module defects: a) broken glass, b) oxidation and discoloration [18], c) delamination and d) bubbles.

5
A. Mellit et al.
meaningful results for a period of time over which new information will
be produced. As part of a European Commission-funded project, PV
monitoring guidelines have been recently developed in order to take
into account the system performance over its lifetime [50,51].
Several FDD methods have been proposed in literature, the main
features that can characterise such methods are: to detect the mal-
functions quickly, the input data required (climatic and electrical data),
and selectivity (i.e., ability to distinguish between different faults).
They can be globally classified into two main categories [8]:
– Visual & thermal methods [15,52], which can be used for detecting
discoloration, browning, surface soiling, hot spot, breaking, and
delamination, and
– Electrical methods that can be used for detecting and diagnosing
faulty PVM, strings and arrays including arc fault, grounding fault,
diodes fault, etc.
Most electrical-based FDD methods rely on some type of PVS model
to detect various types of faults. In this section, only electrical methods
will be reviewed and discussed. Electrical methods can be also classified
into five groups [53]:
– statistical and signal processing approaches (SSPA);
– I-V characteristics analysis (I-VCA);
– power losses analysis (PLA);
– voltage and current measurement (VCM);
– artificial intelligence techniques (AIT).
The different FDD electrical methods are summarized in Table 2.
The applications as well as the methods employed for this purpose,
stating from 1993 till 2017 are presented and classified. Some FFD
methods based on electrical techniques are recently reported in [54].
3.1. Methods based on statistical signal and processing approaches (SSPA)
Signal processing methods are mainly based on the analysis of the
waveform signals; for example, Time Domain Reflectometry (TDR),
Speared Spectrum TDR (SSTDR) and Earth Capacitance Measurement
(ECM) are used to detect and localise faulty PV modules.
In [55], TDR technique is employed to determine the position of the
failed PVM in a PVA. The authors highlighted that the method can be
used for fault detection and localisation, however the method is easily
affected by the installation conditions such as modules mounting,
wiring, or PVA components materials.
An electrical method [56,57] is based on the ECM and TDR to detect
which PVM in a string is disconnected. According to the authors, the
ECM can be used to detect the disconnection positions between the
PVM in the string without the effects of the irradiance change, and the
TDR could detect the position of the degradation, such as the increase in
the Rs between the PVs. The work done by Takashim et al. [57] showed
that the ECM method could be applied not only to a PV string consisting
of crystalline Si PVM, but also to a string consisting of amorphous
–Silicon (a-Si) PVM. The TDR technique [58,59] was used to detect
breaks of the circuit, insulation defects, wiring anomalies in strings
including open circuit and inversed polarity. A 1 MW PVP was tested for
a couple of days. The method is basically based on the analysis of the
waveform of the output voltage.
A frequency response analyser was used to measure Radio
Frequency (RF) propagation through arrays of varying irradiance and
size [60,61]. Based on the RF effects in PVS, it is recommended that AF
circuit interrupter manufacturers select detection frequencies below
100 kHz. As reported in [61], Eaton Corporation (EC) and Sandia Na-
tional Laboratories (SNL) have demonstrated the effectiveness of PV arc
detection testing with a wide range of PVS and topologies. TDR tech-
nique is applied to PVStr in order to detect the faults and locate its
position. It has been observed that, the method can clearly detect and
6
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
localise disconnection faults between modules, detect the in-module
open faults but cannot locate clearly their position [62].
A diagnostic method in a DC-DC boost converter and the voltage
source full-bridge inverter, based on monitoring the magnitudes of
some significant measurable frequency components is presented in
[63]. Three faults can be isolated by this method, disconnection of the
positive point of the DC link capacitor, short circuit of the diode of the
boost converter and short circuit of the boost converter switch. Al-
though, with the signal approach used it is not possible to locate se-
parately all the faults, due to the similarity of the signatures and the
practical limits in sensor measurements, the location of a fault be-
longing to a group that has the same signature is still useful for main-
tenance purposes, as the diagnosis is reduced to a group of components,
instead of the examination of the entire PVS [63].
An AF circuit protection technique is proposed in [64]. This tech-
nique can be used to detect and interrupt arcing faults resulting from a
failure in the intended continuity of a conductor, connection, module,
or other system components in the direct current PV source and output
circuits. In [65], the authors developed a FDM for a GCPVS using wa-
velet transform (WT). The merit of the method is the simple calculation
and precise diagnostic capabilities of the FDe. A MS which provides
real-time measurements of each PVM's voltage and current is con-
sidered in [66]. The presented method employs a classical approach to
outlier detection, employing more recent work in robust statistics to
overcome the problem of multiple clustered anomalous observations. It
has been examined only for two types of faults, AF and GF [66]. As
reported in [67], it is possible to detect AF using Fast Fourier Transform
(FFT), but it is not as significant as using WT, particularly when it
comes to the problem for a threshold setting for AF determination. The
proposed method is validated experimentally with good results.
A novel method of monitoring and FDe in PVS is proposed in [68]. It
is based mainly on the analysis of the power losses of the PVS by using
statistical signal processing. The authors introduced a model residual
based on Wireless Sequential Probability Ratio Test (SPRT) framework
for electrical FD in PVS. It has been demonstrated that the results are
influenced by the noise. A methodology based on parameter extraction
techniques for automatic failure detection in PVS has been presented in
[69]. Several failures have been introduced in a real system, including
series resistances (Rs) in connections, current leak to ground, SC of a
battery vessel, and disconnection of a branch of modules. Vergura et al.
[70,71] investigated the use of a statistical approach based on the
analysis of variance (ANOVA) test and non-parametric Kruskale Wallis
(KW) test. The method can detect a disconnection in PVM as well as to
forecast fault in the PVP.
Recently, a statistical method named exponentially weighted
moving average (EWMA) chart is developed in [72], the method is used
to investigate the following faults: short-circuit, open circuit and
shading in PVS. The method has been tested for a GCPV system and the
results are very promising. The same method improved for multivariate
(MEWMA) and it has been applied in [73] for fault detection and
identification in DC side of PVS.
Table 3 summarizes the reviewed methods-based on SPA, with re-
spect to the type of faults investigated, detectability and diagnosis,
complexly integration, capability generalisation and efficiency, ex-
perimental validation and economic cost.
3.2. Methods based on the I-V characteristic analysis (I-VCA)
FDi of PVA based I-V characteristic is firstly introduced in [39], in
which a procedure for the detection of PVA faults is proposed. It con-
sists of comparing the actual to expected electrical parameters from I-V
characteristics. Faulty disconnection in PVA is investigated, the key
elements of the method are PVNode and shadow programs. Experi-
mental testing has demonstrated the applicability of the procedure to
detect some PVA faults. The analysis of the shape of the I-V char-
acteristic of PVA cannot always detect the presence of faults, for
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Table 2
Faults detection and diagnosis methods in PVSs.
Authors Ref. Year Subjects Methods

Schirone and Califano [58,59] 1994 Breaks of the circuit, insulation defects and wiring anomalies in strings Statistical and signal processing
Guasch et al. [69] 2003 Series resistances in connections, current leak to ground, short circuit of a battery vessel, approaches
disconnection of a branch of module
Takashima et al. [55] 2006 Disconnection in PVM in a string and determining the position of failed PVM in a PVA
Takashima et al. [62] 2008 Detection the disconnection faults between modules and the in-module open fault.
Vergura et al. [70,71] 2008 Detect disconnection in PVM and forecast fault in the PVP
2009
Takashima et al. [56,57] 2008 Disconnection in strings, increase series resistance and PVM degradation
2009
Johnson et al. [60,61] 2011 Series AFs in PVMs and cables
González et al. [63] 2011 Detection of disconnection of the positive point of the DC link capacitor. SC of the diode of the
boost converter and short circuit of the switch of the boost converter.
Il-Song [65] 2011 Inverter failures in PV systems
Dini et al. [64] 2011 Detect and interrupt arcing faults resulting from a failure in the intended continuity of a
conductor, connection, module, or other system components
Barun et al., [66] 2012 Arc and ground fault
Davarifar et al. [68] 2013 Monitoring and fault detection in PVs
Wangand and Balog [67] 2013 Arc faults
Garoudja et al. [72] 2017 Short-circuit, open circuit and shading for a DC side PVS
Harrou et al. [73] 2017 Short-circuit, open circuit and shading for a PVS
Stellbogen [39] 1993 Detection faults in module short circuited, module shaded, contact resistance in series, cells in I-V characteristic analysis
a module short circuited.
Chao et al. [78] 2008 Fault diagnosis in PV string
Kaplanis and Kaplani [75] 2011 PVMs degradation
Kang et al. [77] 2012 Partial shading effect
Gokmen et al. [79] 2012 Detect number of open and short circuited PVMs in a string
Tina et al. [8] 2014 Shading effect, breakdown diode, aging soiling delamination
Daliento et al. [81] 2015 Shading effect, BpD fault
Rezgui et al. [83] 2015 SC, OC, impedance and reversed polarity in a PVM
Chine et al. [76] 2015 Connection, shadow and breakdown diodes
Fezzani et al. [80] 2015 Mismatch fault, BpD fault, BkD fault, connectivity fault, module faults and ground fault
Hu et al. [82] 2015 Shading effect on a PVA
Hachana et al. [84,140] 2015, BpD, shading effect, disconnected sub-string, short circuited substring
2016
Wang et al. [85] 2016 Cracked PVMs
Hassan Ali et al. [86] 2017 Faulty PVMs
Zhiqiang et al. [94] 2009 Detection fault section in a branch of PVA Voltage and current measurement
Houssein et al. [97] 2010 Faults in PVMs method
Davarifar et al. [95] 2013 Detection faults occurring in DC side of a PVS
Gokmen et al., [101] 2013 Identification of the number of PVMs presenting a SC or OC in a string regardless of the
number of PVM
Silvestre et al., [99] Fault detection in a GCPVS
Mahendran et al. [100] 2015 Mismatch faults in solar cells
Muñoz et al. [98] 2015 Faults in PVMs
Zhao et al., [96] 2015 Specific normal conditions, LLFs, and OC faults in real-working conditions
Chen and Wang [102] 2017 Fault localisation in a PVA (LLF, GF and shading fault)
Dhar et al. [103] 2017 DC cable and PV series arc faults
Chouder et al. [87] 2010 Faulty PVMs in a string, faulty string, and a group of different faults such as partial shadow, Power losses analysis
ageing, and MPPT error
Shimakage et al. [92] 2011 Faulty PVMs in string and BkD faults
Solórzano and Egido [89] 2013 Fixed object shading, localized dirt, generalized dirt, possible hot-spots, module degradation
and excessive losses in DC cables
Xu et al. [91] 2013 Faults in PVA (short time OC, long time OC, SC, degradation, dust, etc), inverter (MPPT,
energy imbalance and shut down)
Badi et al. [126] 2014 Three–phase balanced faults and unbalanced faults
Stauffer et al. [90] 2015 PV strings faults
Wu et al. [105] 2009 PVAs and power devises Artificial Intelligent Techniques
Chao et al. [112] 2010 Faulty PVMs in branches
Ducange et al. [108] 2011 Detection and diagnosis fault in one broken cell on a string, gradual shading, two broken cells
in two different zones of the same module
Cheng et al., [114] Monitor the status and detect faults for PVMs
Syafaruddin et al. [106] 2011 Identify the SC location of PVMs in one string Independently
Coleman, and Zalewski [107] 2011 Different failure in PVSs (PVMs, inverter, etc.)
Spataru et al. [110] 2012 Increased series losses in PVAs
Chao et al., [115] 2014 Fault diagnosis meter based on a ZigBee wireless sensor network for PVs
Rezgui et al. [109] 2014 SC in a cell, bypass and blocking diodes
Mohamed et al. [113] 2015 Failure operation of the faulty module and failure appeared in the charger or the battery
module
Jones et al. [116] 2015 Single PVM failure in a PVS
Chine and Mellit [9,53] 2015 PVM faults by extracting the intrinsic parameters
Grichting et al. [117] 2015 Arc faults
Spataru et al. [111] 2015 Partial shading effect, increased series-resistance losses, and PID in string PVSs
Chine et al. [53] 2016 SC, inversed BpD, shunted BpD fault or (cell or module), OC and shadow effect
(continued on next page)

7
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Table 2 (continued)

Authors Ref. Year Subjects Methods

Mekki and Mellit [118] 2016 Shading effect (different patterns).

Belaout et al. [119,141] 2016 Bypass and blocking diodes, and increase in series resistance
Zhehan et al. [120] 2017 Line to Line and Line to Ground faults
Karmacharya et al. [42] 2017 Ungrounded PVS
Dhimish et al. [121,122] 2017 Faulty PVM with partial shading effect
Chen et al. [123] 2017 Degradation, open circuit, shunt circuit and shading effect
Yagi et al. [124] 2003 Detection shading effect and inverter stop in a PVS Other methods
Hernández et al. [134,135] 2010 Detection of leakage and insulation current in a PVS
Lin et al., [130] 2012 Degradation faults in a PVM
Hu et al., [137] 2013 PVMs faults
Zhao et al. [129] 2013 LLF in a GCPVS
Chavan et al., [132] 2014 DC-DC converters
Badi et al., [126] 2014 LLF in a GCPVS
Chine et al. [125] 2014 Fault detection in a GCPVS
Jamshidpour et al. [133] 2015 DC-DC converter
Fern et al. [127] 2015 Fault in DC-AC inverter
Wang et al. [131] 2015 Cracked PVMs
Georgijevic et al. [136] 2016 AFs

example the partial shadow of one or two cells in a string cannot be
detected by an appearance of a peak in the I-V characteristic. For this
reason, Miwa et al. [74] have proposed a method based on the analysis
of the (dI/dV)-V characteristic in order to evaluate automatically the
output drop of PVS caused by different loss factors. It was demonstrated
that an appearance of a peak of the (-dI/dV)-V characteristic is effective
to diagnose the power output drop of a PVS.
In [75], the authors calculated the FF, Rs, and Rsh from the light I-V
characteristics, the analysis of such indicators gives an indication of
module degradation. It has been shown that a progressive degradation
of 11% within a period of more than 20 years has been detected. A
method based on the estimation of the key parameters of a PVM is
proposed in [8]. The method allows to detect and locate faults occurred
in PVM by comparing the partially measured I-V with the one esti-
mated. According to the authors, the technique can provide good results
if the irradiance is greater than 500 W/m2, in addition, the calculated
parameters are used not only to detect long term faults (e.g. aging,
soling, delamination, etc.), but also to build an I-V curve reference
when an impromptu fault happens (e.g. shading, breakdown diode, and
so on).
A simple automatic technique for FDi in PVA, based on the analysis
of the anomalies observed on the I-V characteristic was proposed in
[76]. Table 4 reports the various faults investigated. The results de-
monstrate that the proposed technique performs well and accurately
identifies the investigated fault classes. The work done by Kang et al.
[77] presents a novel method to diagnose output power lowering in a
PVA, the method is based on the adjustment of the I-V relationship by
using Kalman-filters. The performance of the proposed diagnosis
method (DM) is evaluated and verified with a laboratory experiment
and field tests. The method can be used for detecting partial shading
faults. A circuit-based simulation model of a photovoltaic module by
PSIM (Software for Power Electronics Simulation) is developed in [78].
Results indicate that the proposed FDM can detect the malfunction
types correctly (95–99%) and promptly. It has been shown that the
method is employed for FDi in PVStr. In [79], the authors proposed a
technique based on the analysis of the relative power of PVStr, which is
defined as the ratio between the actual PVStr power and the power that
the PVStr should generate if there is not any damaged PVM. Such study
aimed to determine the number of open and short-circuited PVM in a
string.
Five common types of faults (mismatch, DF, connectivity, PVMF and
GF) in PVA were investigated in [80]. A Matlab/Simulink based model
is developed for this study. Simulation results show that the designed
model can simulate the different faults investigated. Daliento et al. [81]
developed a novel method to detect failures based on simple electrical
measurements. The authors analysed the first and the second derivation
of I-V curve in order to detect possible faults in RS and BpD. The method
is simulated and validated experimentally, although the applicability of
the method is limited. A low-cost online PVA FDi with optimized vol-
tage sensor locations is introduced by Hu et al. [82]. The authors
claimed that the developed method may increase the output and reduce
the capital and maintenance costs by reducing the number of sensors
and by developing an effective FDi technique. The work done by Rezgui
et al. [83] shows a new algorithm for fault detection in PVM. The de-
signed algorithm is capable to model the PVM into healthy and faulty
operation. It is also able to detect the fault type that affects the PVA
power SC, OC, impedance and reversed polarity.
Hachana et al. [84] developed a diagnosis tool for Building In-
tegrated PV (BIPV) systems, based mainly on a look-up table. The de-
signed tool can be used for detecting possible faults in PVA by analysing
the I-V characteristics. Several fault scenarios have been carried out.
Artificial Bee Colony-Deferential evolution (ABC-DE) technique is em-
ployed for PV parameters extraction. Results showed the capability of
the method for fault diagnosis in BIPV. The authors concluded that, the
system efficiency is increased by using a centralized inverter.
A prototype for FDi in PVMs based on dynamic I-V characteristics
was developed by Wang et al. [85]. Modified Particle Swarm Optimi-
sation (MPSO) technique is employed to extract intrinsic parameters,
and the prototype can be used for detecting the cracked PVM. Ac-
cording to the authors, the merits of the proposed concept lie in the
modularity, scalability, and remote sensing capability without mod-
ifying existing infrastructure. They expected that all electronic com-
ponents will be fabricated inside the module in the future. In Ref. [86] a
method based on real time measuring of I-V curve is proposed. The
method consists of comparing the performance of the faulty module
with its accurate model by quantifying the specific differential residue
that is associated with it.
Table 5 summarizes the methods presented based on I-VCA, with
respect to the type of fault investigated, detectability and diagnosis,
complexly integration, capability generalisation and efficiency, ex-
perimental validation and economic cost.
3.3. Methods based on the power losses analysis (PLA)
An automatic supervision and FDe procedure, based on the PLA is
proposed in [87], which permits to identify three groups of faults and a
false alarm: faulty modules in a string, faulty string, and a group of
different faults such as partial shadow, ageing, and MPPT error. The
automatic supervision method is based on the analysis of the power
losses present in the DC side of the PVA and capture losses (difference
between reference and array yield [88]). Two kinds of capture losses
have been introduced: thermal capture losses and miscellaneous

8
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Table 4

Medium cost as TDR equipment is

List of various types of faults in PVStr [76].

High cost as TDR equipment is

Low-cost, as some sensors are

Types of fault Name

Connected fault Connected resistance between PV modules

Shadow effect with faulty by Shadow effect in a group of cells equipped by a

Relatively high cost

Relatively low-cost
pass diode BpD open circuited

Low-cost system
Economic/cost

Medium cost Shadow effect in a group of cells equipped by a

Medium cost

Medium cost
Medium cost

Medium cost

Medium cost
BpD inversed
required

required

Shadow effect in a group of cells equipped by a

omitted
BpD shunted
Partial shadow effect Shadow effect in the PVMs with normal
–

operation of different components of PVStr

Cell/Module fault Short circuit fault in any cell
Short circuit fault in any PVM
Small, medium and large scale

Inversed module fault

Bypass diode fault Short circuit fault in any BpD
Shunted BpD fault

Medium Scale-PV array

Shadow effect with connected Shadow effect in a group of PVM connected by a
Large size PV plant

fault connected resistance

GCPV system

GCPV system
application

SS-PV array
SS-PV array

SS-PV array

SS-PV array

SS-PV array

capture losses. A procedure for FDi in PVS with distributed MPPT at

PVStr

PVStr

PVAs

module level, power optimizers DC-DC or micro-inverters DC-AC, is

PVM

proposed in [89]. It has been shown that the designed procedure can
diagnose a large scope of failures including: fixed object shading, pos-
Possible integration into low cost system

sible HSs, small localized dirt, module degradation, generalized dirt and
cable losses. The method was experimentally verified.
Possible integration into power

Easy implemented and verified

An easy and cost efficient method for detection faults on PV in-

Easy and possible integration
Easily implemented and test

stallations based on the measured power is proposed in [90]. The

Complexity/ Integration

Easley to be implemented

Verified experimentally

method consists of comparing continuously the measured power with

Easy implementation

the one simulated and then raises a fault flag if a discrepancy is noticed
Easy to implement

(more than 5%). But the method is not able to identify the type of faults
experimentally

experimentally

as well as the location. A simple FDM to check the operating status of

conditioners

the PVS is proposed in [91]. This is done by comparing the simulation

Medium

Medium

parameters (i.e., output direct current and voltage of PV array, output

Easy

alternating current and voltage of inverter) with those measured. Some

–

sensors and several power meters are required for the MS. The proposed
Online/ offline

FDM was verified experimentally.

Shimakage et al. [92] developed a fault detection system by com-
Offline

Offline
Online

Online

Online
Online
Online
Online
Online

Online
Online

Online
Online

paring the present and past conditions in a faulty PVA, and the pro-
posed method was evaluated at specific fault conditions based on the
assumption that some modules are bypassed by the behaviour of a BpD
because of a module fault or a partial shadow on modules in a string.
Temporally or Permanent

Multiple faults can be detected in the algorithm proposed in [93], in

Reviewing papers on the fault detection and diagnosis of PVSs based on SSPA.

which they have used two indicators PR and VR in order to determine

the fault type, time and the location where this fault occurred in the PV
system. The method is based on a statistical analysis of data and the-
Temporally
Both faults
Permanent

Permanent

Permanent
Permanent
Permanent

Permanent
Permanent

Permanent

Permanent

oretical thresholds. The method is not able to detect any fault occurred
in AC side of the system.
fault

Both

Both

Table 6 lists the reviewed methods based on PLA, with respect to the
type of faults investigated, detectability and diagnosis, complexly in-
tegration, capability generalisation and efficiency, experimental vali-
Diagnosis/fault detection

dation and economic cost.

Detection and localisation

Detection and localisation

Detection and localisation

Detection and forecasting

Detection procedure

3.4. Method based on voltage and current measurements (VCM)

In [94] the authors presented a method to identify the approximate

technique

Diagnosis

Diagnosis
Detection

Detection
Detection

Detection

Detection
detection

location of faulty PVM in series or parallel PVA. The analysis of the

experimental results allows the identification of the location of the
faulty PVM. The method can be easily implemented on low cost mi-
[72, 73]

crocontroller for online fault detection in a PVA. Possible faults in DC

[58,59]

[56,57]
[60,61]

[70,71]

side of PVS are identified and classified in [95]. A simplified hybrid

[55]

[63]
[64]
[65]

[67]
[66]

[69]

[68]
Ref.

model of PVM based-MATLAB environment is used. The authors used a

flash test to improve the model behaviour. In [96] the authors devel-
Type of faults

PVMF & GFs

AFs & LLFs

oped a graph-based semi-supervised learning model for fault detection

AF and GF

and classification in PVS. A graph-based semi-supervised learning has

Table 3

PVMF

PVMF

PVMF
PVAF
PVAF

been proposed for possible detection of hidden faults in PVA. Experi-

AFs

AFs

AFs
CIF

CIF

ments demonstrated that the proposed method can correctly detect and

9
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

classify specific normal conditions, LLFs, and OC faults in real-working

conditions. A fault diagnosis technique based on Matlab/Simulink is

Medium cost, sensors are

Low cost and economic

developed in [97]. The method consists of comparing the electrical

Relatively low cost

parameters (Voc, Isc and Pm) in order to detect possible faults in PVM.
Economic/cost

Some measurement at different sites as well as for variable climatic

Medium cost

Medium cost

Medium cost

Medium cost
Medium cost
conditions should be done in order to fix a threshold; the method has

Low cost

Low cost
required

been simulated and verified using Matlab/Simulink.

In [98] a detection and isolation faults method for PVS is designed.
–
–

–
The method doesn’t require any behaviour model and it permits to
detect all possible kind of faults based on signal residuals. In [99] the
Small, medium and large scale

authors define new current and voltage indicators (named: NRc and
NRv) as well as the thresholds for both parameters to identify PV string
and inverter failures. Investigated faults are: faulty string (one string in
open circuit) and bypassed module (one PVM bypassed in one string).
Medium scale PVA

With respect to the authors, the proposed method is simple but effective
application

while considering the minimum number of sensors and minimizes the

PV array

monitoring and supervision system, which can be included in the in-

SS-PVS

SS-PVS
PVStr

PVStr

PVM
PVM
PVA
PVA

PVA
PVA

PVA
PVA

verter. Moreover, the supervision of the PVS could be carried out in real
time by the inverter itself.
to implement in a real-time fault detecting device

A hardware realisation based on Arduino device has been realized

Relatively easy to implement & verified by simulation
Easy to implement, verified by Matlab-Simulink and

for mismatch identification of solar cells [100]. Three parameters have

Possible integration into a data acquisition system

Easy to implement and checked experimentally

Easy to implement and checked experimentally

been measured: voltage, temperature and resistance of the module. The

to implement & validated experimentally
to implement & validated experimentally
to implement & checked experimentally

Easy to implement & checked experimentally

Easy to implement & checked experimentally

method can detect easily the mismatch fault. A method for detecting the
number of open and short circuit faults, and discriminate between them
and partial shading condition is proposed in [101]. The method is based
on the measurement of the operating voltage of PV string and ambient
temperature. A case study for a PVA formed by 8 × 3 PVM was also
also checked experimentally
Complexity /Integration

presented and results showed that the algorithm is able to identify the
Validated experimentally

actual fault of the system with high probability. An adaptive fast fault
localisation method based on the measured power and voltage is pro-
posed in [102]. The method doesn’t require additional equipment. The
investigated faults are LLF, GF and shading. The method can localise a
fault up to PV string level. Recently, a new differential current-based
Easy

Easy
Easy
Easy

fast detection and accurate fault location estimation is proposed [103].

–

The method is developed mainly to detect DC cable faults and PV series

arc fault.
Table 7 lists the reviewed methods-based on VCM, with respect to
Online or

the type of faults investigated, detectability and diagnosis, complexly

offline

Offline

Offline

Offline
Offline

Offline
Offline
Online
Online
Online
Online

Online
Online

Online
Online

integration, capability generalisation and efficiency, experimental va-

Reviewing papers on the fault detection and diagnosis method for PVSs based on I-VCA.

lidation and economic cost.

Temporally or Permanent

3.5. Methods based on artificial intelligence techniques (AIT)

In the last decade artificial intelligence techniques (AIT) have

proven their capability for modelling, control, prediction and fore-
Temporally

Temporally

Temporally
Temporally

Temporally
Permanent

Permanent

Permanent

Permanent

casting in PVS [104]. A few papers are recently published in applica-

tions of ANN for FDD in PVS. For example, the use of ANN technique
fault

Both

Both

Both

Both

Both

was demonstrated for diagnosing faults in a GCPVS [105], five FDi

modes have been examined (i.e., power device cannot be delivered
normally, fault of power device, overheating power devices, too low
Diagnosis/fault detection

value of output voltage and abnormal readings of GC voltage meter). An

ANN is also used for the identification of the SC location of PVM in one
string [106], the proposed method can detect the exact location of SC
condition of PVM in a PVA with good accuracy, it can be used for on-
technique

line system monitoring in the context of smart-grid perspectives.

Diagnosis
Diagnosis
Diagnosis

Diagnosis

Diagnosis

Diagnosis

Diagnosis
Detection

Detection

Detection

Detection
Detection
Detection

Detection

Bayesian Belief Network (BBN) is built to intelligently reason about the

potential causes of the faults detected [107]. The current im-
plementation establishes a framework for validating sensor readings
[84,140]

and mapping them to a BBN using the Netica API. Another method
[39]

[78]
[75]
[77]
[79]

[76]
[80]
[82]
[81]
[83]

[85]
[86]
Ref.

[8]

based on Takagi Sugeno Kahn Fuzzy Rule (TSKFR), was described in

[108]. The method consists of comparing measured and estimated
PVMF, DF &GF
Type of faults

power; a threshold value should be determined. Results showed that the

PVMF, HSF &

PVMF & DF

PVMF & DF

system is able to recognize more than 90% of fault conditions, even in

the presence of noisy data. The investigated faults are: broken cells,
DF
Table 5

PVMF
PVMF
PVMF

PVMF
PVMF
PVAF

PVAF

PVAF

PVAF
PVAF

gradual shading and short circuit.

A fault detection and localisation algorithm is proposed in [109].

10
A. Mellit et al.
Table 6
Reviewing papers on the fault detection and diagnosis methods for PVSs based on PLA.
Type of Ref. Diagnosis/fault Temporally or Online or
faults detection technique Permanent fault offline
PVMF & [87] Detection and Both Online
PVAF supervision
PVMF [89] Diagnosis Both Online
PVAF [90] Detection Both Online
PVAF [91] Detection Both Online
PVMF & DF [92] Fault detection Permanent Online
PVMF [93] Fault detection Temporally Online
The method is used mainly to detect and localise the SCs, BpD and BkD
faults in a PVM. It combines two approaches: the first one is a hybrid
Support Vector Machine (SVM) and the second one is k-Nearest
Neighbour (k-NN) approach aiming to improve the ability of the
method to detect and localise the investigated faults. It has been re-
ported by the authors that a high classification rate with a low error
rate is obtained (0.36–0.55%). In [110], a diagnostic method that is
able to detect an increase in Rs based on FL classifiers is presented and
experimentally validated. Results showed a good detection rate over a
wide range of irradiance levels, and the presence of various sizes and
patterns of partial shadows.
A complete diagnostic methodology for PVStr inverter systems,
which takes advantage of the I-V measurement capability of the string
inverter itself to perform the fault diagnosis was proposed in [111]. FL
is applied for evaluating the diagnostic rules. The designed method can
be used for detecting partial shading, increased series-resistance losses,
and PID in PVS, based on the measurement and analysis of the full I–V
characteristic of the PV string/array by the string inverter. A matter-
element model is combined with an ANN to build an intelligent fault
diagnosis system as shown in [112]. According to the authors, the
proposed fault diagnosis method was adopted to identify the faulty
types of a 3.15 kW PVS. The simulation results indicate that the pro-
posed fault diagnosis method can detect the malfunction types of PVS
rapidly and accurately with lower time and memory consumption.
In [113] the authors presented a method on ANN-based GA to di-
agnose and repair the PVS dynamically. They have shown that, the
proposed method proved how good it is for the practical applications.
The designed approach can be used for detecting the following faults:
SC, OC and degradation in PVM faults. A new approach that can
monitor the status and detect faults for large-scale PVS has been pro-
posed in [114]. A new type of PVA connection was designed, and they
establish its framework for fault diagnosis. A PV cell is used to verify the
effectiveness of the proposed method. It has been shown that the
method has high accuracy and reliability for the fault diagnosis system.
A fault diagnosis meter based on a ZigBee Wireless Sensor Network
(WSN) for PV power generation systems is proposed in [115]. An Ex-
tension Neural Network (ENN) fault diagnosis method is used to iden-
tify whether the PV power generation system is operating normally or a
fault has occurred. The method includes as inputs the solar irradiation
and module temperature of the PVM and then using this information
together with the characteristics captured from the PV power genera-
tion system, provide fault diagnosis, including Pm, Im, Vm and Voc of the
PVA during operation. Investigated faults are reported in Table 8. Ac-
cording to the authors, the proposed method has the advantages of
small learning time (22 iterations), high accuracy, and lower memory
consumption. In addition, a microcontroller and a ZigBee wireless
sensor network are combined to reduce the hardware circuit size and
conduct remote fault diagnostic.
An intelligent technique for FDD based on Laterally Primed
Adaptive Resonance Theory (LAPART) is introduced by Jones et al.
11
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
Complexity/Integration Small, medium and large Economic/cost
scale PV application
Easy to integrate into PV PVA Medium cost
system
Easy to implement & verified PVA Low-cost (as more sensors
experimentally are removed)
Easy to implement & verified PVA Low-cost (only two sensors
experimentally are required)
Easy to implement & tested GCPVS Medium cost (several
experimentally power meters are required)
Easy to implement. GCPVS Low-cost
Easy to implement & verified GCPVS Medium cost
in real time.
[116]. The method can be used to detect automatically the single
module failure in a PVS. The technique has been verified experimen-
tally, and the highest probability of detection was 86%. It has been also
demonstrated that the algorithm was able to interpret both smooth and
intermittent normal behaviour caused by clouds. In [117] the authors
presented a cost effective algorithm based on FL for detecting AF in a
PVS. According to the authors, good results can be expected, when the
detector works together with the DC-AC and not as stand-alone device.
A novel fault diagnostic technique for PVA based on ANN is pro-
posed recently in [9,53]. Table 9 reports the investigated faults. The
analysis is performed using two different Algorithms:
– Algorithm 1, implements a signal threshold approach and isolates
the faults that have a different combination of attributes;
– Algorithm 2, consists of an ANN based approach and detects the
faults that are characterised by the same combination of attributes.
It has been demonstrated that the designed technique is able to
detect and identify accurately the investigated fault categories in the PV
string, using only the parameters of the I-V characteristic as well as the
irradiance level and cell temperature.
A fault detection method for photovoltaic module under partially
shaded conditions is introduced in [118]. It uses an ANN in order to
estimate the output photovoltaic current and voltage under variable
working conditions. The results confirm the ability of the technique to
correctly localise and identify the different types of faults. The designed
diagnostic method is cheap because it requires as input only the fol-
lowing parameters: solar irradiance, PV module's temperature, and PV
array's current and voltage [118]. In [119] the authors developed a
fuzzy logic technique for fault detection in a PVA. The designed algo-
rithm is able to discriminate between the most frequently occurring
PVM module faults, such as increased series losses, BpD and BkD with
good accuracy (90–98%).
A method to detect line-to-line and line-to-ground fault, mainly
based on the application of a multi-resolution signal decomposition
(MSD) technique on fuzzy inference system is developed in [120]. Re-
sults show that the method is able to detect faults in a PV array, and it
was demonstrated experimentally for a SS-PVA. In [42] a fault detec-
tion method based on WT and ANN is developed for an ungrounded PV
system. The designed method is able to detect and localise GF and LL
faults in a PVA. Finally, the effectiveness of the designed fault locator is
tested with a variety of system parameters. The results demonstrate that
the proposed approach has accurate and robust performance even with
noisy measurements and changes in operating conditions. A method
based on the theoretical I-V curves analysis and FL classification system
for fault detection in DC-side of a 1.1kWp GCPV system is developed in
[121,122]. The investigated fault is partial shading effect in PV mod-
ules. The classification rate is more than 98%. Recently, the authors in
[123] developed a novel fault diagnosis approach to detect and classify
the following faults: degradation, open circuit, short circuit and partial
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

shading effect on a PVA. The approach is based on the use of I-V curves
and the emerging kernel based extreme learning machine. With re-

Low cost (few sensors are

ference to the authors, both the simulation and experimental results
show that the designed approach can achieve high accuracy.
Table 10 lists the reviewed methods based on AIT, with respect to
Economic/cost

Medium cost

Medium cost

Medium cost
the type of faults investigated, detectability and diagnosis, complexly

required)
integration, capability generalisation and efficiency, experimental va-
Low cost

Low cost

Low cost

Low cost
lidation and economic cost.

–
–
3.6. Other techniques
Small, medium and large scale PV

Some publications are focused on the identification of faults oc-

curring in both the DC and AC side of PVS. In [124] a learning method
is developed to identify two types of faults: shading effect and inverter
failure. The technique enables classification of energy losses in four
different categories: sustained zero efficiency faults, brief zero effi-
ciency faults, shading, and non-zero efficiency non-shading fault. The
application

absolute error on the performance ratio is calculated and compared

SS-PVA

SS-PVA

SS-PVA

SS-PVA
SS-PVA
SS-PVP

SS-PVS
GCPVS

with a threshold in order to detect the fault [125], and then three in-
PVAs

PVM

dicators are defined to determine the type of fault: the DC-AC power
ratio and the ratio between the simulated and the measured DC output
Verified by simulation with low complexity
Easy to implement, tested experimentally

Easy to implement, tested experimentally

current and voltage. An automatic system is developed graphically

Relatively easy to implement, checked

under MatLab/GUI to display and monitor instantaneously the PVP. In

Possible integration into low cost

[126] the authors presented a methodology for determining the un-

Easy to implement and verified

balanced fault in GCPV systems using Matlab/Simulink. It was observed

Complexity / integration

that double line to ground fault produces more severe effects on the
GCPV system than a single line–to–ground fault. In addition, it was
Easy to implement
Easy to implement

found that faults existence in the system dipped the voltage which re-
microcontroller

experimentally

experimentally

sults in a low voltage during fault condition and in a real-life situation.

A FDD method for open switch fault in a T-type three level inverter
is presented by Fern et al. [127]. It is based on the analysis of the AC
Easy

voltage patterns. The results showed that several possible open switch
–

faults can be detected accurately. Ancuta and Cepisca [128] in-

vestigated a non-destructive method based on infrared (IR) thermo-
graphy for observing hot spot formation early. The current paper in-
Online or

vestigates a non-destructive method based on infrared thermography

offline

Offline
Offline

Offline
Offline
Online

Online

Online

Online

Online

Online

with the help of which solar modules measurements were made. HS

invisible to naked eye are put into evidence and local, dust accumulated
particles where detected and analysed. In [36,129] the authors, in-
Temporally or permanent

Permanent and Temporally

Reviewing papers on the fault detection and diagnosis of PVSs based on VCM method.

vestigated a line and LLF with zero fault impedance in PVA by taking
into account the overcurrent protection devices. It is noted that a fault
may remain undetected and be hidden in the PVA until the whole
system fails. Chao et al. [41] proposed a modelling and fault diagnosis
based on the extended correlation function and the matter element
Temporally

Temporally

Permanent

Permanent

Permanent

model. According to the authors, the system operators probably need

fault

extra work to find the module that experiences fault conditions and this
Both
Both

Both
Both

is not so easy and it is dangerous in large size PVA. In [130] the authors
Fault detection and classification

Fault detection and classification

presented a modified PVM structure and an efficient method for the

Fault detection and isolation

online FDe and tolerance. It has been shown that the proposed FDe and
Diagnosis/fault detection

tolerance technique reduces output power degradation due to PV cell

faults in a PVS by up to 81.31%.
Wang et al. [131] designed a prototype device for fault diagnosis
Fault diagnosis
Fault detection

Fault detection

PVM. The device can be used for extracting the intrinsic parameters of
Localisation
technique

PVM, which is a crucial issue. It has been shown that the intrinsic
Detection
Detection

Detection

parameters, such as the output resistance, the reverse saturation cur-

rent, and the junction capacitance, of the damaged PVM may sig-
nificantly deviate from the nominal values. A model based approach for
[101]

[100]

[102]
[103]
[94]

[99]
[95]

[96]

[98]

[97]
Ref.

FDi in full bridge DC-DC converters, used for PV applications, is re-

ported in [132]. The designed approach can be used for detecting the
DC cable and PV series

fault in the open circuit power switch. As a conclusion, the authors

recommended the use of a high-speed controller for better accuracy and
LL, LG and PVM
Type of faults

quick fault diagnosis. A very fast FDM based on the shape of the in-
ductor current associated to fault tolerant operation for boost converter
used in PVS, is proposed in [133]. The authors pointed that the current
AF
Table 7

PVMF
PVMF

PVMF
PMVF
PVAF

PVAF

PVAF

sensor which is considered for control loops is used also for FDM, thus
LLF

no additional sensor is required. This approach can be applied on a

12
A. Mellit et al. Renewable and Sustainable Energy Reviews 91 (2018) 1–17

Table 8
Faults type and condition [115].
Fault type Fault condition

PF1 Normal operation

PF2 Two series branches a fault has occurred in a module in either of the two branches
PF3 Two series branches a fault has occurred in two modules in one branch
PF4 Two series branches a fault has occurred in three modules in one branch
PF5 Two series branches a fault has occurred in one module in each branch
PF6 Two series branches a fault has occurred in two modules in each branch
PF7 Two series branches a fault has occurred in three modules in each branch
PF8 Two series branches a fault has occurred in one module, one branch and two modules in the other branch
PF9 Two series branches a fault has occurred in one module, one branch and three modules in the other branch
PF10 Two series branches a fault has occurred in two modules, one branch and three modules in the other branch

Table 9 and cost implementation, but suffer from the lack of fault tolerance,
Investigated faults [9]. feasibility to implement in commercial monitoring system and
Fault Symbol
possibility of integration in the LS-PV plant.
– FDi based electrical methods are more advantageous and promising
- Short-circuited BpD F1 for the monitoring and diagnosis of PVS, while non-electrical
- Inverted BpD F2 methods (like visual, IR camera, etc.) are much suitable for SS-PV
- Shunted BpD F3
plants.
- Increased Rs F4
- Partial shadow while the BpDs work correctly F5 – FDi techniques differ in term of speed to detect the malfunctions,
- Partial shadow when a BpD diode is faulted F6 data requirement, and ability to distinguish among different faults.
- A string disconnection F7 – Most available FDD methods work offline and tested only for SS-PV
- A fault in the JB F8
plants. In addition, some of them cannot accurately identify the
cause of the fault or its location.
– The main advantages of the applications of FDD-based AIT in PVS
system with several parallel DC-DC converters connected to a common
are that they can be used for the on-line system monitoring and they
DC bus. In this case only one redundant switch is needed for the whole
don’t require any additional efforts.
system. In [134,135] the authors proposed a method for the calculation
– Methods-based on ANN and FL are able to distinguish between faults
of the PV generator's insulation resistance and leakage current. This
that have the same signature and can classify possible faults.
model can be employed to analyse the risk of electric shock and for the
However, the main drawbacks are: they require more advanced
design of system protections.
skills with respect to the implementation in real time (experimental
A detector prototype for series AF is designed in [136] and its op-
realisation) and databases (of different faults), which are not always
eration was experimentally tested in a GC 1.6 kW PVS, without any
available.
prior analysis of the target PVS or any additional algorithm parameter
– The main advantage of automatic FDi method based power losses
adjustment. A parameter model for FDi of PVM based on the measured
and I-VCA is that they usually have low hardware requirements and
current and voltage characteristic and module temperature using a
they can detect practically all faults that can appear in the PVA.
thermal camera, is proposed in [137]. The model is validated experi-
– Methods based on SASP can also accurately detect and localise the
mentally by measuring the electrical faulty parameters of PVM and
faults, and they are most suitable for SS-PV plants. However, they
comparing them with the ones estimated by the model. The results
require some expensive equipment.
showed the capability of the method for fault diagnosis in PVM. A cost
– It is important to point out that electronics instruments have been
effective FDM in PVS was proposed in [138]. A model-based state is
proposed for fault detection, especially for arc fault, and grounding
developed for identifying faults (e.g. PVM, switch and sensor faults).
faults. These faults are integrated based on statistical and numerical
The method is validated experimentally and precise fault identification
methods.
was done within 31 ms. Firth et al. [139] have used empirical models
– So far, the available commercialized components concern the fault
for failure detection in PVS. The techniques distinguish four categories
protection of PVS, especially from ground, arc and line-to-line faults,
of faults: sustained zero efficiency faults; brief zero efficiency faults;
but most of them are not able to diagnose and locate faults occurring
shading faults; and non-zero efficiency non-shading faults. However,
in PVS.
they noticed that it was not possible to identify all of the seven types of
faults individually from the monitored data and so the faults were
Challenging issues in fault diagnosis of photovoltaic systems are:
placed into four fault categories, which are based on the duration of
faults and partial shading faults by monitoring the efficiency of the
– Cost effective hardware/software implementation and ability of the
system.
designed MS to integrate advanced FDD methods, especially the
ones based on the application of AIT.
4. Concluding remarks
– Appropriate fault detection and diagnosis techniques, which should
be characterised by their efficiency, simplicity in terms of im-
In this paper, an overview of fault detection and diagnosis methods
plementation, fast fault detection and diagnosis algorithms, cap-
of PVS are reviewed. To increase the competitiveness of PV in the en-
ability generalisation for LS-PVP, adaptability for different PV
ergy market, their reliability, efficiency and operating cost have be-
technologies, fault classification, ability to identify multiple faults
come crucial factors. As shown in this paper many methods for fault
and ability to detect new faults.
detection and diagnosis have been investigated in order to solve this
– Emerging faults are the specific faults which are more challenging to
issue. With reference to the aforementioned techniques, the following
detect. These emerging faults cause major problems because they
points can be highlighted:
can remain undetected by existing equipment.
– The effectiveness of a monitoring /faults detection system is related
– MS have some advantages and limitations in terms of complexity

13
 A. Mellit et al.

Table 10
Reviewing papers on the fault detection and diagnosis of PVSs based on AITs.
Type of faults Ref. Diagnosis/fault detection Temporally or permanent Online or Complexity / Integration Small, medium and large scale PV Economic/cost
technique fault offline application

CIF [105] Fault diagnosis Temporally Offline Relatively easily to implement SS-GCPVS Medium cost
PVAF [112] Fault detection Temporary Offline Relatively easily to implement PVA –
PVAF [106] Detection and localisation Permanent Offline Relatively complex SS-PVA Medium cost
PVAF &CIF [107] Diagnosis Both fault Online Complex MS-PVS High cost
PVMF [108] Detection Both fault Offline Relatively complex SS-PVS –
PVMF [110] Detection Temporary Offline Easy to implement SS-PVA Medium cost
PVMF &DF [109] Detect and classify Permanent Offline Relatively easy to implement, fast, and low computation SS-PVA Medium cost
PVMF & CIF [111] Detect and classify Both Online Relatively easy to Implement, fast, and low computation. SS-PVA Medium cost
PVMF [113] Detect and analysis Both Offline Easy to implement and tested under Matlab/Simulink PVM Medium cost
PVAF [116] Detection Temporally fault Offline Relatively easy to implement & verified based experimental PVStr Medium cost
data.

14
AF [117] Detection Permanent Online Easy to implement, tested under Matlab/Simulink SS-PVP Low cost
PVAF & DF [9] Detect and diagnosis Both Offline Relatively easy to implement. Validated and tested based on PVstr Medium cost
measured I-V curves
PVMF Detection Temporally Offline Easy to implement. Tested and validated under Matlab/ PVM Low cost
Simulink
PVMF [114] Diagnosis Both Offline Relatively easy to implement PVM Low cost
PVAF [115] Diagnosis Both Online Relatively easy. Validated and tested experimentally PVA Low cost
PVMF [118,133] Detection Temporally Offline Relatively easy to implement. Validated under Matlab/ PVA –
Simulink
PVMF [119] Detection Temporally Offline Relatively easily to implement PVA –
LL, Line to ground [120] Detection Permanent Online Relatively easy. Validated and tested experimentally SS-PVA Medium cost
GF, PVA [42] Localisation Permanent Online Relatively easy to implement. Validated under Matlab/ SS-PVA –
Simulink
PVA [121,122] Detection Both Online Relatively easy to be implemented & verified SS-PVA –
experimentally
PVA [123] Detect & classify Both Online Relatively easy to implement. Validated under simulation SS-PVA Medium cost
with experimental data
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
A. Mellit et al.
to the precision of measurement; the inverters normally do not
provide very precise measurements so extra sensors are required.
– Decentralized control for fault tolerance and mitigation based on
fast diagnostics.
– Database collection for different faults that can occur in LS-PVP.
Recommendations and future directions:
– The experimental implementation of fault diagnosis techniques into
low-cost integrated circuit is necessary for real time applications,
where the input/output data of the adopted techniques can be
managed via internet in order to reduce the computational and
transmission costs.
– Integration into reconfigurable hardware such as reconfigurable
FPGA, for rapid prototyping of the advanced fault detection algo-
rithms and diagnosis techniques may improve the reliability of the
fault diagnosis techniques in terms of rapidity, flexibility and ro-
bustness.
– Large-scale integration of the designed methods and experimental
validation, by taking into account the compromise between cost and
complexity when the appropriate method is chosen.
– The firmware of inverter should be modified in such a way to per-
form also functions connected with diagnosis (such as on-line
measurements of I-V curves).
– Data mining algorithms able to analyse long term PVS operation
data could be useful in diagnostics specially to detect aging phe-
nomena.
– Using Internet of Things (IoT) technology for remote sensing and
monitoring of LS-PVP.
Acknowledgements
The reprint of the various figures used in this manuscript is done
with the agreement of the copyright owner (Elsevier numbers:
4280201190652, 4280210861539).
References
[1] IEA. A snapshot of global Photovoltaic Markets. 〈http://www.iea-pvps.org〉.
[Accessed March 2017].
[2] Taylor M, Ralon, Ilas A. The power to change: solar and wind cost reduction po-
tential to 2025. Int Renew Energy Agency (IRENA) 2016 [Accessed June 2016].
http://www.irena.org/.
[3] Alam MK, Khan F, Johnson J, Flicker J. A comprehensive review of catastrophic
faults in PV arrays: types, detection, and mitigation techniques. IEEE J Photovolt
2015;5:982–97.
[4] Colli A. Failure mode and effect analysis for photovoltaic systems. Renew Sustain
Energy Rev 2015;50:804–9.
[5] Fiorentini L, Danzi E, Puccia V. Fires in Photovoltaic Systems: Lessons Learned
from Fire Investigations in Italy; 2016. 〈http://www.sfpe.org/?Page=FPE_ET_
Issue_99〉. [Assessed May].
[6] Falvo MC, Capparella S. Safety issues in PV systems: design choices for a secure
fault detection and for preventing fire risk. Case Stud Fire Saf 2015;3:1–16.
[7] Brooks W. Field guide for testing existing photovoltaic systems for ground faults
and installing equipment to mitigate fire hazards. NREL National Renewable
Energy Laboratory (NREL); 2015.
[8] Tina GM, Cosentino F, Ventura C. Monitoring and diagnostics of photovoltaic
power plants. In: Sayigh A, editor. Renewable Energy in the Service of Mankind,
Vol II. Cham: Springer; 2016. p. 505–16.
[9] Chine W, Mellit A, Lughi V, Malek A, Sulligoi G, Pavan AM. A novel fault diagnosis
technique for photovoltaic systems based on artificial neural networks. Renew
Energy 2016;90:501–12.
[10] Krueger J, Smith D. A practical approach to fire hazard analysis for offshore
structures. J Hazard Mater 2003;104:107–22.
[11] Al-Sheikh H, Moubayed N. Fault detection and diagnosis of renewable energy
systems: An overview. In: Proceedings of the IEEE International Conference on
Renewable Energies for Developing Countries (REDEC). 2012. p. 1–7.
[12] Hasenfus GD. Smart sensors for solar panels. Google Patents; 2012. 〈http://www.
google.com/patents/US8294451〉, [Assessed October].
[13] Grasso AD, Pennisi S, Ragusa M, Tina GM, Ventura C. Performance evaluation of a
multistring photovoltaic module with distributed DC–DC converters. IET Renew
Power Gener 2015;9:935–42.
[14] Tsanakas JA, Vannier G, Plissonnier A, Ha L, Barruel F. Fault diagnosis and clas-
sification of large-scale photovoltaic plants through aerial orthophoto thermal
15
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
mapping. In: Proceedings of the 31st European photovoltaic solar energy con-
ference and exhibition (EU PVSEC) Hamburg, Germany; 2015.
[15] Tsanakas JA, Ha LD, Al Shakarchi F. Advanced inspection of photovoltaic in-
stallations by aerial triangulation and terrestrial georeferencing of thermal/visual
imagery. Renew Energy 2017;102:224–33.
[16] Triki-Lahiani A, Abdelghani AB-B, Slama-Belkhodja I. Fault detection and mon-
itoring systems for photovoltaic installations: a review. Renew Sustain Energy Rev
2017;82:2680–92.
[17] Kato K. PV module failures observed in the field- solder bond and bypass diode
failures; 2015. 〈http://iea-pvps.org/index.php?Id=223&eID=dam_frontend〉,
[Assessed June].
[18] Djordjevic S, Parlevliet D, Jennings P. Detectable faults on recently installed solar
modules in Western Australia. Renew Energy 2014;67:215–21.
[19] Munoz M, Alonso-Garcia M, Vela N, Chenlo F. Early degradation of silicon PV
modules and guaranty conditions. Sol Energy 2011;85:2264–74.
[20] Forman S. Performance of experimental terrestrial photovoltaic modules. IEEE
Trans Reliab 1982;31:235–45.
[21] Köntges M, Kurtz S, Jahn U, Berger K, Kato K, Friesen T, et al. Review of failures of
photovoltaic modules. IEA PVPS Task 2014;13.
[22] Massi Pavan A, Mellit A, De Pieri D, Lughi V. A study on the mismatch effect due to
the use of different photovoltaic modules classes in large‐scale solar parks. Prog
Photovolt: Res Appl 2014;22:332–45.
[23] Pavan AM, Tessarolo A, Barbini N, Mellit A, Lughi V. The effect of manufacturing
mismatch on energy production for large-scale photovoltaic plants. Sol Energy
2015;117:282–9.
[24] Pavan AM, Mellit A, De Pieri D, Kalogirou S. A comparison between BNN and
regression polynomial methods for the evaluation of the effect of soiling in large
scale photovoltaic plants. Appl Energy 2013;108:392–401.
[25] Cristaldi L, Faifer M, Lazzaroni M, Khalil MMAF, Catelani M, Ciani L. Diagnostic
architecture: a procedure based on the analysis of the failure causes applied to
photovoltaic plants. Measurement 2015;67:99–107.
[26] Pavan AM, Mellit A, De Pieri D. The effect of soiling on energy production for
large-scale photovoltaic plants. Sol Energy 2011;85:1128–36.
[27] Lorenzo E, Moretón R, Luque I. Dust effects on PV array performance: in‐field
observations with non‐uniform patterns. Prog Photovolt: Res Appl
2014;22:666–70.
[28] Kalogirou SA, Agathokleous R, Panayiotou G. On-site PV characterization and the
effect of soiling on their performance. Energy 2013;51:439–46.
[29] Adinoyi MJ, Said SA. Effect of dust accumulation on the power outputs of solar
photovoltaic modules. Renew Energy 2013;60:633–6.
[30] Ndiaye A, Charki A, Kobi A, Kébé CMF, Ndiaye PA, Sambou V. Degradations of
silicon photovoltaic modules: a literature review. Sol Energy 2013;96:140–51.
[31] Yihua H, Wenping C, Jien M, Finney SJ, Li D. Identifying PV module mismatch
faults by a thermography-based temperature distribution analysis. IEEE Trans
Device Mater Reliab 2014;14:951–60.
[32] Yang H, Xu W, Wang H, Narayanan M. Investigation of reverse current for crys-
talline silicon solar cells—New concept for a test standard about the reverse cur-
rent. In: Proceedings of the 35th IEEE Photovoltaic Specialists Conference, p.1–7.
[33] Simon M, Meyer EL. Detection and analysis of hot-spot formation in solar cells. Sol
Energy Mater Sol Cells 2010;94:106–13.
[34] Winter C-J, Sizmann RL, Vant-Hull LL. Solar power plants: fundamentals, tech-
nology, systems, economics. Springer Science & Business Media; 2012.
[35] Rezgui W, Mouss NK, Mouss LH, Mouss MD, Amirat Y, Benbouzid M. Faults
modeling of the impedance and reversed polarity types within the PV generator
operation. In: Proceedings of the 3rd International Symposium on Environmental
Friendly Energies and Applications (EFEA). 2014. p. 1–6.
[36] Zhao Y, Lehman B, De Palma J-F, Mosesian J, Lyons R. Challenges to overcurrent
protection devices under line-line faults in solar photovoltaic arrays. IEEE Energy
Convers Congr Expo (ECCE) 2011:20–7.
[37] Chang M, Chen C, Hsueh C, Hsieh W, Yen E, Ho K. et al. The reliability in-
vestigation of PV junction box based on 1GW worldwide field database. In:
Proceedings of the 42nd IEEE Photovoltaic Specialist Conference (PVSC). 2015.
p. 1–4.
[38] Jakobi WN K-M, Paterna M, Ansorge F, Baar C. K. Ring Faults of Contacts in PV
Module Junction Boxes due to Fretting Corrosion. In: Proceedings of the 29th
European Photovoltaic Solar Energy Conference and Exhibition, p.2505 –10; 2014.
[39] Stellbogen D. Use of PV circuit simulation for fault detection in PV array fields. In:
Proceedings of the 20th IEEE: Photovoltaic Specialists Conference, 1993, p.
1302–7.
[40] Ye Z, Lehman B, de Palma JF, Mosesian J, Lyons R. Fault analysis in solar PV arrays
under: Low irradiance conditions and reverse connections. In: Proceedings of the
37th IEEE Photovoltaic Specialists Conference (PVSC), 2011, p. 1–5.
[41] PVPN1: Ground-Fault Analysis and Protection in PV Arrays Tech topics:
Photovoltaic protection; 2011. 〈http://www.ep-ru.mersen.com〉 [Assessed].
[42] Karmacharya IM, Gokaraju R. Fault location in ungrounded photovoltaic system
using wavelets and ANN. IEEE Trans Power Deliv 2017;99:1–7.
[43] Flicker J, Johnson J, Albers M, Ball G. Recommendations for isolation monitor
ground fault detectors on residential and utility-scale PV systems. In: Proceedings
of the 42nd IEEE Photovoltaic Specialist Conference (PVSC). 2015. p. 1–6.
[44] Johnson J, Montoya M, McCalmont S, Katzir G, Fuks F, Earle J. et al.
Differentiating series and parallel photovoltaic arc-faults. In: Proceedings of the
38th IEEE Photovoltaic Specialists Conference (PVSC). 2012. p. 720–26.
[45] McCalmont S. Low cost arc fault detection and protection for PV systems. Contract
2013;303:275–3000.
[46] Ye Z, Lehman B, Ball R, de Palma JF. Graph-based semi-supervised learning for
fault detection and classification in solar photovoltaic arrays. IEEE Energy Convers
A. Mellit et al.
Congr Expo (ECCE) 2013:1628–34.
[47] IEC61724 S. Photovoltaic System Performance Monitoring – Guidelines for
Measurement, Data Exchange and Analysis, IEC; 1998.
[48] Blaesser GM. D. Guidelines for the Assessment of Photovoltaic Plants, Document A,
Photovoltaic System Monitoring. Commission of the European Communities, JRC,
Ispra, Italy, EUR 16338 EN; 1993.
[49] Blaesser GMD. Guidelines for the assessment of photovoltaic plants, document B,
Analysis and presentation of monitoring data. Commission of the European
Communities, JRC, Ispra, Italy, EUR 16339 EN; 1993.
[50] Pearsall N, Atanasiu B. The European PV system monitoring guidelines-
Modernisation under the PERFORMANCE project. In: Proceedings of the 34th IEEE
Photovoltaic Specialists Conference (PVSC), 2009, p. 256–61.
[51] Pearsall N, Atanasiu B. Assessment of PV System Monitoring Requirements by
Consideration of Failure Mode Probability. In: Proceedings of the 24th EUPVSEC.
2009: pp. 3896–903.
[52] Tsanakas JA, Ha L, Buerhop C. Faults and infrared thermographic diagnosis in
operating c-Si photovoltaic modules: a review of research and future challenges.
Renew Sustain Energy Rev 2016;62:695–709.
[53] Chine W, Mellit A. ANN-Based Fault Diagnosis Technique for Photovoltaic Strings.
In: Proceedings of the 5th International on Electrical Engineering (ICEE-B), 2017,
p.1–6.
[54] Daliento AC S, Guerriero P, Pavan AMassi, Mellit A, Moeini R, Tricoli P.
Monitoring, Diagnosis, and Power Forecasting for Photovoltaic Fields: A Review.
International Journal of Photoenergy(In press).
[55] Takashima T, Yamaguchi J, Otani K, Kato K, Ishida M. Experimental Studies of
Failure Detection Methods in PV Module Strings. In: Proceedings of the 4th IEEE
World Conference Photovoltaic Energy Conversion. 2006. p. 2227–30.
[56] Takashima T, Yamaguchi J, Otani K, Oozeki T, Kato K, Ishida M. Experimental
studies of fault location in PV module strings. Sol Energy Mater Sol Cells
2009;93:1079–82.
[57] Takashima T, Yamaguchi J, Ishida M. Disconnection detection using earth capa-
citance measurement in photovoltaic module string. Prog Photovolt: Res Appl
2008;16:669–77.
[58] Schirone L, Califano F, Moschella U, Rocca U. Fault finding in a 1 MW photovoltaic
plant by reflectometry. In: Proceedings of the 24th IEEE Photovoltaic Specialist on
Energy Conversion, 1994, p. 846–9.
[59] Schirone L, Califano F, Pastena M. Fault detection in a photovoltaic plant by time
domain reflectometry. Prog Photovolt: Res Appl 1994;2:35–44.
[60] Johnson J, Kuszmaul S, Bower W, Schoenwald D. Using PV Module and Line
Frequency Response Data to Create Robust Arc Fault Detectors. In: Proceedings of
the 26th European Photovoltaic Solar Energy Conference and Exhibition Hamburg,
Germany. 2011. p. 3745–50.
[61] Johnson J, Pahl B, Luebke C, Pier T, Miller T, Strauch J, Kuszmaul S, Bower W.
Photovoltaic DC Arc Fault Detector testing at Sandia National Laboratories. In:
Proceedings of the 37th IEEE Photovoltaic Specialists Conference (PVSC), 2011, p.
3614–19.
[62] Takashima T, Yamaguchi J, Ishida M. Fault detection by signal response in PV
module strings. In: Proceedings of the 33rd IEEE Photovoltaic Specialists
Conference, 2008 PVSC'08. 2008. p. 1–5.
[63] González M, Raison B, Bacha S, Bun L. Fault diagnosis in a grid-connected pho-
tovoltaic system by applying a signal approach. IECON −In: Proceedings of the
37th Annual Conference on IEEE Industrial Electronics Society. 2011. p. 1354–59.
[64] Dini DA, Brazis PW, Yen K-H. Development of arc-fault circuit-interrupter re-
quirements for photovoltaic systems. In: Proceedings of the Photovoltaic
Specialists Conference (PVSC), 2011 37th IEEE: IEEE. 2011. p. 001790–94.
[65] Il-Song K. Fault detection algorithm of the photovoltaic system using wavelet
transform. In: Proceedings of the 2010 India International Conference on Power
Electronics (IICPE). 2011. p. 1–6.
[66] Braun H, Buddha ST, Krishnan V, Spanias A, Tepedelenlioglu C, Yeider T. et al.
Signal processing for fault detection in photovoltaic arrays. In: Proceedings of the
2012 IEEE International Conference on Acoustics, Speech and Signal Processing
(ICASSP). 2012. p. 1681–84.
[67] Wang Z, Balog RS. Arc fault and flash detection in DC photovoltaic arrays using
wavelets. In: Proceedings of the 39th IEEE Photovoltaic Specialists Conference
(PVSC). 2013. p. 1619–24.
[68] Davarifar M, Rabhi A, El-Hajjaji A, Dahmane M. Real-time model base fault di-
agnosis of PV panels using statistical signal processing. IEEE Int Conf Renew
Energy Res Appl (ICRERA) 2013:599–604.
[69] Guasch D, Silvestre S, Calatayud R. Automatic failure detection in photovoltaic
systems. In: Proceedings of the 3rd IEEE World Conference on Photovoltaic Energy
Conversion. 2003. p. 2269–71.
[70] Vergura S, Acciani G, Amoruso V, Patrono G. Inferential statistics for monitoring
and fault forecasting of PV plants. In: Proceedings of the IEEE International
Symposium on Industrial Electronics, ISIE, 2008, p. 2414–19.
[71] Vergura S, Acciani G, Amoruso V, Patrono GE, Vacca F. Descriptive and inferential
statistics for supervising and monitoring the operation of pv plants. IEEE Trans Ind
Electron 2009;56:4456–64.
[72] Garoudja E, Harrou F, Sun Y, Kara K, Chouder A, Silvestre S. Statistical fault de-
tection in photovoltaic systems. Sol Energy 2017;150:485–99.
[73] Harrou F, Sun Y, Taghezouit B, Saidi A, Hamlati M-E. Reliable fault detection and
diagnosis of photovoltaic systems based on statistical monitoring approaches.
Renewa Energy 2017.
[74] Miwa M, Yamanaka S, Kawamura H, Ohno H, Kawamura H. Diagnosis of a power
output lowering of PV ARRAY with a (-dI/dV)-V characteristic. In: Proceedings of
the IEEE 4th World Conference on Photovoltaic Energy Conference. 2006. p.
2442–45.
16
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
[75] Kaplanis S, Kaplani E. Energy performance and degradation over 20 years per-
formance of BP c-Si PV modules. Simul Model Pract Theory 2011;19:1201–11.
[76] Chine W, Mellit A, Pavan AM, Lughi V. Fault diagnosis in photovoltaic arrays. In:
Proceedings of the IEEE International Conference on Clean Electrical Power
(ICCEP). 2015. p. 67–72.
[77] Kang B-K, Kim S-T, Bae S-H, Park J-W. Diagnosis of output power lowering in a pv
array by using the kalman-filter algorithm. IEEE Trans Energy Convers
2012;27:885–94.
[78] Chao K-H, Ho S-H, Wang M-H. Modeling and fault diagnosis of a photovoltaic
system. Electr Power Syst Res 2008;78:97–105.
[79] Gokmen N, Karatepe E, Celik B, Silvestre S. Simple diagnostic approach for de-
termining of faulted PV modules in string based PV arrays. Sol Energy
2012;86:3364–77.
[80] Fezzani A, Mahammed IH, Drid S, Chrifi-alaoui L. Modeling and analysis of the
photovoltaic array faults. In: Proceedings of the 3rd IEEE International Conference
on Control, Engineering & Information Technology (CEIT). 2015. p. 1–9.
[81] Daliento S, Di Napoli F, Guerriero P, d’Alessandro V. A modified bypass circuit for
improved hot spot reliability of solar panels subject to partial shading. Sol Energy
2016;134:211–8.
[82] Hu Y, Zhang J, Cao W, Wu J, Tian GY, Finney SJ, Kirtley JL. Online two-section PV
array fault diagnosis with optimized voltage sensor locations. IEEE Trans Ind
Electron 2015;62:7237–46.
[83] Rezgui W, Mouss H, Mouss N, Mouss D, Benbouzid M, Amirat Y. Photovoltaic
module simultaneous open-and short-circuit faults modeling and detection using
the I–V characteristic. In: Proceedings of the 24th IEEE International Symposium
on Industrial Electronics (ISIE). 2015. p. 855–60.
[84] Hachana O, Tina GM, Hemsas KE. PV array fault DiagnosticTechnique for BIPV
systems. Energy Build 2016;126:263–74.
[85] Wang W, Liu AC-F, Chung HS-H, Lau RW-H, Zhang J, Lo AW-L. Fault diagnosis of
photovoltaic panels using dynamic current–voltage characteristics. IEEE Trans
Power Electron 2016;31:1588–99.
[86] Ali MH, Rabhi A, El Hajjaji A, Tina GM. Real time fault detection in photovoltaic
systems. Energy Procedia 2017;111:914–23.
[87] Chouder A, Silvestre S. Automatic supervision and fault detection of PV systems
based on power losses analysis. Energy Convers Manag 2010;51:1929–37.
[88] Mayer D, Heidenreich M. Performance analysis of stand alone PV systems from a
rational use of energy point of view. In: Proceedings of the 3rd IEEE World
Conference on:Photovoltaic Energy Conversion. 2003. p. 2155–58.
[89] Solórzano J, Egido M. Automatic fault diagnosis in PV systems with distributed
MPPT. Energy Convers Manag 2013;76:925–34.
[90] Stauffer Y, Ferrario D, Onillon E, Hutter A. Power monitoring based photovoltaic
installation fault detection. IEEE Int Conf Renew Energy Res Appl (ICRERA)
2015:199–202.
[91] Xu P, Hou JM, Yuan DK. Fault diagnosis for building grid-connected photovoltaic
system based on analysis of energy loss. Adv Mater Res: Trans Tech Publ
2013:93–8.
[92] Shimakage T, Nishioka K, Yamane H, Nagura M, Kudo M. Development of fault
detection system in PV system. In: Proceedings of the 33rd IEEE International
Telecommunications Energy Conference (INTELEC). 2011. p. 1–5.
[93] Dhimish M, Holmes V. Fault detection algorithm for grid-connected photovoltaic
plants. Sol Energy 2016;137:236–45.
[94] Zhiqiang H, Li G. Research and implementation of microcomputer online fault
detection of solar array. In: Proceedings of the 4th International Conference on
Computer Science & Education, 2009 ICCSE'09. 2009. p. 1052–55.
[95] Davarifar M, Rabhi A, El Hajjaji A. Comprehensive modulation and classification
of faults and analysis their effect in DC side of photovoltaic system. Energy Power
Eng 2013;5:230.
[96] Ye Z, Ball R, Mosesian J, de Palma JF, Lehman B. Graph-based semi-supervised
learning for fault detection and classification in solar photovoltaic arrays. IEEE
Trans Power Electron 2015;30:2848–58.
[97] Houssein A, Heraud N, Souleiman I, Pellet G. Monitoring and fault diagnosis of
photovoltaic panels. In: Proceedings of the IEEE International Energy Conference
and Exhibition (EnergyCon). 2010. p. 389–94.
[98] Muñoz M, Correcher A, Ariza E, García E, Ibañez F Fault detection and isolation in
a photovoltaic system.
[99] Silvestre S, da Silva MA, Chouder A, Guasch D, Karatepe E. New procedure for
fault detection in grid connected PV systems based on the evaluation of current
and voltage indicators. Energy Convers Manag 2014;86:241–9.
[100] Mahendran M, Anandharaj V, Vijayavel K, Winston DP permanent mismatch fault
identification of photovoltaic cells using arduino. ICTACT Journal on
Microelectronics, p. 79–82.
[101] Gokmen N, Karatepe E, Silvestre S, Celik B, Ortega P. An efficient fault diagnosis
method for PV systems based on operating voltage-window. Energy Convers
Manag 2013;73:350–60.
[102] Chen L, Wang X. Adaptive fault localization in photovoltaic systems. IEEE Trans
Smart Grid 2017. [in press].
[103] Dhar S, Patnaik R, Dash P. Fault detection and location of photovoltaic based DC
microgrid using differential protection strategy. IEEE Trans Smart Grid 2017. [In
press].
[104] Mellit A, Kalogirou SA. Artificial intelligence techniques for photovoltaic appli-
cations: a review. Prog Energy Combust Sci 2008;34:574–632.
[105] Wu Y, Lan Q, Sun Y. Application of BP neural network fault diagnosis in solar
photovoltaic system. Int Conf Mechatron Autom, ICMA 2009:2581–5.
[106] Syafaruddin S, Karatepe E, Hiyama T. Controlling of artificial neural network for
fault diagnosis of photovoltaic array. In: Proceedings of the 16th International
Conference on Intelligent System Application to Power Systems (ISAP) 2011.
A. Mellit et al.
2011. p. 1–6.
[107] Coleman A, Zalewski J. Intelligent fault detection and diagnostics in solar plants.
In: Proceedings of the 6th International Conference on Intelligent Data Acquisition
and Advanced Computing Systems (IDAACS). 2011. p. 948–53.
[108] Ducange P, Fazzolari M, Lazzerini B, Marcelloni F. An intelligent system for de-
tecting faults in photovoltaic fields. In: Proceedings of the 11th International
Conference on Intelligent Systems Design and Applications (ISDA). 2011. p.
1341–46.
[109] Rezgui W, Mouss L-H, Mouss NK, Mouss MD, Benbouzid M. A smart algorithm for
the diagnosis of short-circuit faults in a photovoltaic generator. In: Proceedings of
the First International Conference on Green Energy ICGE, 2014, p. 139-43.
[110] Spataru S, Sera D, Kerekes T, Teodorescu R. Detection of increased series losses in
PV arrays using Fuzzy Inference Systems. In: Proceedings of the 38th IEEE
Photovoltaic Specialists Conference (PVSC). p.464-69; 2012.
[111] Spataru S, Sera D, Kerekes T, Teodorescu R. Diagnostic method for photovoltaic
systems based on light I–V measurements. Sol Energy 2015;119:29–44.
[112] Chao K-H, Chen C-T, Wang M-H, Wu C-F. A novel fault diagnosis method based-on
modified neural networks for photovoltaic systems. Advances in Swarm
Intelligence. Springer; 2010. p. 531–9.
[113] Mohamed A, Nassar A. New algorithm for fault diagnosis of photovoltaic energy
systems. Int J Comput Appl 2015;114.
[114] Cheng Z, Zhong D, Li B, Liu Y. Research on fault detection of PV array based on
data fusion and fuzzy mathematics. In: Proceedings of the Power and Energy
Engineering Conference (APPEEC) Asia-Pacific, IEEE. 2011. p. 1–4.
[115] Chao K-H, Chen P-Y, Wang M-H, Chen C-T. An intelligent fault detection method
of a photovoltaic module array using wireless sensor networks. Int J Distrib Sens
Netw 2014.
[116] Jones CB, Stein JS, Gonzalez S, King BH. Photovoltaic system fault detection and
diagnostics using Laterally Primed Adaptive Resonance Theory neural network. In:
Proceedings of the 42nd IEEE Photovoltaic Specialist Conference (PVSC). 2015.
p. 1–6.
[117] Grichting B, Goette J, Jacomet M. Cascaded fuzzy logic based arc fault detection in
photovoltaic applications. In: Proceedings of the IEEE International Conference on
Clean Electrical Power (ICCEP), 2015, p. 178–83.
[118] Mekki H, Mellit A, Salhi H. Artificial neural network-based modelling and fault
detection of partial shaded photovoltaic modules. Simul Model Pract Theory
2016;67:1–13.
[119] Krim BelaoutAF, Mellit A. A Neuro-fuzzy classifier for fault detection and classi-
fication in photovoltaic module. In: Proceedings of the 8th IEEE International
Conference on Modelling, Identification and Control, p.1–6; 2016.
[120] Yi Z, Etemadi AH. Fault detection for photovoltaic systems based on multi-re-
solution signal decomposition and fuzzy inference systems. IEEE Trans Smart Grid
2017;8:1274–83.
[121] Dhimish M, Holmes V, Mehrdadi B, Dales M. Diagnostic method for photovoltaic
systems based on six layer detection algorithm. Electr Power Syst Res
2017;151:26–39.
[122] Dhimish M, Holmes V, Mehrdadi B, Dales M, Mather P. Photovoltaic fault detec-
tion algorithm based on theoretical curves modelling and fuzzy classification
system. Energy 2017;140:276–90.
[123] Chen Z, Wu L, Cheng S, Lin P, Wu Y, Lin W. Intelligent fault diagnosis of photo-
voltaic arrays based on optimized kernel extreme learning machine and IV
17
Renewable and Sustainable Energy Reviews 91 (2018) 1–17
characteristics. Appl Energy 2017.
[124] Yagi Y, Kishi H, Hagihara R, Tanaka T, Kozuma S, Ishida T, Waki M, Tanaka M,
Kiyama S. Diagnostic technology and an expert system for photovoltaic systems
using the learning method. Sol Energy Mater Sol Cells 2003;75:655–63.
[125] Chine W, Mellit A, Pavan AM, Kalogirou S. Fault detection method for grid-con-
nected photovoltaic plants. Renew Energy 2014;66:99–110.
[126] Badi LWA, Zakaria Z, Nordin AHM, Mustapa RF. Unbalanced faults analysis in grid
- Connected PV system. In: Proceedings of the IEEE International Conference on
Power and Energy (PECon), 2014, p. 360–65.
[127] Fern V, Foito D, Amaral TG. Fault detection and diagnosis in a PV grid-connected
T-type three level inverter In: Proceedings of the IEEE International Conference on
Renewable Energy Research and Applications (ICRERA). 2015. p. 933–37.
[128] Ancuta F, Cepisca C. Fault analysis possibilities for PV panels. In: Proceedings of
the IEEE 3rd International Youth Conference on Energetics (IYCE), 2011, p. 1–5.
[129] Zhao Y, de Palma J-F, Mosesian J, Lyons R, Lehman B. Line–line fault analysis and
protection challenges in solar photovoltaic arrays. IEEE Trans Ind Electron
2013;60:3784–95.
[130] Lin X, Wang Y, Zhu D, Chang N, Pedram M. Online fault detection and tolerance
for photovoltaic energy harvesting systems. In: Proceedings of the International
Conference on Computer-Aided Design: ACM. 2012. p. 1–6.
[131] Wang W, Liu AC-f, Chung HS-h, Lau RW-h, Zhang J, Lo AW-l. Fault diagnostic
device for photovoltaic panels. In: Proceedings of the IEEE Applied Power
Electronics Conference and Exposition (APEC), p.2609–16; 2015.
[132] Chavan S, Chavan M. A model based approach for fault diagnosis in converter of
photovoltaic systems. In: Proceedings of the IEEE Global Conference on Wireless
Computing and Networking (GCWCN), p.112–5; 2014.
[133] Jamshidpour E, Poure P, Saadate S. Photovoltaic systems reliability improvement
by real-time FPGA-based switch failure diagnosis and fault-tolerant DC–DC con-
verter. IEEE Trans Ind Electron 2015;62:7247–55.
[134] Hernández J, Vidal P, Medina A. Characterization of the insulation and leakage
currents of PV generators: relevance for human safety. Renew Energy
2010;35:593–601.
[135] Garcia O, Hernandez J-C, Jurado F. Guidelines for protection against overcurrent
in photovoltaic generators. Adv Electr Comput Eng 2012;12:63–70.
[136] Georgijevic NL, Jankovic MV, Srdic S, Radakovic Z. The detection of series arc
fault in photovoltaic systems based on the arc current entropy. IEEE Trans Power
Electron 2016;31:5917–30.
[137] Hu Y, Gao B, Song X, Tian GY, Li K, He X. Photovoltaic fault detection using a
parameter based model. Sol Energy 2013;96:96–102.
[138] Jain P, Xu J-X, Panda SK, Poon J, Spanos C, Sanders SR. Fault diagnosis via PV
panel-integrated power electronics. In: Proceedings of the 17th IEEE Workshop on
Control and Modeling for Power Electronics (COMPEL), 2016, p. 1–6.
[139] Firth SK, Lomas KJ, Rees SJ. A simple model of PV system performance and its use
in fault detection. Sol Energy 2010;84:624–35.
[140] Hachana O, Hemsas KE, Tina GM. Fault diagnosis of building integrated PV gen-
erator: A metaheuristic technique. In: Proceedings of the 6th IEEE
International:Renewable Energy Congress (IREC), 2015, p.1–6.
[141] Belaout A, Krim F, Salhi A, Mellit A. Multi-class neuro-fuzzy classifier for photo-
voltaic array faults diagnosis. In: Proceedings of the 5th International Conference
on Electrical Engineering, (ICEE-B),p.1–7; 2017.