Fault Detection - VIT Papers
Fault Detection - VIT Papers
A R T I C LE I N FO A B S T R A C T
Keywords: With the exponential growth in global photovoltaic (PV) power capacity, protection of PV systems has gained
Photo voltaic (PV) systems prodigious importance in last few decades. Even with the use of standard protection devices in a PV system,
Fault detection faults occurring in a PV array may remain undetected. Inspired by the ever increasing demand for a reliable fault
Protection challenges detection technique, several advanced techniques have been proposed in literature; especially in the last few
Grid connected PV systems
years. Hence, this paper carries out an in depth analysis of various fault occurrences, protection challenges and
Partial shading
ramifications due to undetected faults in PV systems. Furthermore, with a widespread literature, the paper
critically reviews numerous fault detection algorithms/techniques available for PV systems which are proven to
be effective and feasible to implement. The proposed study is not only limited to surveying the existing tech-
niques but also analyzes the performance of each technique with an emphasis on its: 1) Approach, 2) Sensor
requirements, 3) Ability to diagnose and localize faults, 4) Integration complexity, 5) Accuracy, 6) Applicability
and 7) Implementation cost. Overall, this paper is envisioned to avail the researchers working in the field of PV
systems with a valuable resource, which will assist them to enrich their research works.
⁎
Correspondence to: School of Electrical Engineering, VIT University, Vellore 632014, Tamil Nadu, India.
E-mail addresses: [email protected] (D.S. Pillai), [email protected] (N. Rajasekar).
https://doi.org/10.1016/j.rser.2018.03.082
Received 29 July 2017; Received in revised form 19 March 2018; Accepted 26 March 2018
1364-0321/ © 2018 Elsevier Ltd. All rights reserved.
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
blind spots that lead to hazardous electric fires. However, these devices for performance analysis in PV installations. In addition, a
guidelines involved multiple combinations of protective devices which survey on fault diagnostics in micro grids is presented in [29]. How-
are not cost-effective and do not guarantee accurate fault detection ever, all these studies were limited to the analysis of only physical and
either. Moreover, the presence of MPPT controllers in PV systems also environmental faults that occur in PV systems. Furthermore, the ad-
adversely affects these design strategies [13,24]. Above all, none of the vanced detection techniques were not reviewed for its performance
protection devices available till date can detect partial shading in PV evaluation. Moreover, electrical faults occurring in PV arrays are more
modules which is a very common and an intense fault [25,26]. Al- hazardous that demand immediate attention at the first sight. In this
though partial shading is treated as a fault, more often it is temporary in regard, the evolution of reliable fault detection schemes started with
nature and the protection system must be capable to discriminate this the development of improved protection devices followed by the phy-
phenomenon to avoid false tripping. On the whole, to counter act the sical testing of PV systems using fault detection tools/instruments. In
difficulties and to accurately detect, diagnose and localize multiple the last decade, numerous advanced fault detection algorithms have
faults in a PV system, advanced fault detection techniques are manda- evolved that are capable to achieve accurate and rapid fault detection
tory. in PV systems. An overview of the conventional methods for fault de-
Over the years, numerous fault detection techniques have been tection along with a discussion on the advantages and disadvantages of
proposed for detection and diagnosis of faults in PV systems. Authors in each method is presented in [30]. However, a detailed analysis of all
[27] presented a detailed analysis on various fault detection possibi- these methods especially based on their accuracy, integration com-
lities in PV systems and provided a brief review on the various mon- plexity and cost effectiveness were not discussed. Moreover, the com-
itoring systems available to monitor the performance. In addition, [28] putational effectiveness, detection procedure and the applicability of
also furnishes some key details regarding the selection of monitoring each algorithm/technique were not analyzed either. In addition, a
19
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
handful of advanced fault detection techniques that are efficient to as well as all possible faults that can occur in a typical PV power plant
detect, diagnose and localize multiple faults in a PV system have been can be classified using a tree diagram presented in Fig. 1. The block
proposed recently in the last few years. Hence, with a wide spread lit- diagram of a typical grid-connected PV generation system is shown in
erature, this paper reviews various fault detection techniques which are Fig. 2. As shown, the system comprises of two main stages of power
proven to be effective and feasible to implement. The proposed study flow viz. the DC stage and the AC stage. In the DC stage, unlike other
analyzes the performance of each technique with an emphasis on its: 1) generating sources, PV systems have the following peculiarities: 1) The
Approach, 2) Sensor requirement, 3) Ability to diagnose and localize voltage and current of a PV module is limited and is highly dependent
faults, 4) Integration complexity, 5) Accuracy and 6) Implementation on insolation level and temperature, 2) Usually, with the presence of
cost. Furthermore, an in depth investigation has been made to analyze MPPTs, all PV panels operate at a point which is very near to its ISC and
various faults occurring in PV systems, their effects on the electrical 80% of its VOC . Hence, any fault occurring in the DC side of PV systems
characteristics, protection challenges and the limitations of conven- possess disparate characteristics, especially low fault current magni-
tional protection devices. Rest of the paper is organized as follows; Sub tudes that makes it very difficult to detect and distinguish fault oc-
Section 2 discusses various faults occurrences in a PV system, the currences.
standard protection schemes recommended for detecting the faults and Meanwhile, faults occurring in the distribution side of a grid con-
the protection challenges of the conventional detection techniques. nected or standalone PV system affects only the AC side; which are easy
Section 3 provides a detailed study on recent technological advance- to detect and isolate using the protection standards specified in [9–11].
ments in fault detection of PV systems while an overall comparison of For a clear understanding, a PV schematic is depicted in Fig. 3 to show
advanced fault detection techniques is presented in Section 4. Summary various frequent fault scenarios that occur at different power conver-
and some suggestions for further improvements are discussed in Section sion stages in PV systems. Faults in a PV system can occur due to many
5. reasons [31] and a deep insight is necessary to: 1) Distinguish each fault
occurrence, 2) Understand the necessity of fault detection and 3)
Analyze the protection challenges. Hence, a detailed analysis of various
2. Typical faults in a PV system and protection challenges faults is provided in this section for a better understanding. In short,
this section gives an insight to various faults occurring in a PV system,
Like any other power system, PV generating systems are also very the effects of each fault occurrence on the electrical characteristics,
sensitive to fault occurrences. Various faults investigated in this paper
20
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
their protection challenges and the necessity of detection. PV systems, protection is enabled with the help of suitable grounding
electrodes. However, with this arrangement, isolating the PV array
2.1. Ground faults during ground faults is not possible. Hence, for an ungrounded PV
array, a Residual Current Monitoring Device (RCD), connected between
In order to protect the users from potential electric shock, the non- the positive and negative terminal of each string is usually required to
current carrying parts of a PV array i.e. the metallic parts are usually detect ground faults. Alternative to RCD, offline measurement of dc
grounded using Earth Ground Conductors (EGC) [11,32]. Any acci- insulation resistance using Insulation Monitoring Devices (IMDs) can
dental connection between EGC and Current Carrying Conductors also be employed to detect ground faults. On the other hand, in
(CCC) in a PV array leads to rapid and continuous current flow into the grounded PV systems, continuous flow of current to the ground melts
ground; which is often termed as a “ground fault”. Various potential the GFDI fuse, which is connected in series to the ground conductor to
reasons for the occurrence of ground faults in PV systems are discussed isolate the PV array in the presence of a fault. Hence, grounded PV
in [24,31,33,34]. systems are globally preferred over ungrounded systems. Usually, GFDI
fuses are rated for a current of 1A-5A depending upon the rating of DC-
DC converter used in the system.
2.1.1. Standard protection schemes for detecting ground faults
To provide adequate safety and to avoid permanent failure of the PV
systems, ground faults must be detected and isolated. Based on the 2.1.2. Protection challenges in grounded PV systems
plant location and installation standards, PV systems can be classified For a clear understanding of ground faults and its protection chal-
into two: 1) Grounded PV systems and 2) Un-grounded PV systems. As lenges, two different ground fault scenarios are discussed in this paper:
illustrated in Fig. 4, even though the metallic parts are grounded in 1) Single ground faults and 2) Double ground faults. In Fig. 5, case 1
both the configurations, grounded PV systems have a live connection and case 2 shows two different configurations of single ground faults
between EGC and CCC via GFDI fuse. While in the case of ungrounded (indicated in red) and case 3 illustrates the combined occurrence two
21
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
single ground faults that in turn create a double ground fault. In case 1, will be substantial enough to blow out the GFDI fuse. Even though the
the potential difference between the fault point and the ground is suf- GFDI fuse is blown out, the fault current still has a low impedance path
ficiently high (low impedance) to create a substantial current flow to to flow through the initially undetected fault point. Therefore, the array
the ground. Now, the maximum fault current in this case is given by operates as a complete short circuit resulting in a heavy back-fed cir-
culating current that ignites a rapid fire in the PV system even before
I fmax 1 = (n − 1) In (1) the OCPD clears the fault. Under such circumstances, the maximum
fault current, I fmax 3 is given by;
On the other hand, in case 2, the impedance between the fault point
and the ground is very high since the fault point is already near to the I fmax 3 = 2n × In (2)
ground potential. Hence, the fault current will be significantly low and
is usually not substantial enough to blow the GFDI fuse. However in
case 3, there are two ground faults: 1) With low impedance and 2) With 2.1.3. Effect of ground faults in the output characteristics
high impedance with respect to the ground. Let us assume that the From the above analysis, it is clear that the severity of ground faults
faults occur in the sequence of high impedance fault followed by the depends upon the location of the fault point and may even remain
low impedance fault. Since the fault current is not significant enough to undetected. Undoubtedly, an uninterrupted ground fault in a PV array
blow the 5 A GFDI fuse during the high impedance fault, this fault re- induces changes in the electrical characteristics of the PV system as
mains undetected in the PV array; usually referred as ‘blind spots’ well. Anomalies present in the I-V curve as a result of ground fault are
[35–39]. In sequence, if a low impedance fault occurs, the fault current extensively studied in [19,24,40,41] and is shown in Fig. 6; where
Fig. 5. Analysis of ground faults. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
22
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
Combining (3) and (4), installed fuse rating of any OCPD is ex-
pressed as,
Io = 2.1 × ISC (5)
23
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
24
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
inverter for small PV systems and in the combiner box for large PV sometimes get hampered in the following facets: 1) Installation of AFCIs
systems [30]. However, as illustrated in Fig. 10, using multiple com- and AFDs in the inverter attenuates the arc signal since the signal has to
binations of AFDs/AFCIs is a more reliable option since it provides an propagate through various junctions and interconnections in a PV
enhanced string level protection [52,60]. On the other hand, GFDIs and array, 2) Reliable arc fault detection can only be achieved with the help
OCPDs are expected to be efficient enough to address majority of par- of multiple AFDs and AFCIs installed in a PV array; which is highly
allel arc faults to the ground [59]. expensive, particularly for small PV systems, 3) Switching frequencies
of power electronic devices in a PV systems can interfere with AFCIs/
2.3.2. Significance of extinguishing arc faults in PV systems AFDs leading to nuisance tripping, 4) PV systems with single AFCI
Timely isolation of inverter terminals usually plays a crucial role in isolate the whole PV array; irrespective of the location of the arc fault
extinguishing arc faults in PV systems. In this regard, isolating the in- and 5) Even though parallel arcs are more severe and difficult to ex-
verter terminals completely blocks the current flow through the PV tinguish, AFCIs/AFDs can only detect series arc faults.
array and hence, usually de-energizes a series arc fault. In the case of a
parallel arc fault, isolating the inverter terminals increases the circu- 2.4. Shade faults
lating current through the parallel arc path; as the operating voltage of
the PV array reaches near to the open circuit voltage of the PV array. For optimum performance and energy yield, PV modules are sui-
Hence, distinguishing parallel and series arc faults is also extremely tably interconnected to form a PV array [67,68]. However, the per-
important to avail timely isolation of the PV array. [48,61–63] have formance of PV systems varies as the irradiation profile varies. PV ar-
carried out extensive researches in this regard and the following points rays can be subjected to shading, either homogenous or non-
were suggested to discriminate between the two: 1) Apart from series homogenous. Homogenous shading refers to balanced but reduced ir-
arc faults, the combination of arcing noise and drop in array current is a radiation over all PV panels while, non-homogenous shading is the
good indicator of parallel arc faults, 2) Presence of arcing noise at an uneven distribution of irradiation across the panels in a PV array. Non
operating point near to the open circuit voltage of PV array indicates a homogenous shading most often occurs due to the following: 1) Shadow
parallel arc and 3) A combination of differential current and arc in- of tall buildings or trees, 2) Movement of clouds and 3) Bird droppings
duced noise suggests the presence of a parallel-arc fault. [69,70].
2.3.3. Effect of arc faults in the output characteristics 2.4.1. Effect of shading faults and limitations of mitigation techniques
Unlike ground faults and line-line faults, arc faults have little effect In addition to the reduction in output power, partial shading also
on the electrical characteristics of a PV array [63]. However, arc faults leads to the formation of hotspots in the shaded panel; there by accel-
induce serious distortions in the output current and voltage wave forms erating its aging process. Hence, to diminish the effects of partial
of the PV system. For a clear picture, the distortions in the output shading and to improve the power flow, bypass diodes have been
current and voltage wave forms of a 6 × 1 PV array consisting of proposed [71,72]. Even though bypass diodes improve the power flow,
170 W, 44.3 V and 5.7 A manually adjustable MSK solar modules during bypassing of panels will introduce multiple peaks in the I-V and P-V
series and parallel arc fault operation is depicted in Fig. 11. For each curves; making it a challenging task for the MPPT to catch the global
case, the waveforms for ‘ipa ’, ‘upa ’, ‘iiv ’, ‘uiv ’, ‘i fa ’ and ‘ufa ’ are shown. peek. For the PV configuration illustrated in Fig. 12, the effect of
shading on the PV characteristics shown in Fig. 13 that indicates the
2.3.4. Limitations of conventional protection schemes presence of multiple peaks due to bypassing of panels. In addition,
An arc fault protection device should be fast enough to detect the shading faults impose the following challenges in a PV system as well
abnormal patterns in the output waveforms in order to isolate the faulty [73–76]: 1) Additional power loss persists as long as current is flowing
string even before the fault ignites a fire. Unfortunately, [64–66] sug- through the bypass diode, 2) Despite of bypass diodes, infrared ther-
gests that the satisfactory operation of arc fault protection devices can mography has shown that local spots can still exist in the shaded panel
Fig. 11. Current and voltage waveforms during; parallel arc (a) and series arc (b) [59].
25
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
due to the avalanche effect, 3) Providing bypass diodes for each module
is very expensive and makes the system structure more complex and 4)
• Rather than the type of fault, the magnitude of fault current depends
upon the location of faults in a PV array.
Permanent partial shading due to dust accumulation and bird droppings
can cause degradation of the panels, if it remains undetected for a long
• Among all the faults occurring in a PV system, double ground faults
and arcing faults pose severe impacts and hence, need firm protec-
time. tion rules.
• Conventional protection devices are not sensitive enough to protect
the PV array from all fault occurrences and hence, new fault de-
2.4.2. Necessity of detecting shading faults tection tools are mandatory.
Till date, no protection devices are available which can detect
shading faults in a PV system. Even though shading faults are common
• Apart from detection, distinguishing faults is also important for
proper isolation and maintenance of the system.
and is usually temporary, detecting shading faults is extremely im-
portant in the context of protection of PV systems due to the following
facts: 1) Rapid detection of permanent shades increases the life span of 3. Advanced fault detection techniques
a PV system, 2) Detecting shading faults reduces the complexity in the
selection of MPPTs [77] and 3) More importantly, since shading faults As concluded in the previous section, there are three conventional
have similar signatures to that of other electrical faults, temporary schemes to protect a PV system from catastrophic electrical faults: 1)
shading faults must be detected and discriminated to avoid false trip- Use of standard protection devices, 2) Combined use of multiple pro-
ping of PV systems. tection devices in a single PV array and 3) Offline/real-time testing of
PV systems with the help of testing tools to detect the presence of faults.
However, none of these protection devices have shown trustworthy
Fig. 13. I-V characteristics (a) and P-V characteristics (b); due to different irradiation patterns.
26
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
Table 2
Standard protection devices and its features.
Sl. No. Protection devices/methods Type Rating criterion Parameter analyzed
27
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
Table 3
Advantages and limitations of conventional fault detection devices/methods.
Device/method Advantages Limitations
3.1. Model based difference measurement (MBDM) 3.1.1. Survey on MBDM techniques
Following MBDM, [99] proposed an automatic fault detection
PV modeling is vital in design, optimization and control of PV sys- scheme for grid connected PV systems. The method was based on
tems. The capability of a PV model to accurately emulate the electrical analyzing the difference between model predicted power ratio (PR) and
characteristics of a PV system, especially in varying environmental real-time PR. The error value was then used to compute the absolute
conditions has made PV modeling a very handy tool in the analysis and power ratio error (APRE) to detect the fault. Furthermore, the error in
detection of faults [157–159]. MBDM techniques rely on the compar- RC and RV defined in [174] was utilized to distinguish between string
ison of real-time parameters with the model predicted data to identify faults, inverter faults and general fault conditions of the PV system;
faults in the system. Based on the sensed instantaneous irradiance and while the ratio between string current and AC power was analyzed to
temperature levels, theoretical parameters are calculated using a PV locate the fault point. Similarly in [100], the real time system losses and
model and these parameters are then compared with the real-time model predicted system losses were compared to detect the fault; while
parameters to identify the presence of faults in the system. The com- the measured and predicted difference between DC voltage and DC
plete cycle of MBDM based fault detection is shown in Fig. 16. Many PV current was used as an indicator to differentiate shading faults from
models have been proposed over the years for PV characteristic pre- other faults in a system. In another similar approach presented in [101],
diction [160–169]. Among many, the two basic conventional models, the difference between the actual power and predicted power was the
single diode (SD) and double (DD) model [160–162] are the most only detection variable utilized to detect hotspots, ground faults and
commonly used PV models for fault detection; owing to its simplicity open circuit faults in a PV system. However, a Wald test was utilized to
and accuracy. However, the reliability of the PV model depends upon eliminate erroneous detection. Furthermore, diagnosis of faults was
Fig. 15. Different investigation areas considered for analyzing advanced fault detection techniques in PV systems.
28
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
performed with the help of a flash test device that has the ability to positive and negative peaks in the derivative of standard error curve to
scrutinize the I-V and P-V curves. A different approach was put forward detect shading faults; thus eliminating the effect of thermal power
in [102] where multiple PV models were used to detect shading faults, dissipation. [110] recommended a dynamic state estimation algorithm;
diode faults and ground faults. Three PV models representing ideal, which analyzed the difference between real-time data and model pre-
normal and faulty operation of a PV system were designed using an dicted data for terminal voltage and current residuals to detect high
innovative Neuro-Fuzzy approach such that the model has hybrid impedance and low impedance ground faults in a PV array. A DD model
ability to learn from experience and to establish precise relationships was used to predict the PV characteristics and the model parameters
between input and output variables. The training data for the model were extracted using least square method specified in [189]. An alter-
was obtained using a hybrid simulator PV model presented in native and interesting method based on the parameter extraction of PV
[175–177] and the detection rule was based on the comparison of norm models was proposed in [111]; where the change in intrinsic para-
values of six different output parameters with respect to the threshold meters of a PV model with junction capacitance [190] was analyzed to
limits set by normal and faulty PV models. To reduce the in- detect cracks in PV modules. A modified particle swarm optimization
strumentation errors while measuring irradiance, [103] proposed a (MPSO) based technique presented in [191] extracted the intrinsic
GISTEL (Solar Radiation by Tele Detection) model [178] to predict the parameters based on the dynamic I-V characteristics of the PV model. In
irradiation profile which served as the input to the PV model. Si- order to accurately emulate PV characteristics in shaded conditions,
multaneously, the difference measurement between actual and simu- [112] introduced a Bishop PV model incorporating the avalanche effect
lated DC power was utilized as an indicator for faults; while the current present in a PV cell [192]. In addition, a least squares technique was
and voltage ratios defined in [179] was used to discriminate shading also proposed to magnify the temporal variations in power and voltage
faults from string faults. In a basic MBDM approach presented in [104], residuals in the presence of shading. The fuzzy based detection rules
the error between measured and simulated values of ‘ISC ’, ‘IMP ’, ‘VOC ’, compared the change in residuals and irradiation levels to detect the
‘VMP ’ and ‘N ’ were analyzed to detect the fault. The proposed diagnostic fault. Similarly in [113], the decision making ability of fuzzy classifiers
tool used an SD PV model realized by Newton Raphson algorithm [180] were used for fault detection [193]; where the model predicted and
to identify short circuit faults, connection faults and shading faults in a real-time values of open circuit voltage and output power were com-
PV array. Rather than using the conventional PV model, an AC power pared to detect multiple faults like: 1) Bypass diode fault, 2) Blocking
model was proposed in [105]; based on a parametric approach used to diode fault and 3) Open circuits in the array. In [114], the control limits
model the PV system efficiency with respect to the solar irradiance in of the residual vectors generated using MBDM technique were mon-
the PV module plane [181,182]. The fault detection rule was to com- itored using Exponentially Weighted Moving Average (EWMA) mon-
pare the measured AC power of the PV system with the model predicted itoring chart [194,195] to check for any violations to detect the fault.
AC power. Further, in validation, the detection error was found as low The proposed technique used the change in output power as the re-
as 7.25%. In [106], the effect of shading faults in various system con- sidual vector to detect the fault while the current and voltage residuals
figurations were analyzed and two algorithms were suggested to: 1) corresponding to the maximum power point were utilized to distinguish
Determine the maximum available solar power in all shading conditions between open circuit, short circuit and shading faults. A similar tech-
using a predictive control strategy [183], 2) Detect the shading and nique with EWMA is presented in [115] that can identify the type of the
converter faults in PV systems with MPPT and without MPPT. Subse- faults using an additional Univariate EWMA statistic. Recently, [116]
quently, fault detection and diagnosis were done based on the differ- proposed an advanced technique to detect MPPT faults in PV systems
ence between measured and predicted output power as well as the along with bypass diode, open circuit and short circuit faults. The
voltage ratio. Distinctively in [107], to detect general fault conditions method used the statistical T test to determine the fault condition and
and to localize the fault point, a physical PV model was developed to distinguished faults using the difference measurement in VR and PR. On
predict the DC side and AC side losses in PV systems; while the AC the other hand, [117] proposed the same technique without T test
power prediction was based on the polynomial regression models where, the fault detection was based on the decision making ability of
[184,185]. The fault detection rule was based on a new merit of figure; Fuzzy systems. However, compared to [116,117] does not have the
non uniformity index (NUI) and the localization of fault point was capability to distinguish the fault occurrences.
based on the comparative difference between AC and DC powers. Si-
milar to [99,108] compared the measured and simulated PR to detect 3.2. Real-time difference measurement (RDM)
faults. In addition, diagnosis of faults was based on the comparative
evaluation between model predicted voltage ratio (VR) and actual VR; An RDM technique doesn’t rely on PV models for real-time detection
while a statistical T test was deployed to locate the fault. Apart from of faults in a PV system. RDM technique compares real time entities
other MBDM techniques, [109] utilized I-V curve analysis to detect with their threshold limits to detect faults. However, PV modeling or
shading faults. Unlike [186–188], instead of unanimously arriving at a real time experimentations are usually used to set threshold limits for
decision on the presence of shade, the proposed technique analyzes the the defined detection variables to detect the fault. Compared to MBDM,
29
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
RDM possesses the advantage of rapid detection; which is especially current and voltage indicators were defined with respect to the in-
useful in grid connected PV systems for timely isolation and main- stantaneous maximum power point co-ordinates. The threshold limits
tenance of the faulty panels. Note that the accuracy of RDM techniques were set for these indicators using simulation analysis and real time
depends on the quality of the threshold limits set; that once again is violation of these limits indicated the presence of fault. Further, the
dependent on the accuracy of the PV model. For clarity, the block relation between voltage ratios (see equation (23) in [123]) was scru-
diagram of RDM based fault detection is depicted in Fig. 17. tinized to distinguish partial shading effects. While in [124], a novel
approach to detect and isolate faulty panels in a PV array deploying
array reconfiguration was proposed. The algorithm had three main
3.2.1. Survey on RDM techniques steps: 1) Monitor changes in the instantaneous value of output power to
Based on RDM, [118] defined three different detection rules to detect fault, 2) Use electrical array reconfiguration [206–208] to locate
identify faulty operation of a PV array: 1) Measure the AC output power the fault and 3) Select suitable configuration to bypass the faulty
and compare it with the threshold limits, 2) Compute the performance module. [125] has also propounded a novel technique based on optimal
ratio and compare it with that of the previous instant and 3) Compute placement of sensors; where only voltage sensors were used and the
the power difference between present instant and previous instant. Out change in terminal voltage was measured to detect the fault. Further-
of the three, the strong co-relation of performance ratio with the en- more, for locating the faulty panels, two mathematical techniques viz.
vironmental variation [196] makes the second rule the foremost. Fur- the low voltage and the high voltage diagnosis based on the number of
thermore, similar to [197], the authors of the proposed work concluded healthy panels in the array was also proposed. Similarly in [126],
that fault detection is necessary if there is 6% loss in the output power current of each string and voltages corresponding to specific zones were
of a PV system. On the other hand, [119] used a similar technique as sensed to detect the fault. Any mismatch between the sensed para-
proposed in [198] for detecting shading faults. Numerous sensors were meters indicated the presence of fault; while the mismatch level was the
used to measure the module voltages; while the current between par- key to locate fault points. On the other hand in [127], two new para-
ticular groups of panels were also measured. The control rule was based meters, ‘λ ’ and ‘Larray ’ were defined based on the instantaneous and
on the instantaneous difference between current and voltage data. standard irradiation levels (see Eqs. (1) and (2) in [127]). Threshold
Furthermore, a three level alarm system was also proposed to monitor limit for each parameter was set using simulation analysis and in va-
the severity of the faults; utilizing the decision making ability of a fuzzy lidation, the comparative analysis proved the effectiveness of the pro-
inference system. In [120], a control circuitry based on logical gates posed technique to detect line-line faults and shading faults in a PV
was recommended for detecting faults in the control bocks of a PV system. The diagnostic tool was capable to differentiate the system
system. Faults in MPPT was identified by comparing the difference performance between: 1) Normal operation, 2) Temporary fault mode
between instantaneous values of ‘VMP ’ and ‘VOC ’; while inverter faults and 3) Permanent fault mode. Another simple technique to detect
were determined by comparing the magnitude of the filtered output shading faults in a PV array was introduced in [128], where the change
voltage with the threshold value. In another research work carried out in output power and the duration of power dip was analyzed to detect
in [121], the change in relative power (Prel ) was utilized as the detection the presence of various shadows. Similar method was adopted in [129]
variable to detect short circuit and open circuit faults in the string. as well; where current, voltage and power ratios corresponding to the
Further, based on Prel , a mathematical model was realized to find out the maximum power point were defined. The threshold limits for these
number of faulty modules in the PV array. Moreover, the method was parameters were set using an offline PV model optimized by Artificial
validated successfully for three different array sizes as well. [122] Bee Colony (ABC) optimization [209] and the real-time violations were
presented a fault detection technique capable of detecting even minute checked to detect and diagnose faults. Recently in [130], an RDM based
module level failures in PV systems with distributed MPPTs [199]. The fault detection technique was proposed to detect degradation faults and
change in normalized power was used as an indicator to detect the line-line faults in a PV array. Real time difference in the Module De-
fault. Further, the energy loss in each string/module was the indicator gradation Factor (MDF) was used as an indicator for detecting de-
to analyze the severity of the fault and change in average values of gradation of panels while; the difference between array current and
output current and voltage were monitored for diagnosis. Apart from string currents was measured to detect line-line faults.
shading faults and hotspots, the proposed scheme was potentially
capable to predict effects of panel degradation [200–204]. Inspired by
the model presented in [205], an automatic fault detection scheme was
recommended for grid connected PV systems in [123]; where new
30
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
3.3. Output signal analysis (OSA) Whenever the modified Tsallis entropy of current violates the threshold
limits, an arcing state is detected. Apart from other OSA techniques,
Any fault occurrence in the PV system is expected to have a sub- [136] extended the work presented in [219] to give a full description of
stantial effect on the terminal output characteristics; particularly dis- hotspot detection in a particular PV string with the help of two fre-
tortions in the output current and voltage waveforms. OSA techniques quency measurements: 1) Capacitance measurement of the PV string at
usually apply suitable analysis on the output signal to discover the 50 kHz and 2) DC impedance measurement of the string at 50 Hz. Ex-
anomalies present. For an example, as explained earlier, arcing faults perimental results indicated that both the parameters will show sig-
are usually associated with transients in the voltage and current wave nificant increase when there are hot spots present in the string. While in
forms. Hence, analyzing these signals using Fast Fourier Transforms [137], without any prior knowledge on the behavior of the fault, a
(FFT) can detect the anomalies in the wave form; which can be used as sequential technique was proposed based on the Generalized Likelihood
an indicator to detect the fault. For a better understanding, a basic Ratio (GLLR) test [220–222]; where both the time co-relation of faulty
block diagram for OSA based fault detection procedure is presented in signals and co-relation among simultaneously measured multiple cur-
Fig. 18. Even though OSA techniques are usually preferred for detecting rent, voltage and power signals were exploited to characterize the
arc faults, some recent researches have shown its potential to detect faulty signal using an autoregressive (AR) model [223–225]. The pre-
other faults too. Hence, this section analyzes different OSA techniques diction involved two steps: 1) Use AR model to detect sequential
for fault detection in PV systems and a comprehensive survey is also changes in output parameters using multiple meters and 2) Conduct
presented. GLLR test to check for fault. Another fault detection approach based on
Multi Signal Decomposition (MSD) and Fuzzy Inference System (FIS)
was recommended in [138] for DC side fault detection in a PV system.
3.3.1. Survey on OSA techniques Four specific features of the output voltage and current signal (see Eqs.
Following OSA, instantaneous samples of the output current signal (5), (6) and (7) in [138]) was extracted using MSD based signal pro-
from a PV string was analyzed in [131] to check for the presence of cessing [226] and then, fuzzy logic rules were applied for non-erro-
outliers [210]; usually defined as an anomalous data with respect to the neous detection of line-line faults and ground faults. Further, diagnosis
theoretical standards. Three different outlier detection rules were tested of faults was also possible based on the magnitude of the fuzzy output.
for optimum performance: 1) Three Sigma Rule, 2) Hampel Identifier In [139], another method for series arc detection based on both time
and 3) Box Plot Rule. Among the three, Box Plot Rule showed superior domain and frequency domain analysis of the output current waveform
accuracy to detect faults in all operating conditions of a PV. Alter- was proposed. A time domain detection variable, usually non-stationary
natively, to increase the accuracy in outlier detection, a quantitative in nature, was predicted using variance statistical analysis along with a
approach based on machine learning Local Outlier Factor (LOF) [211] frequency domain detection variable obtained from Short Time Fourier
was proposed in [132] to detect line-line faults; which was effective to Transformation (STFT). These variables were then compared with their
eliminate false tripping that is usually associated with statistical basic dynamic threshold limits to detect the presence of series arcs. However,
outlier rules. [133] introduced a Common Mode (CM) model [212] of in [140], a discrete wavelet transformation [227] was used to detect the
the PV systems to detect high impedance ground faults. CM model presence of arcs in PV systems. Discrete wavelet transformation was
analyzed the leakage current flow in a PV system and the detection rule applied to the instantaneous power samples and the detection rule was
was based on FFT based spectral analysis of CM current. The compar- based on the variation of Arc-Signal to Noise Ratio (ASNR). Further-
ison of resonant frequency of the instantaneous CM current with that of more, the method was validated for the synthetic data as well as the
normal operation was utilized as the indicator for fault detection while, real-time arc fault data to prove its effectiveness even in the presence of
the magnitude of the resonant frequency was analyzed to locate the inverter noise. Recently, a novel non-iterative fault detection technique
fault point. A new fault detection algorithm based on the Multi-Level based on DC cable modeling was introduced in [141] for low voltage
Decomposition (MLD) of the wavelet transform [213,214] was pro- DC micro-grids that had the capability to detect arc faults as well as
pounded in [134]; where the wavelet coefficients of the output current other micro grid faults. The proposed fault detection and diagnosis
signal were obtained using MLD [215]. Detection of faults was based on scheme was based on the violations in Cumulative Sum (CuSum)
the standard deviation and mean of wavelet coefficients. Further, as threshold index. On the other hand, for locating the distance of fault
explained in [216], the diagnosis of the fault was done by analyzing the location, a differential current strategy based on Moore Penrose pseudo
energy levels of wavelet coefficients. Another technique based on OSA inverse technique [228] was utilized. In an entirely distinctive ap-
is presented in [135] for detecting series electric arcs in a PV system. proach, DC arc fault in a PV system was detected using electromagnetic
The proposed research was based on a quantum probability model radiation analysis in [142]. The frequency spectrum of the
[217] and modified Tsallis entropy [218] of PV panel current.
31
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
electromagnetic radiation during arc fault was analyzed using a fourth Network (PNN) which is a radial basis neural network [231] was used
order Hilbert antenna [229] to obtain the frequency corresponding to to detect open circuit faults and line-line faults in a PV array. In a vision
the highest amplitude of the radiation signal. Experimental analysis to improve the accuracy in detection, a mathematical PV model was
conducted in the proposed research work has shown that, in the pre- built by incorporating the standard and nominal operating conditions of
sence of arc faults, this characteristic frequency will always be in the a PV; which in turn has been utilized to train the PNN network. While in
range of 39 MHz. Further, in validation, the method was found capable a similar approach, [147] utilized a three layered feed forward Artifi-
to distinguish arc faults in the presence of inverter noise too. With the cial Neural Network (ANN) to detect short circuit faults in a PV system.
already available data of array current and voltage, [143] proposed a Compared to radial basis neural network, the proposed structure has a
new method to detect line-line faults in a PV system based on wavelet simple structure and a better accuracy [232]. Furthermore, the pro-
packets. Discrete Wavelet Transforms (DWT) of the sampled array posed algorithm has the capability to locate faults too. Another ANN
voltage, energy of array voltage and energy of the change in impedance based technique was proposed in [148], where a Multi-Layer Percep-
are extracted and are compared with the threshold limits to detect the tron ANN (MLPANN) was utilized to detect shading faults. [233] vali-
fault. The proposed method has shown good accuracy in detecting line- dated the superior performance of MLPANN towards approximation
line faults with no need of additional sensors. However, extracting problems. The advantage with the proposed method was that the
wavelet packets is extremely difficult and it requires additional soft- training data was obtained from real-time faulty conditions in a PV
ware platforms. [144] proposed an improved method of SSTDR to de- system and was trained using a Resilient Back Propagation (RP) algo-
tect line-ground faults in PV systems. The propounded method was rithm. Similar ANN was used in [149]; but the training data was col-
experimentally validated using WILMA LWG40414 SSTDR hardware lected using the simulation analysis of an SD PV model. Furthermore,
device. Even though the method has shown promise in detecting even analytical approximations were used to find the open circuit voltage of
high impedance ground faults, the capability of the technique to in- the PV array after the occurrence of a fault; which served as the tool for
stantaneously detect the fault is questionable. fault diagnosis. On the other hand, [150] introduced a Laterally Primed
Adaptive Reference Theory (LAPART) based ANN [234] to identify the
general faulty conditions of in a PV system. Limited on site data was
3.4. Machine learning techniques (MLT)
collected to train the network which was later calibrated using PV si-
mulation package. Furthermore, the method was validated using both
In MLT based fault detection, machine learning algorithms are
experimental data and model predicted data. Distinctively, [151] has
trained to learn the relationship between input and output parameters
propounded a Reliability Centered Maintenance Scheme (RCM)
of a PV system. Training data can be collected either experimentally or
[235,236] for grid connected PV systems. This research work identified
with the help of accurate PV models. However, anomalous data during
and analyzed the severity of six different faults through Failure Modes
fault occurrences is also required for accurate training and prediction.
Effects and Criticality Analysis (FMECA) tool [237]. However, MLPANN
Once an accurately trained model is built, it can easily identify faulty
was used as a key tool for pattern recognition in the proposed research
states by simple error evaluations. MLT based fault detection helps to
to diagnose and identify faults with same signatures. The disadvantages
overcome the difficulty of defining thresholds and aids in accurate
discussed earlier stands for all MLT based fault detection methods as
detection and classification of faults. However, the MLT techniques
well.
have the following disadvantages too: 1) Accuracy depends upon the
quality in training data used, 2) Training data; especially for fault oc-
currences are extremely difficult to collect, 3) Training machine
3.5. Infrared thermography (ITH)
learning algorithms using PV models is not reliable; as the accuracy
depends entirely on the PV model used and 4) Training data is not
As discussed earlier, both non-homogenous shading and accumu-
globally available and it varies with the type, size and location of PV
lation of dirt in a PV panel can bring about thermal imbalance to the
systems. The steps involved in fault detection using MLT are depicted in
panel structure; particularly due to the formation of hot spots. It is
Fig. 19.
obvious that hotspots can drastically increase the temperature of PV
modules which are subjected to undetected faults. Hence, it is very
3.4.1. Survey on MLT based fault detection important to monitor the temperature of PV modules in a PV array to
A Decision Tree (DT) algorithm [230] which is an effective method account for the efficiency analysis and maximum power prediction.
to implement classification methods was trained to build an array level Infrared thermography is typically a graphical measurement obtained
PV model in [145]. Furthermore, in validations, the proposed DT model via infrared cameras to analyze the infrared energy emitted from an
has shown 99% accuracy in detecting line-line faults, shading faults and object; which is then converted into temperature to display a thermo
ground faults. On the other hand, in [146], a Probabilistic Neural graphic image of temperature distribution. [152] employed this
32
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
technique to detect hotspots in a PV module and concluded with the 4. Overall review on the performance of various fault detection
following observations: 1) Hot spots not even visible to human eye can technologies
be accurately detected, 2) Faults can be detected even before they be-
come failures, 3) Structural defects in a PV module can also be detected; In order to analyze the performance of various advanced fault de-
since these detects cause abnormal distribution of temperature over the tection techniques, this section comprehensively compares each tech-
surface of the module, 4) Even though infrared thermography is highly nique in a common platform. As expressed in Table 4, each technique
accurate, it requires additional manpower to monitor all PV modules in has been analyzed on the basis of eight important features which play
a large PV system and 5) ITH technique is highly expensive. Analysis of an extremely crucial role in making a fault tolerant PV system efficient
abnormal distribution of temperature on the surface of a PV panel and and compact. The following features have been analyzed and are briefly
detection of hotspots using ITH is shown in Fig. 20 (a) and (b) re- discussed: 1) Approach: Indicates the general procedure followed for
spectively. achieving fault detection, 2) Sensor requirement: Analyzed to find
whether additional sensors are required for fault detection other than
3.6. Hybrid detection techniques (HDT) the usually used current and voltage sensors for MPPT operation and if
so, how many are required, 3) Fault diagnosis capability: Indicates
Hybrid techniques usually combine two fault detection techniques whether the method can efficiently distinguish between different faults
in a single algorithm and have been proposed for fault detection in PV of same signatures, 4) Ability to locate faults: Corresponds to the cap-
systems in order to: 1) Improve the accuracy in detection, 2) Reduce the ability of a method to locate the fault point, such that maintenance
computational complexity, 3) Accurately distinguish between faults of becomes easy and efficient, 5) Integration complexity: Indicates how
same signatures and 4) Detect multiple faults occurrences. Following easily the proposed method can be integrated with existing PV systems;
this strategy, a few fault detection algorithms are also available in lit- which depends mainly upon the detection rule, components required
erature. and structure of the proposed technique, 6) Applicability: Analyzed
based on the detection parameters used; which indicates the method's
applicability towards PV systems with and without MPPT. For instance,
3.6.1. Survey on hybrid detection techniques
some methods necessitate the MPP data itself for fault detection; in-
[153] proposed a novel PV model by incorporating an energy bal-
dicating its applicability towards PV systems with MPPT only, 7) Ac-
ance equation to the conventional cell model to relate the electrical and
curacy: This is the main feature of any detection technique and is
thermal characteristics of a PV. The hybridization was made with ITH
analyzed based on the accuracy of detection rule, selection of detection
technique and its accuracy was utilized to measure the temperature of
variables and performance of the technique in all environmental con-
each panel for achieving reliable predictions from PV model. The de-
ditions and 8) Cost of realization: Another important feature which is
tection rule was based on the change in ′UPV ′ between different panels.
analyzed by considering the number of additional components and
Another hybrid technique propounded in [154] combined MLT and
sensors required, cost of the components and the hardware platform
RDM technique together to identify four different faults in a PV system.
required for implementation.
The proposed algorithm had two parts: 1) For diagnosing faults char-
acterized by different fault signatures, conventional RDM approach was
utilized and 2) For the faults characterized by the same combination of 5. Future scope for fault detection techniques
fault signatures, MLT based on ANN was used for diagnosis. Switching
between two modes was achieved by monitoring the change in different After conducting a thorough analysis on various faults, protection
attributes. In [155], reverse DC bias voltage was injected into the PV challenges and advanced detection techniques available for PV system,
string to identify open circuits in the bypass path while hotspots were it is worth to mention that protection of PV systems is an area that is
detected using the thermal images obtained via ITH technique. A less explored by the researchers around the globe compared to MPPT
completely different approach based on OSA and MLT was recently techniques, array reconfiguration methods and grid inter connection
introduced in [156] where a two stage Support Vector Machine (SVM) issues. However, PV protection is extremely crucial to ensure long
was utilized to identify line-line faults. Interestingly, the wavelet lasting reliability and safety of PV systems. In this regard, advanced
packets obtained using MSD was used to train the SVM. Then, SVM was detection techniques studied in this paper are way superior to the
deployed to detect the fault. However, it is not be noted that, in the case conventional protection devices in terms of its detection accuracy and
of line-line faults with less than 20% mismatch level, the accuracy was these techniques can be used as a supplementary detection scheme in
not substantially high to provide accurate detection. all PV systems. Among the different advanced detection approaches,
Fig. 20. Analysis of PV panel temperature using Infrared Thermography: (a) abnormal distribution of temperature [75] and (b) detection of hotspots [152].
33
Table 4
Overall comparison of various fault detection techniques.
Ref Approach Additional sensors No: of sensors required Fault diagnosis Locating faults Integration Applicability Accuracy Cost of realization
required possible possible complexity
[99] MBDM Yes 2 and 1 additional current sensor for each string Yes Yes Easy MPPT and non-MPPT based High Low
D.S. Pillai, N. Rajasekar
systems
[100] MBDM Yes 5 Yes No Easy MPPT and non-MPPT based Low High
systems
[101] MBDM Yes 2 Yes No Easy MPPT based systems High Low
[102] MBDM Yes 2 Yes No Difficult MPPT and non-MPPT based Low High
systems
[103] MBDM No – Yes No Easy MPPT and non-MPPT based Low Low
systems
[104] MBDM No – Yes No Low MPPT based systems High Low
[105] MBDM Yes 4 No No Easy MPPT and non-MPPT based Low High
systems
[106] MBDM Yes 2 Yes No Low MPPT and non-MPPT based High Low
systems
[107] MBDM Yes 2 No Yes Difficult MPPT and non-MPPT based Low High
systems
[108] MBDM Yes 2 Yes Yes Easy MPPT and non-MPPT based High Low
systems
[109] MBDM Yes 1 Yes No Low MPPT and non-MPPT based Low Low
systems
[110] MBDM No – No No Easy MPPT and non-MPPT based Low Low
systems
[111] MBDM Yes 2 for each panel No Yes Very Difficult MPPT based systems High Very High
[112] MBDM yes 1 Yes No Easy MPPT and non-MPPT based High Difficult
34
systems
[113] MBDM Yes 2 Yes No Difficult MPPT and non-MPPT based Low High
systems
[114] MBDM Yes 2 and a reference solar cell Yes No Easy MPPT based systems Low Low
[115] MBDM No – Yes No Difficult MPPT based systems Low Low
[116] MBDM Yes 1 Yes No Easy MPPT and non-MPPT based Low High
systems
[117] MBDM Yes Irradiance sensor and 1 temperature sensor for each No No Easy MPPT and non-MPPT based Low High
module systems
[118] RDM Yes 3 No No Difficult Non-MPPT based systems Low Low
[119] RDM Yes 1 voltage sensor for each string and 1 current sensor for Yes Yes Difficult Non-MPPT based systems Low High
a group of 'n' modules.
[120] RDM Yes 2 Yes No Difficult MPPT based systems Low High
[121] RDM Yes 2 sensors and a power meter for each string Yes No Easy MPPT based systems Low Low
[122] RDM No – Yes Yes Difficult MPPT based systems High High
[123] RDM No – Yes No Low MPPT based systems High Low
[124] RDM No – No Yes Difficult MPPT and non-MPPT based Low High
systems
[125] RDM Yes Based on the size of PV system Yes Yes Easy MPPT and non-MPPT based High Low
systems
[126] RDM Yes Based on the size of PV system No Yes Difficult Non-MPPT based systems High Very High
[127] RDM Yes 2 Yes No Easy MPPT and non-MPPT based High Low
systems
[128] RDM No – Yes No Easy MPPT and non-MPPT based Low Low
systems
[129] RDM Yes 2 Yes No Difficult MPPT based systems Low Low
[130] RDM Yes 1 for each string No No Difficult MPPT and non-MPPT based Low Low
systems
(continued on next page)
Renewable and Sustainable Energy Reviews 91 (2018) 18–40
Table 4 (continued)
Ref Approach Additional sensors No: of sensors required Fault diagnosis Locating faults Integration Applicability Accuracy Cost of realization
required possible possible complexity
[131] OSA Yes 1 for each string No No Easy MPPT and non-MPPT based Low Low
systems
D.S. Pillai, N. Rajasekar
[132] OSA Yes 1 for each string No No Easy MPPT and non-MPPT based High Low
systems
[133] OSA Yes 2 Yes Yes Easy MPPT based systems High Low
[134] OSA No – Yes No Easy MPPT and non-MPPT based High High
systems
[135] OSA No – Yes No Difficult MPPT based systems High High
[136] OSA NO _ Yes No Very Difficult MPPT based systems High High
[137] OSA Yes 1 Yes No Easy MPPT based systems High Low
[138] OSA Yes 1 Yes No Easy MPPT and non-MPPT based High High
systems
[139] OSA No _ Yes No Difficult MPPT and non-MPPT based High Low
systems
[140] OSA No _ Yes No Easy MPPT and non-MPPT based High High
systems
[141] OSA Yes 1 Yes Yes Difficult MPPT and non-MPPT based High High
systems
[142] OSA Yes 1 Hilbert Antenna Yes No Easy MPPT and non-MPPT based High High
systems
[143] OSA No – Yes No Difficult MPPT based systems High High
[144] OSA No – Yes No Difficult MPPT and non-MPPT based High High
systems
[145] MLT Yes 2 Yes No Difficult MPPT and non-MPPT based High High
systems
35
[146] MLT Yes 2 Yes No Difficult MPPT and non-MPPT based High High
systems
[147] MLT Yes 2 Yes Yes Difficult MPPT and non-MPPT based High High
systems
[148] MLT Yes 2 Yes No Difficult MPPT and non-MPPT based High High
systems
[149] MLT Yes 3 for each string Yes No Difficult MPPT and non-MPPT based High High
systems
[150] MLT No – No No Difficult MPPT and non-MPPT based High High
systems
[151] MLT Yes 2 Yes No Easy MPPT and non-MPPT based Low Low
systems
[152] ITH No – Yes Yes Difficult MPPT and non-MPPT based High High
systems
[153] HDT Yes 1 for each string Yes Yes Difficult MPPT and non-MPPT based Low High
systems
[154] HDT Yes 2 for each string Yes No Difficult MPPT and non-MPPT based High High
systems
[155] HDT No – Yes Yes Very Difficult Non-MPPT based systems High High
[156] HDT No – Yes No Very Difficult Non-MPPT based systems Low Low
Renewable and Sustainable Energy Reviews 91 (2018) 18–40
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
RDM, OSA and MLT techniques are product oriented and have the Acknowledgements
potential to become future solutions for error free fault detection.
Nevertheless, these techniques also require research modifications such The authors would like to thank the Management, VIT University,
that it can be easily integrated with any PV system. Overall, the authors Vellore, India for providing the support to carry out research work. This
feel that there is a potential research gap that exists in fault detection of work is carried out at Solar Energy Research Cell (SERC), School of
PV systems and the arena will witness huge research advancements in Electrical Engineering, VIT University, Vellore. Further, the authors
the forthcoming years. also would like to thank the reviewers for their valuable comments and
recommendations to improve the quality of the paper.
6. Summary
References
Inspired by the interest towards fault detection, this paper has re-
viewed a handful of fault detection techniques proposed for PV systems. [1] Biagi M, Falvo MC Smart micro grid programming for renewable resources: from
The attempt made is envisioned to provide the researchers and in- communication to dispatching.
[2] Brenna M, Falvo MC, Foiadelli F, Martirano L, Poli D. Sustainable energy micro-
dustrialists working in the PV arena, with some valuable data and system (SEM): preliminary energy analysis. In: Innovative Smart Grid Technologies
benchmarks to work with. In this paper, the authors have conducted an (ISGT), IEEE PES; 26 Jan 2012. p. 1–6.
in depth analysis of various fault occurrences and its catastrophic ef- [3] Capparella S, Falvo MC. Secure faults detection for preventing fire risk in PV
systems. In: Proceedings of International carnahan conference on security tech-
fects in a PV system. Furthermore, the challenges in detecting each fault nology (ICCST); 13 Oct 2014. p. 1-5.
and the limitations of conventional protection devices were also ana- [4] EPIA. Connecting the Sun: Solar Photovoltaics on the road to large scale grid in-
lyzed. Moreover, the fault detection approaches were broadly classified tegration. [Online], Sept; 2012. Available: 〈http://www.epia.org/news/
publications/〉.
into six categories and the techniques proposed in each category were
[5] Masson G, Latour M, Biancardi D. Global market outlook for photovoltaics until
studied in detail with extensive comparative analysis. On the whole, to 2016. European Photovoltaic Industry Association; 2012.
conclude the work, the following points are likely to be summarized. [6] Zahedi A. Solar photovoltaic (PV) energy; latest developments in the building in-
tegrated and hybrid PV systems. Renew Energy 2006;31(5):711–8.
[7] Protection against electric shock—common aspects for installation and equipment,
➢ The standard ratings of conventional protection devices are not IEC Standard 61140; 2005.
suitable enough to achieve reliable fault detection. Hence, potential [8] LV Electrical Installations—Part 41: protection for safety—protection against
research works must be carried out for the optimal selection of fuse electric shock, IEC Standard 60364-4-41; 2005.
[9] Electrical Installations of Buildings—Part 7: requirements for special installations
ratings for fault detection in PV systems to improve the reliability of or locations—section 712: solar PV power supply systems, IEC Standard 60364-7-
existing PV installations. 712; 2002.
➢ For already installed PV systems comprising of conventional pro- [10] Installation and safety requirements for PV generators, IEC Standard 62548.
[11] Article 690—solar photovoltaic systems of national electrical code, NFPA70; 2011.
tective devices, extreme care must be given to protect the system [12] Albers MJ, Ball G. Comparative evaluation of DC fault-mitigation techniques in
from double ground faults that are usually associated with blind large PV systems. IEEE J Photovolt 2015;5(4):1169–74.
spots. It seems that installing RCDs along with the existing GFDIs [13] Zhao Y, de Palma JF, Mosesian J, Lyons R, Lehman B. Line–line fault analysis and
protection challenges in solar photovoltaic arrays. IEEE Trans Ind Electron
will be a better and feasible option. 2013;60(9):3784–95.
➢ Even though many researches were carried out to detect various [14] Alam MK, Khan FH, Johnson J, Flicker J PV faults: Overview, modeling, preven-
faults in a PV system, a practically feasible fault detection technique tion and detection techniques. In: Proceedings of the 14th workshop on control
and modeling for power electronics (COMPEL). IEEE; 23 Jun 2013. p. 1–7.
to detect the presence of parallel arcs in is not available. Therefore, [15] Capparella S, Falvo MC Secure faults detection for preventing fire risk in PV sys-
any detection scheme must fulfill this motive to achieve effective tems. In: Proceedings of international Carnahan conference on security technology
protection. (ICCST). IEEE; 13 Oct 2014. p. 1–5.
[16] Brooks B. The bakersfield fire-A lesson in ground-fault protection. Sol Pro Mag
➢ Most of the researches proposed in literature are focused on de-
2011:62–70.
tecting and diagnosing the faults. However, in the case of large PV [17] Collier DE, Key TS Electrical fault protection for a large photovoltaic power plant
systems, it is very hard to locate the fault point even if a fault is inverter. In: Proceedings of conference record of the twentieth photovoltaic spe-
detected. In this regard, a fault detection technique must be capable cialists conference. IEEE; 26 Sep 1988. p. 1035–42.
[18] Zhao Y, Lehman B, DePalma JF, Mosesian J, Lyons R Fault evolution in photo-
to locate the fault points in the system for availing proper main- voltaic array during night-to-day transition. In: Proceedings of the 12th workshop
tenance and mitigation. on control and modeling for power electronics (COMPEL). IEEE; 28 Jun 2010.
➢ Selecting detection variables in a fault detection algorithm play an p. 1–6.
[19] Zhao Y, Lehman B, de Palma JF, Mosesian J, Lyons R Fault analysis in solar PV
important role in the diagnosis of faults too. Care should be given to arrays under: Low irradiance conditions and reverse connections. In: Proceedings
select detection variables which are not affected by multiple faults of the 37th photovoltaic specialists conference (PVSC). IEEE; 19Jun 2011. p.
since, the faults with same signatures are difficult to distinguish 002000–5.
[20] Firth SK, Lomas KJ, Rees SJ. A simple model of PV system performance and its use
with the same detection variables. in fault detection. Sol Energy 2010;84(4):624–35.
➢ Even though the system structure becomes complex, detection [21] Colli A. Failure mode and effect analysis for photovoltaic systems. Renew Sustain
procedure utilizing adequate number of sensors is found to be more Energy Rev 2015;50:804–9.
[22] Hernandez JC, Vidal PG. Guidelines for protection against electric shock in PV
accurate, effective and reliable. Hence, to reduce the structural generators. IEEE Trans Energy Convers 2009;24(1):274–82.
complexity, improving these techniques using optimal placement of [23] Falvo MC, Capparella S. Safety issues in PV systems: design choices for a secure
sensors is expected to provide compact and rapid fault detection fault detection and for preventing fire risk. Case Stud Fire Saf 2015;3:1–6.
[24] Zhao Y Fault analysis in solar photovoltaic arrays (Doctoral dissertation,
techniques.
Northeastern University).
➢ Apart from other power sources solar PVs are colossally dependent [25] Ram JP, Babu TS, Rajasekar N. A comprehensive review on solar PV maximum
on the instantaneous irradiation levels and the temperature of the power point tracking techniques. Renew Sustain Energy Rev 2017;67:826–47.
panels. Hence, it is always necessary to consider the effect of these [26] Ram JP, Manghani H, Pillai DS, Babu TS, Miyatake M, Rajasekar N. Analysis on
solar PV emulators: a review. Renew Sustain Energy Rev 2018;81:149–60.
parameters while designing a fault detection algorithm so that rapid [27] Triki-Lahiani A, Abdelghani AB, Slama-Belkhodja I. Fault detection and mon-
fault detection is possible in all working conditions of a PV. itoring systems for photovoltaic installations: a review. Renew Sustain Energy Rev
➢ As a future scope, the performance of each fault detection technique 2017.
[28] Madeti SR, Singh SN. Monitoring system for photovoltaic plants: a review. Renew
introduced in literature can be analyzed for detecting similar faults Sustain Energy Rev 2017;67:1180–207.
in various environmental conditions to identify the best technique [29] Hare J, Shi X, Gupta S, Bazzi A. Fault diagnostics in smart micro-grids: a survey.
available to detect a particular fault. Renew Sustain Energy Rev 2016;60:1114–24.
[30] 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
36
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
2015;5(3):982–97. [64] Dhere NG, Shiradkar NS. Fire hazard and other safety concerns of photovoltaic
[31] Forman SE. Performance of experimental terrestrial photovoltaic modules. IEEE systems. J Photonics Energy 2012;2(1):022006.
Trans Reliab 1982;31(3):235–45. [65] Spooner ED, Wilmot N Safety issues, arcing and fusing in PV arrays.
[32] Bower WI, Wiles JC Analysis of grounded and ungrounded photovoltaic systems. [66] Johnson J, Pahl B, Luebke C, Pier T, Miller T, Strauch J, Kuszmaul S, Bower W
In: Proceedings of the first world conference on photovoltaic energy conversion, Photovoltaic DC arc fault detector testing at Sandia National Laboratories. In:
conference record of the twenty fourth. IEEE photovoltaic specialists conference. Proceedings of the 37th photovoltaic specialists conference (PVSC). IEEE; 19 Jun
IEEE, Vol. 1; 5 Dec 1994. p. 809–12. 2011. p. 003614–9.
[33] Zhao Y, Lyons Jr R. Ground-fault analysis and protection in PV arrays. Proc [67] Deshkar SN, Dhale SB, Mukherjee JS, Babu TS, Rajasekar N. Solar PV array re-
Photovolt Prot 2011:1–4. configuration under partial shading conditions for maximum power extraction
[34] McEvoy A, Markvart T, Castañer L, Markvart T, Castaner L, editors. Practical using genetic algorithm. Renew Sustain Energy Rev 2015;43:102–10.
handbook of photovoltaics: fundamentals and applications. Elsevier; 2003. [68] Rani BI, Ilango GS, Nagamani C. Enhanced power generation from PV array under
[35] Brooks B. The ground-fault protection blind spot: Safety concern for larger PV partial shading conditions by shade dispersion using Su Do Ku configuration. IEEE
systems in the US. Solar American Board for Codes and Standards Jan; 2012. Trans Sustain Energy 2013;4(3):594–601.
[36] Sabbaghpur Arani M, Hejazi MA. The comprehensive study of electrical faults in [69] Ram JP, Rajasekar N. A new global maximum power point tracking technique for
PV arrays. J Electr Comput Eng 2016. solar photovoltaic (PV) system under partial shading conditions (PSC). Energy
[37] Ball G, Brooks B, Johnson J, Flicker J, Rosenthal A, Wiles J, Sherwood L, Albers M, 2017;118:512–25.
Zgonena T. Inverter ground-fault detection‘blind spot’ and mitigation methods. Sol [70] Ram JP, Rajasekar N. A novel flower pollination based global maximum power
Am Board Codes Stand Rep 2013. point method for solar maximum power point tracking. IEEE Trans Power Electron
[38] Flicker J, Johnson J. Analysis of fuses for blind spot ground fault detection in 2017;32(11):8486–99.
photovoltaic power systems. Report. Sandia National Laboratories; 2013. [71] Ram JP, Rajasekar N. A new robust, mutated and fast tracking LPSO method for
[39] Flicker J, Johnson J. Photovoltaic ground fault and blind spot electrical simula- solar PV maximum power point tracking under partial shaded conditions. Appl
tions. Sandia National Laboratories; 2013. Energy 2017;201:45–59.
[40] Wiles JC, King DL Blocking diodes and fuses in low-voltage PV systems. In: [72] Jung TH, Ko JW, Kang GH, Ahn HK. Output characteristics of PV module con-
Proceedings of conference record of the twenty-sixth photovoltaic specialists sidering partially reverse biased conditions. Sol Energy 2013;92:214–20.
conference. IEEE; Sep 1997. p. 1105–8. [73] Nguyen D, Lehman B. An adaptive solar photovoltaic array using model-based
[41] Wu H Fault diagnosis testbed for plug and play photovoltaic system. reconfiguration algorithm. IEEE Trans Ind Electron 2008;55(7):2644–54.
[42] Bower W, Wiles J Investigation of ground-fault protection devices for photovoltaic [74] Zaihidee FM, Mekhilef S, Seyedmahmoudian M, Horan B. Dust as an unalterable
power system applications. In: Proceedings of conference record of the twenty- deteriorative factor affecting PV panel's efficiency: why and how. Renew Sustain
eighth photovoltaic specialists conference. IEEE; 2000. p. 1378–83. Energy Rev 2016;65:1267–78.
[43] Czapp S The effect of earth fault current harmonics on tripping of residual current [75] Sharma V, Sastry OS, Kumar A, Bora B, Chandel SS. Degradation analysis of a-Si,
devices. In: Proceedings of international school on non sinusoidal currents and (hit) hetro-junction intrinsic thin layer silicon and mC-Si solar photovoltaic tech-
compensation. ISNCC. IEEE; 10 Jun 2008. p. 1–6. nologies under outdoor conditions. Energy 2014;72:536–46.
[44] M. Cotterell. Installation guidelines: electrical. In: Practical Hand book of [76] Sharma V, Chandel SS. Performance and degradation analysis for long term re-
Photovoltaics, 2nd ed. Boston. MA, USA: Academic, 2012, ch. IIC-3. p. 819–34. liability of solar photovoltaic systems: a review. Renew Sustain Energy Rev
[45] Outline of investigation for low-voltage fuses—fuses for photovoltaic systems. UL 2013;27:753–67.
subject 2579; 2010. [77] Manickam C, Raman GR, Raman GP, Ganesan SI, Nagamani C. A hybrid algorithm
[46] Catalog No. HP6M10. Special purpose fuses—photovoltaic fuses: HP6M—600VDC for tracking of GMPP based on P&O and PSO with reduced power oscillation in
PV rated fuses. Newburyport, MA: MERSEN; 2011. string inverters. IEEE Trans Ind Electron 2016;63(10):6097–106.
[47] Armijo KM, Johnson J. Characterizing fire danger from low power PV arc-faults. [78] Xu P, Hou JM, Yuan DK. Fault diagnosis for building grid-connected photovoltaic
Albuquerque, NM (United States): Sandia National Laboratories (SNL-NM); 2014. system based on analysis of energy loss. Adv Mater Res 2013;805:93–8.
[48] Johnson J, Montoya M, McCalmont S, Katzir G, Fuks F, Earle J, Fresquez A, [79] Kuznetsova VA, Gaston RS, Bury SJ, Strand SR Photovoltaic reliability model de-
Gonzalez S, Granata J Differentiating series and parallel photovoltaic arc-faults. In: velopment and validation. In: Proceedings of the 34th photovoltaic specialists
Proceedings of the 38th photovoltaic specialists conference (PVSC). IEEE; 3 Jun conference (PVSC). IEEE; 7Jun 2009. p. 000432–6.
2012. p. 000720–6. [80] Dhople SV, Davoudi A, Chapman PL, Domínguez-García AD Reliability assessment
[49] Johnson J, Bower W, Quintana M. Electrical and thermal finite element modeling of fault-tolerant Dc-Dc converters for photovoltaic applications. In: Proceedings of
of arc faults in photovoltaic bypass diodes. World Renew Energy Forum 2012;16. energy conversion congress and exposition, ECCE. IEEE; 20 Sep 2009. p. 2271–6.
[50] Johnson J, Kuszmaul S, Bower W, Schoenwald D Using PV module and line fre- [81] Bazzi AM, Kim KA, Johnson BB, Krein PT, Dominguez-García A Fault impacts on
quency response data to create robust arc fault detectors. In: Proceedings of the solar power unit reliability. In: Proceedings of twenty-sixth annual applied power
26th european photovoltaic solar energy conference and exhibition; 1 Sep 2011. p. electronics conference and exposition (APEC). IEEE; 6 Mar 2011. p. 1223–31.
05–9. [82] Chan F, Calleja H. Design strategy to optimize the reliability of grid-connected PV
[51] Johnson J, Schoenwald D, Kuszmaul S, Strauch J, Bower W Creating dynamic systems. IEEE Trans Ind Electron 2009;56(11):4465–72.
equivalent PV circuit models with impedance spectroscopy for arc fault modeling. [83] Kase R, Nishikawa S Fault detection of bypass circuit of PV module—Detection
In: Proceedings of 37th photovoltaic specialists conference (PVSC). IEEE; 19 Jun technology of open circuit fault location. In: Proceedings of the 19th international
2011. p. 002328–33. conference on electrical machines and systems (ICEMS). IEEE; 13 Nov 2016.
[52] Johnson J, Oberhauser C, Montoya M, Fresquez A, Gonzalez S, Patel A Crosstalk p. 1–4.
nuisance trip testing of photovoltaic DC arc-fault detectors. In: Proceeding of the [84] Jain P, Xu JX, Panda SK, Poon J, Spanos C, Sanders SR Fault diagnosis via PV
38th photovoltaic specialists conference (PVSC). IEEE; 3 Jun 2012. p. 001383–7. panel-integrated power electronics. In: Proceedings of the 17th workshop on
[53] Johnson J, Kang J Arc-fault detector algorithm evaluation method utilizing pre- control and modeling for power electronics (COMPEL). IEEE; 27 Jun 2016. p. 1–6.
recorded arcing signatures. In: Proceeding of the 38th photovoltaic specialists [85] Takashima T, Yamaguchi J, Otani K, Oozeki T, Kato K, Ishida M. Experimental
conference (PVSC). IEEE; 3 Jun 2012. p. 001378–82. studies of fault location in PV module strings. Sol Energy Mater Sol Cells
[54] Flicker J, Johnson J Electrical simulations of series and parallel PV arc-faults. In: 2009;93(6):1079–82.
Proceeding of the 39th photovoltaic specialists conference (PVSC). IEEE; 16 Jun [86] Extech TG20: Wire Tracer Kit; 2013, Nov. [Online]. Available: 〈http://www.
2013. p. 3165–72. extech.com/instruments/product.asp?Catid=2&prodid=768〉.
[55] Schimpf F, Norum LE Recognition of electric arcing in the DC-wiring of photo- [87] Amprobe AT-4004 advanced wire tracer with clamp attachment; 2013, Nov.
voltaic systems. In: Proceedings of the 31st international telecommunications en- [Online]Available: 〈http://www.amprobe.com/amprobe/usen/Wire-Tracers-and-
ergy conference, INTELEC. IEEE; 18 Oct 2009. p. 1–6. CableLocators/Wire-Tracers/AT-4004CON.htm?PID=73167〉.
[56] Dini DA, Brazis PW, Yen KH Development of arc-fault circuit-interrupter re- [88] Penkey P, Alhajeri F, Johnson BK Modeling, analysis and detection of faults in
quirements for photovoltaic systems. In: Proceedings of the 37th photovoltaic grid-connected PV systems. In: Proceedings of the 10th international conference on
specialists conference (PVSC).IEEE 19 Jun 2011. p. 001790–4. intelligent systems and control (ISCO). IEEE; 7Jan 2016. p. 1–5.
[57] Strauch J, Quintana MA, Granata J, Bower W, Kuszmaul S Solar module arc fault [89] Varma RK, Rahman SA, Atodaria V, Mohan S, Vanderheide T. Technique for fast
modeling at Sandia National Laboratories. In: 2011 NREL Module Reliability detection of short circuit current in PV distributed generator. IEEE Power Energy
Workshop; 2010 Aug. Technol Syst J 2016;3(4):155–65.
[58] Seo GS, Bae H, Cho BH, Lee KC Arc protection scheme for DC distribution systems [90] Katiraei F, Holbach J, Chang T. Impact and Sensitivity Studies of PV Inverters
with photovoltaic generation. In: Proceedings of international conference on re- Contribution to Faults based on Generic PV Inverter Models. Quanta Technology; 2
newable energy research and applications (ICRERA). IEEE; 11 Nov 2012. p. 1–5. May 2012.
[59] National Electrical Code(R) (NEC) and Handbook Set (NFPA 70); 2014 edition. [91] Bravo RJ, Yinger R, Robles S Three phase solar photovoltaic inverter testing. In:
[60] Zgonena T, Ji L, Dini D Photovoltaic DC arc-fault circuit protection and UL Subject Proceedings of power and energy society general meeting (PES). IEEE; 21 Jul
1699B. In: Photovoltaic Module Reliability Workshop, Golden, CO; 2011 Feb. 2013.p. 1–5.
[61] Johnson J, Gudgel B, Meares A, Fresquez A. Series and parallel arc-fault circuit [92] Jenkins N, Allan R, Crossley P, Kirschen D, Strbac G. Embedded generation in-
interrupter tests. Tech. Rep. SAND2013-5916 SAND2013-5916]. Albuquerque, stitute of electrical engineering. London, UK; 2000.
NM, USA: Sandia Nat. Lab.; 2013. [93] Hooshyar H, Baran ME. Fault analysis on distribution feeders with high penetra-
[62] McCalmont S. Low cost arc fault detection and protection for PV systems: January tion of PV systems. IEEE Trans Power Syst 2013;28(3):2890–6.
30, 2012-September 30, 2013. Golden, CO.: National Renewable Energy [94] Baran ME, El-Markaby I. Fault analysis on distribution feeders with distributed
Laboratory (NREL); 2013. generators. IEEE Trans Power Syst 2005;20(4):1757–64.
[63] Strobl C, Meckler P Arc faults in photovoltaic systems. In: Proceedings of the 56th [95] Schirone L, Califano FP, Moschella U, Rocca U Fault finding in a 1 MW photo-
IEEE Holm conference on electrical contacts; 4 Oct 2010. p. 1–7. voltaic plant by reflectometry. In: Proceedings of the first world conference on
37
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
photovoltaic energy conversion, Conference record of the twenty fourth. IEEE international conference on electro information technology (EIT). IEEE; 19 May
photovoltaic specialists conference. IEEE, Vol. 1; 5 Dec 1994, pp. 846–9. 2016. p. 0307–12.
[96] Alam MK, Khan F, Johnson J, Flicker J PV ground-fault detection using spread [127] Hariharan R, Chakkarapani M, Ilango GS, Nagamani C. A method to detect pho-
spectrum time domain reflectometry (SSTDR). In: Proceedings of energy conver- tovoltaic array faults and partial shading in PV systems. IEEE J Photovolt
sion congress and exposition (ECCE). IEEE; 15 Sep 2013. p. 1015–102. 2016;6(5):1278–85.
[97] Alam MK, Khan FH, Johnson J, Flicker J PV arc-fault detection using spread [128] Bressan M, El-Basri Y, Alonso C A new method for fault detection and identifi-
spectrum time domain reflectometry (SSTDR). In: Proceedings of energy conver- cation of shadows based on electrical signature of defects. In: Proceedings of the
sion congress and exposition (ECCE). IEEE; 14 Sep 2014. p. 3294–300. 17th European conference on power electronics and applications (EPE'15 ECCE-
[98] Takashima T, Yamaguchi J, Otani K, Oozeki T, Kato K, Ishida M. Experimental Europe). IEEE; 8 Sep 2015. p. 1–8.
studies of fault location in PV module strings. Sol Energy Mater Sol Cells [129] Garoudja E, Kara K, Chouder A, Silvestre S, Kichou S Efficient fault detection and
2009;93(6):1079–82. diagnosis procedure for photovoltaic systems. In: Proceedings of the 8th interna-
[99] Chine W, Mellit A, Pavan AM, Kalogirou SA. Fault detection method for grid- tional conference on modelling, identification and control (ICMIC). IEEE; 15 Nov
connected photovoltaic plants. Renew Energy 2014;66:99–110. 2016. p. 851–6.
[100] Silvestre S, Chouder A, Karatepe E. Automatic fault detection in grid connected PV [130] Dhoke A, Sharma R, Saha TK. PV module degradation analysis and impact on
systems. Sol Energy 2013;94:119–27. settings of overcurrent protection devices. Sol Energy 2018;160:360–7.
[101] Davarifar M, Rabhi A, El Hajjaji A, Dahmane M New method for fault detection of [131] Zhao Y, Lehman B, Ball R, Mosesian J, de Palma JF Outlier detection rules for fault
PV panels in domestic applications. In: Proceedings of the 3rd international con- detection in solar photovoltaic arrays. In: Proceedings of twenty-eighth annual
ference on systems and control (ICSC). IEEE; 29 Oct 2013. p. 727–32. applied power electronics conference and exposition (APEC). IEEE; 17 Mar 2013.
[102] Bonsignore L, Davarifar M, Rabhi A, Tina GM, Elhajjaji A. Neuro-fuzzy fault de- p. 2913–20.
tection method for photovoltaic systems. Energy Procedia 2014;62:431–41. [132] Zhao Y, Balboni F, Arnaud T, Mosesian J, Ball R, Lehman B Fault experiments in a
[103] Tadj M, Benmouiza K, Cheknane A, Silvestre S. Improving the performance of PV commercial-scale PV laboratory and fault detection using local outlier factor. In:
systems by faults detection using GISTEL approach. Energy Convers Manag Proceedings of 40th photovoltaic specialist conference (PVSC). IEEE; 8 Jun 2014.
2014;80:298–304. p. 3398–403.
[104] Chine W, Mellit A, Pavan AM, Lughi V Fault diagnosis in photovoltaic arrays. In: [133] Wang G, Youn CC, Stankovic AM DC-side high impedance ground fault detection
Proceedings of international conference on clean electrical power (ICCEP). IEEE; for transformerless single-phase PV systems. In: Proceedings of North American
16 Jun 2015. p. 67–72. power symposium (NAPS). IEEE; 4 Oct 2015. p. 1–6.
[105] Platon R, Martel J, Woodruff N, Chau TY. Online fault detection in PV systems. [134] Kim IS. On-line fault detection algorithm of a photovoltaic system using wavelet
IEEE Trans Sustain Energy 2015;6(4):1200–7. transform. Sol Energy 2016;126:137–45.
[106] Hosseinzadeh M, Salmasi FR. Determination of maximum solar power under [135] Georgijevic NL, Jankovic MV, Srdic S, Radakovic Z. The detection of series arc
shading and converter faults—a prerequisite for failure-tolerant power manage- fault in photovoltaic systems based on the arc current entropy. IEEE Trans Power
ment systems. Simul Model Pract Theory 2016;62:14–30. Electron 2016;31(8):5917–30.
[107] Ventura C, Tina GM. Utility scale photovoltaic plant indices and models for on-line [136] Kim KA, Seo GS, Cho BH, Krein PT. Photovoltaic hot-spot detection for solar panel
monitoring and fault detection purposes. Electr Power Syst Res 2016;136:43–56. substrings using AC parameter characterization. IEEE Trans Power Electron
[108] Dhimish M, Holmes V. Fault detection algorithm for grid-connected photovoltaic 2016;31(2):1121–30.
plants. Sol Energy 2016;137:236–45. [137] Chen L, Li S, Wang X. Quickest fault detection in photovoltaic systems. IEEE Trans
[109] Bressan M, El Basri Y, Galeano AG, Alonso C. A shadow fault detection method Smart Grid 2016.
based on the standard error analysis of IV curves. Renew Energy [138] Yi Z, Etemadi AH. Fault detection for photovoltaic systems based on multi-re-
2016;99:1181–90. solution signal decomposition and fuzzy inference systems. IEEE Trans Smart Grid
[110] Umana A, Meliopoulos AS Detection of cell-level fault conditions within a pho- 2017;8(3):1274–83.
tovoltaic array system. In: Proceedings of transmission and distribution conference [139] Chen S, Li X PV series arc fault recognition under different working conditions
and exposition (T&D), IEEE/PES. IEEE; 3 May 2016. p. 1–5. with joint detection method. In: Proceedings of the 62nd Holm conference on
[111] Wang W, Liu AC, Chung HS, Lau RW, Zhang J, Lo AW. Fault diagnosis of photo- electrical contacts (Holm). IEEE; 9 Oct 2016. p. 25–32.
voltaic panels using dynamic current–voltage characteristics. IEEE Trans Power [140] Zhu H, Wang Z, Balog RS Real time arc fault detection in PV systems using wavelet
Electron 2016;31(2):1588–99. decomposition. In: Proceedings of 43rd photovoltaic specialists conference
[112] Andrianajaina T, Sambatra EJ, Andrianirina CB, Razafimahefa TD, Heraud N PV (PVSC). IEEE; 5 Jun 2016. p. 1761–6.
fault detection using the least squares method. In: Proceedings of international [141] Dhar S, Patnaik RK, Dash PK. Fault detection and location of photovoltaic based
conference and exposition on electrical and power engineering (EPE). IEEE; 20 Oct DC microgrid using differential protection strategy. IEEE Trans Smart Grid 2017.
2016. p. 846–51. [142] Xiong Q, Ji S, Zhu L, Zhong L, Liu Y. A novel DC arc fault detection method based
[113] Belaout A, Krim F, Mellit A Neuro-fuzzy classifier for fault detection and classifi- on electromagnetic radiation signal. IEEE Trans Plasma Sci 2017;45(3):472–8.
cation in photovoltaic module. In: Proceedings of the 8th international conference [143] Kumar BP, Ilango GS, Reddy MJ, Chilakapati N. Online fault detection and diag-
on modelling, identification and control (ICMIC). IEEE; 15 Nov 2016. p. 144–9. nosis in photovoltaic systems using wavelet packets. IEEE J Photovolt
[114] Garoudja E, Harrou F, Sun Y, Kara K, Chouder A, Silvestre S. Statistical fault de- 2018;8(1):257–65.
tection in photovoltaic systems. Sol Energy 2017;150:485–99. [144] Roy S, Alam MK, Khan F, Johnson J, Flicker J. An irradiance independent, Robust
[115] Harrou F, Sun Y, Taghezouit B, Saidi A, Hamlati ME. Reliable fault detection and ground fault detection scheme for PV arrays based on spread spectrum time do-
diagnosis of photovoltaic systems based on statistical monitoring approaches. main reflectometry (SSTDR). IEEE Trans Power Electron 2017.
Renew Energy 2018;116:22–37. [145] Zhao Y, Yang L, Lehman B, de Palma JF, Mosesian J, Lyons R Decision tree-based
[116] Dhimish M, Holmes V, Mehrdadi B, Dales M. Simultaneous fault detection algo- fault detection and classification in solar photovoltaic arrays. In: Proceedings of
rithm for grid-connected photovoltaic plants. IET Renew Power Gener twenty-seventh annual applied power electronics conference and exposition
2017;11(12):1565–75. (APEC). IEEE; 5 Feb 2012. p. 93–9.
[117] Dhimish M, Holmes V, Mehrdadi B, Dales M. Multi-layer photovoltaic fault de- [146] Akram MN, Lotfifard S. Modeling and health monitoring of DC side of photovoltaic
tection algorithm. High Volt 2017. array. IEEE Trans Sustain Energy 2015;6(4):1245–53.
[118] Shimakage T, Nishioka K, Yamane H, Nagura M, Kudo M Development of fault [147] Karatepe E, Hiyama T Controlling of artificial neural network for fault diagnosis of
detection system in PV system. In: Proceedings of 33rd international tele- photovoltaic array. In: Proceedings of the 16th international conference on in-
communications energy conference (INTELEC). IEEE; 9 Oct 2011. p. 1–5. telligent system application to power systems (ISAP). IEEE; 25 Sep 2011.p. 1–6.
[119] Xu X, Wang H, Zuo Y Method for diagnosing photovoltaic array fault in solar [148] Mekki H, Mellit A, Salhi H. Artificial neural network-based modelling and fault
photovoltaic system. In: Proceedings of Asia-Pacific power and energy engineering detection of partial shaded photovoltaic modules. Simul Model Pract Theory
conference (APPEEC). IEEE; 25 Mar 2011. p. 1–5. 2016;67:1–3.
[120] Omana M, Rossi D, Collepalumbo G, Metra C, Lombardi F Faults affecting the [149] Jiang LL, Maskell DL Automatic fault detection and diagnosis for photovoltaic
control blocks of PV arrays and techniques for their concurrent detection. In: systems using combined artificial neural network and analytical based methods.
Proceedings of international symposium on defect and fault tolerance in VLSI and In: Proceedings of international joint conference on neural networks (IJCNN).
nanotechnology systems (DFT). IEEE; 3 Oct 2012. p. 199–204. IEEE; 12 Jul 2015. p. 1–8.
[121] Gokmen N, Karatepe E, Celik B, Silvestre S. Simple diagnostic approach for de- [150] Jones CB, Stein JS, Gonzalez S, King BH Photovoltaic system fault detection and
termining of faulted PV modules in string based PV arrays. Sol Energy diagnostics using laterally primed adaptive resonance theory neural network. In:
2012;86(11):3364–77. Proceedings of the 42nd photovoltaic specialist conference (PVSC). IEEE; 14 Jun
[122] Solórzano J, Egido MA. Automatic fault diagnosis in PV systems with distributed 2015. p. 1–6.
MPPT. Energy Convers Manag 2013;76:925–34. [151] Rao KU, Parvatikar AG, Gokul S, Nitish N, Rao P A novel fault diagnostic strategy
[123] Silvestre S, da Silva MA, Chouder A, Guasch D, Karatepe E. New procedure for for PV micro grid to achieve Reliability Centered Maintenance. In: Proceedings of
fault detection in grid connected PV systems based on the evaluation of current international conference on power electronics, intelligent control and energy
and voltage indicators. Energy Convers Manag 2014;86:241–9. systems (ICPEICES). IEEE; 4 Jul 2016. p. 1–4.
[124] Lin X, Wang Y, Pedram M, Kim J, Chang N. Designing fault-tolerant photovoltaic [152] Ancuta F, Cepisca C Fault analysis possibilities for PV panels. In: Proceedings of
systems. IEEE Des Test 2014;31(3):76–84. the 3rd international youth conference on energetics (IYCE). IEEE; 7Jul 2011.
[125] Hu Y, Zhang J, Cao W, Wu J, Tian GY, Finney SJ, Kirtley JL. Online two-section PV p. 1–5.
array fault diagnosis with optimized voltage sensor locations. IEEE Trans Ind [153] Hu Y, Gao B, Song X, Tian GY, Li K, He X. Photovoltaic fault detection using a
Electron 2015;62(11):7237–46. parameter based model. Sol Energy 2013;96:96–102.
[126] Alajmi M, Abdel-Qader I Fault detection and localization in solar photovoltaic [154] Chine W, Mellit A, Lughi V, Malek A, Sulligoi G, Pavan AM. A novel fault diagnosis
arrays using the current-voltage sensing framework. In: Proceedings of technique for photovoltaic systems based on artificial neural networks. Renew
38
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
Energy 2016;90:501–12. prediction under complex conditions for a predictive energy management system.
[155] Kase R, Nishikawa S Fault detection of bypass circuit of PV module—Detection J Sol Energy Eng 2015;137(3):031015.
technology of open circuit fault location. In: Proceedings of the 19th international [184] Skoplaki E, Palyvos JA. On the temperature dependence of photovoltaic module
conference on electrical machines and systems (ICEMS), 2016. IEEE; 13 Nov 2016. electrical performance: a review of efficiency/power correlations. Sol Energy
p. 1–4. 2009;83(5):614–24.
[156] Yi Z, Etemadi AH. line-to-line fault detection for photovoltaic arrays based on [185] Mayer D, Wald L, Poissant Y, Pelland S Performance prediction of grid-connected
multiresolution signal decomposition and two-stage support vector machine. IEEE photovoltaic systems using remote sensing.
Trans Ind Electron 2017;64(11):8546–56. [186] Alonso-García MC, Ruiz JM, Herrmann W. Computer simulation of shading effects
[157] Ghali FM, Syam FA, Abdelaziz MM Analysis of interconnected configuration in PV in photovoltaic arrays. Renew Energy 2006;31(12):1986–93.
arrays under fault condition. In: proceedings of the 46th midwest symposium on [187] Karatepe E, Boztepe M, Colak M. Development of a suitable model for char-
circuits and systems. IEEE, Vol. 3; 27 Dec 2003. p. 1095–9. acterizing photovoltaic arrays with shaded solar cells. Sol Energy
[158] Fezzani A, Mahammed IH, Drid S, Chrifi-alaoui L Modeling and analysis of the 2007;81(8):977–92.
photovoltaic array faults. In: Proceedings of the 3rd international conference on [188] Kawamura H, Naka K, Yonekura N, Yamanaka S, Kawamura H, Ohno H, Naito K.
control, engineering & information technology (CEIT). IEEE; 25 May 2015. p. 1–9. Simulation of I–V characteristics of a PV module with shaded PV cells. Sol Energy
[159] Stellbogen D Use of PV circuit simulation for fault detection in PV array fields. In: Mater Sol Cells 2003;75(3):613–21.
Proceedings of conference record of the twenty third photovoltaic specialists [189] Umana A, Meliopoulos AS The extraction of photovoltaic module parameters using
conference. IEEE; 10 May 1993. p. 1302–1307. Fibonacci and Steepest Descent methods. In: Proceedings of North American
[160] Gow JA, Manning CD Development of a model for photovoltaic arrays suitable for power symposium (NAPS). IEEE; 2015 Oct 4 pp. 1–6.
use in simulation studies of solar energy conversion systems. [190] Suskis P, Galkin I Enhanced photovoltaic panel model for MATLAB-simulink en-
[161] Chowdhury S, Taylor GA, Chowdhury SP, Saha AK, Song YH Modelling, simula- vironment considering solar cell junction capacitance. In: Proceedings of the 39th
tion and performance analysis of a PV array in an embedded environment. In: annual conference of the industrial electronics society, IECON. IEEE; Nov 10 2013.
Proceedings of 42nd international universities power engineering conference, p. 1613–8.
UPEC. IEEE; 4 Sep 2007. p. 781–5. [191] Zhan ZH, Zhang J, Li Y, Chung HS. Adaptive particle swarm optimization. IEEE
[162] Gupta S, Tiwari H, Fozdar M, Chandna V Development of a two diode model for Trans Syst Man Cybern Part B (Cybern) 2009;39(6):1362–81.
photovoltaic modules suitable for use in simulation studies. In: Proceedinga of [192] Bishop JW. Computer simulation of the effects of electrical mismatches in pho-
power and energy engineering conference (APPEEC), Asia-Pacific. IEEE; 27 Mar tovoltaic cell interconnection circuits. Sol Cells 1988;25:73–89.
2012. p. 1–4. [193] Kuncheva L. Fuzzy classifier design. Springer Science & Business Media; 2000.
[163] Nishioka K, Sakitani N, Uraoka Y, Fuyuki T. Analysis of multicrystalline silicon [194] Hunter JS. The exponentially weighted moving average. J Qual Technol
solar cells by modified 3-diode equivalent circuit model taking leakage current 1986;18(4):203–10.
through periphery into consideration. Sol Energy Mater Sol Cells [195] Lucas JM, Saccucci MS. Exponentially weighted moving average control schemes:
2007;91(13):1222–7. properties and enhancements. Technometrics 1990;32(1):1–2.
[164] Suskis P, Galkin I Enhanced photovoltaic panel model for MATLAB-simulink en- [196] Nishioka K, Shimakage T, Yamane H, Kudo M, Ueda Y Evaluation of output per-
vironment considering solar cell junction capacitance. In: Proceedings of the 39th formance of various photovoltaic systems in the hokuto mega-solar project. In:
annual conference of the industrial electronics society, IECON. IEEE; 10 Nov 2013. Proceedings of 32nd international telecommunications energy conference
p. 1613–8. (INTELEC). IEEE; 6 Jun 2010. p. 1–7.
[165] Kurobe KI, Matsunami H. New two-diode model for detailed analysis of multi- [197] Orkisz M Estimating effects of individual PV panel failures on PV array output. In:
crystalline silicon solar cells. Jpn J Appl Phys 2005;44(12R):8314. Proceedings of the 16th international conference on environment and electrical
[166] Mazhari B. An improved solar cell circuit model for organic solar cells. Sol Energy engineering (EEEIC). IEEE; 7 Jun 2016. p. 1–5.
Mater Sol Cells 2006;90(7):1021–33. [198] Zhiqiang H, Li G Research and implementation of microcomputer online fault
[167] Castro F, Laudani A, Fulginei FR, Salvini A. An in-depth analysis of the modelling detection of solar array. In: Proceedings of the 4th international conference on
of organic solar cells using multiple-diode circuits. Sol Energy 2016;135:590–7. computer science & education. ICCSE'09. IEEE; 25 Jul 2009. p. 1052–5.
[168] Lumb MP, Bailey CG, Adams JG, Hillier G, Tuminello F, Elarde VC, Walters RJ [199] Orduz R, Solórzano J, Egido MÁ, Román E. Analytical study and evaluation results
Analytical drift-diffusion modeling of GaAs solar cells incorporating a back mirror. of power optimizers for distributed power conditioning in photovoltaic arrays.
In: Proceedings of the 39th photovoltaic specialists conference (PVSC). IEEE; 16 Progress Photovolt: Res Appl 2013;21(3):359–73.
Jun 2013. p. 1063–8. [200] Sánchez‐Friera P, Piliougine M, Pelaez J, Carretero J, Sidrach de Cardona M.
[169] Soon JJ, Low KS, Goh ST Multi-dimension diode photovoltaic (PV) model for Analysis of degradation mechanisms of crystalline silicon PV modules after 12
different PV cell technologies. In: Proceedings of the 23rd international sympo- years of operation in Southern Europe. Progress Photovolt: Res Appl
sium on industrial electronics (ISIE). IEEE; 1 Jun 2014. p. 2496–501. 2011;19(6):658–66.
[170] Ram JP, Babu TS, Dragicevic T, Rajasekar N. A new hybrid bee pollinator flower [201] Reis AM, Coleman NT, Marshall MW, Lehman PA, Chamberlin CE Comparison of
pollination algorithm for solar PV parameter estimation. Energy Convers Manag PV module performance before and after 11-years of field exposure. In:
2017;135:463–76. Proceedings of conference record of the twenty-ninth photovoltaic specialists
[171] Rajasekar N, Kumar NK, Venugopalan R. Bacterial foraging algorithm based solar conference. IEEE; 19 May 2002. p. 1432–1435.
PV parameter estimation. Sol Energy 2013;97:255–65. [202] International Electrotechnical Commission. Crystalline silicon terrestrial photo-
[172] Jordehi AR. Parameter estimation of solar photovoltaic (PV) cells: a review. voltaic (PV) modules—design qualification and type approval. Int Stand IEC
Renew Sustain Energy Rev 2016;61:354–71. 2005;61215(04).
[173] Pillai DS, Rajasekar N. Metaheuristic algorithms for PV parameter identification: a [203] Parretta A, Bombace M, Graditi G, Schioppo R. Optical degradation of long-term,
comprehensive review with an application to threshold setting for fault detection field-aged c-Si photovoltaic modules. Sol Energy Mater Sol Cells
in PV systems. Renew Sustain Energy Rev 2017. 2005;86(3):349–64.
[174] Kymakis E, Kalykakis S, Papazoglou TM. Performance analysis of a grid connected [204] Skoczek A, Sample T, Dunlop ED. The results of performance measurements of
photovoltaic park on the island of Crete. Energy Convers Manag field‐aged crystalline silicon photovoltaic modules. Progress Photovolt: Res Appl
2009;50(3):433–8. 2009;17(4):227–40.
[175] Davarifar M, Rabhi A, El-Hajjaji A, Dahmane M Real-time model base fault di- [205] Castaner L, Silvestre S. Modelling photovoltaic systems using PSpice. John Wiley
agnosis of PV panels using statistical signal processing. In: Proceedings of inter- and Sons; 2002.
national conference on renewable energy research and applications (ICRERA). [206] Salameh ZM, Dagher F. The effect of electrical array reconfiguration on the per-
IEEE; 20 Oct 2013. p. 599–604. formance of a PV-powered volumetric water pump. IEEE Trans Energy Convers
[176] Davarifar M, Rabhi A, Hajjaji AE, Bosche J, Pierre X. Improved real time amor- 1990;5(4):653–8.
phous pv model for fault diagnostic usage. In: Sustainability in energy and [207] Velasco-Quesada G, Guinjoan-Gispert F, Piqué-López R, Román-Lumbreras M,
buildings. Berlin, Heidelberg: Springer; 2013. p. 179–88. Conesa-Roca A. Electrical PV array reconfiguration strategy for energy extraction
[177] Jiang Y, Qahouq JA, Batarseh I Improved solar PV cell Matlab simulation model improvement in grid-connected PV systems. IEEE Trans Ind Electron
and comparison. In: Proceedings of 2010 IEEE international symposium on circuits 2009;56(11):4319–31.
and systems (ISCAS). IEEE; 30 May 2010. p. 2770–3. [208] Velasco G, Guinjoan F, Pique R Grid-connected PV systems energy extraction
[178] Delorme C, Gallo A, Olivieri J. Quick use of WEFAX images from METEOSAT to improvement by means of an Electric Array Reconfiguration (EAR) strategy: op-
determine daily solar radiation in France. Sol Energy 1992;49(3):191–7. erating principle and experimental results. In: Proceedings of power electronics
[179] Chouder A, Silvestre S. Automatic supervision and fault detection of PV systems specialists conference. PESC 2008. IEEE; Jun 15 2008. p. 1983–8.
based on power losses analysis. Energy Convers Manag 2010;51(10):1929–37. [209] Garoudja E, Kara K, Chouder A, Silvestre S Parameters extraction of photovoltaic
[180] Easwarakhanthan T, Bottin J, Bouhouch I, Boutrit C. Nonlinear minimization al- module for long-term prediction using artifical bee colony optimization. In:
gorithm for determining the solar cell parameters with microcomputers. Int J Sol Proceedings of the 3rd international conference on control, engineering & in-
Energy 1986;4(1):1–2. formation technology (CEIT). IEEE; May 25 2015. p. 1–6.
[181] Drews A, De Keizer AC, Beyer HG, Lorenz E, Betcke J, Van Sark WG, Heydenreich [210] Pearson RK. Mining imperfect data: dealing with contamination and incomplete
W, Wiemken E, Stettler S, Toggweiler P, Bofinger S. Monitoring and remote failure records. Soc Ind Appl Math 2005.
detection of grid-connected PV systems based on satellite observations. Sol Energy [211] Zhao Y, Ball R, Mosesian J, de Palma JF, Lehman B. Graph-based semi-supervised
2007;81(4):548–64. [Apr 30]. learning for fault detection and classification in solar photovoltaic arrays. IEEE
[182] Pelland S, Galanis G, Kallos G. Solar and photovoltaic forecasting through post‐- Trans Power Electron 2015;30(5):2848–58.
processing of the Global Environmental Multiscale numerical weather prediction [212] Xiao H, Xie S. Leakage current analytical model and application in single-phase
model. Progress Photovolt: Res Appl 2013;21(3):284–96. transformerless photovoltaic grid-connected inverter. IEEE Trans Electromagn
[183] Schmelas M, Feldmann T, da Costa Fernandes J, Bollin E. Photovoltaics energy Compat 2010;52(4):902–13.
39
D.S. Pillai, N. Rajasekar Renewable and Sustainable Energy Reviews 91 (2018) 18–40
[213] Aktas M, Turkmenoglu V. Wavelet-based switching faults detection in direct detection of inter-area oscillations in power systems. In: Proceedings of innovative
torque control induction motor drives. IET Sci Meas Technol 2010;4(6):303–10. smart grid technologies conference Europe (ISGT Europe), IEEE PES. IEEE; 11 Oct
[214] Kim CH, Aggarwal R. Wavelet transforms in power systems. Part 1: general in- 2010. p. 1–5.
troduction to the wavelet transforms. Power Eng J 2000;14(2):81–7. [226] Etemadi AH, Sanaye-Pasand M. High-impedance fault detection using multi-re-
[215] Gaouda AM, Salama MM, Sultan MR, Chikhani AY. Power quality detection and solution signal decomposition and adaptive neural fuzzy inference system. IET
classification using wavelet-multi resolution signal decomposition. IEEE Trans Gener Transm Distrib 2008;2(1):110–8.
Power Deliv 1999;14(4):1469–76. [227] Narasimhan SV, Basumallick N, Veena S. Introduction to wavelet transform: a
[216] Wilkinson WA, Cox MD. Discrete wavelet analysis of power system transients. signal processing approach. Alpha Science International, Ltd; 2011.
IEEE Trans Power Syst 1996;11(4):2038–44. [228] Barata JC, Hussein MS. The Moore–Penrose pseudo inverse: a tutorial review of
[217] Meyer PA. Quantum probability for probabilists. Springer; 2006. the theory. Braz J Phys 2012;42(1–2):146–65.
[218] Tsallis C. Possible generalization of Boltzmann-Gibbs statistics. J Stat Phys [229] Xiong Q, Ji S, Zhu L, Zhong L, Liu Y. A novel DC arc fault detection method based
1988;52(1):479–87. on electromagnetic radiation signal. IEEE Trans Plasma Sci 2017;45(3):472–8.
[219] Kim KA, Krein PT, Seo GS, Cho BH Photovoltaic ac parameter characterization for [230] Han J, Kamber M. Data mining concepts and techniques. second ed. 2006.
dynamic partial shading and hot spot detection. In: Proceedings of twenty-eighth [231] Yusof Y, Sayuti SH, Latif MA, Wanik MC Modeling and simulation of maximum
annual applied power electronics conference and exposition (APEC). IEEE; 17 Mar power point tracker for photovoltaic system. In: Proceedings of power and energy
2013. p. 109–15. conference, PECon, National. IEEE; 29 Nov 2004. p. 88–93.
[220] Banerjee T, Chen YC, Dominguez-Garcia AD, Veeravalli VV Power system line [232] Hiyama T, Karatepe E Feasibility of artificial neural network for maximum power
outage detection and identification—a quickest change detection approach. In: point estimation of non crystalline-Si photovoltaic modules. In: Proceedings of the
Proceedings of IEEE international conference on acoustics, speech and signal 15th international conference on intelligent system applications to power systems,
processing (ICASSP). IEEE; 4 May 2014. p. 3450–4. ISAP'09. IEEE; 8 Nov 2009. p. 1–6.
[221] Chen YC, Banerjee T, Domínguez-García AD, Veeravalli VV. Quickest line outage [233] Haykin S Neural networks, a comprehensive foundation Second Edition by
detection and identification. IEEE Trans Power Syst 2016;31(1):749–58. Prentice-Hall.
[222] Li S, Yılmaz Y, Wang X. Quickest detection of false data injection attack in wide- [234] Healy MJ, Caudell TP, Smith SD. A neural architecture for pattern sequence ver-
area smart grids. IEEE Trans Smart Grid 2015;6(6):2725–35. ification through inferencing. IEEE Trans Neural Netw 1993;4(1):9–20.
[223] Li S, Wang X. Cooperative change detection for voltage quality monitoring in [235] Dixey M. Putting reliability at the centre of maintenance. Prof Eng
smart grids. IEEE Trans Inf Forensics Secur 2016;11(1):86–99. 1993;6(6):23–5.
[224] Li S, Wang X Monitoring disturbances in smart grids using distributed sequential [236] Afefy IH. Reliability-centered maintenance methodology and application: a case
change detection. In: Proceedings of the 5th international workshop on compu- study. Engineering 2010;2(11):863.
tational advances in multi-sensor adaptive processing (CAMSAP). IEEE; 15 Dec [237] U.S.A. Department of defence, MlL-STD-1629 procedures for performing a failure
2013. p. 432–5. mode. Eff Crit Anal 1980.
[225] Sidorov D, Panasetsky D, Šmádl V Non-stationary autoregressive model for on-line
40