Professional Documents
Culture Documents
Solar Energy
journal homepage: www.elsevier.com/locate/solener
A R T I C LE I N FO A B S T R A C T
Keywords: Due to exponential growth of large-scale PV plants, automatic approaches for PV system protection are gaining
Fault detection prodigious importance. Even with the use of standard protection devices, faults in PV arrays may remain un-
PV systems detected. To address such an important issue, this paper focuses on string level monitoring to develop the
PV protection functionality of automatic fault detection, location and fault type identification. The fault detection is achieved
Fault current
through the generation of fault indicator signals called residuals and comparison with a pre-set threshold. The
String monitoring
automatic identification of fault type is achieved by the development of a procedure reliant on the variations in
the string current profiles relative to the type of fault. Finally, the location of faults is estimated through the
combination of analytical and regression expressions reliant on fault type, irradiance levels and string current
measurements. Various line-line fault cases are tested and verified using the proposed method through simu-
lations and experiments. The proposed method is experimentally evaluated for multiple fault scenarios on an
experimental setup located within an existing solar farm to emulate conditions akin to real world solar farms.
⁎
Corresponding author.
E-mail addresses: a.dhoke@uq.edu.au (A. Dhoke), rahul.sharma@uq.edu.au (R. Sharma), saha@itee.uq.edu.au (T.K. Saha).
https://doi.org/10.1016/j.solener.2020.06.019
Received 22 February 2020; Received in revised form 29 May 2020; Accepted 4 June 2020
Available online 24 June 2020
0038-092X/ © 2020 International Solar Energy Society. Published by Elsevier Ltd. All rights reserved.
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
60%
50% (a) (b)
Failure (%)
40%
30%
20%
10%
0%
Fig. 1. PV system failure statistics based on (a) location of fault and (b) cause of failure (Vargas et al., 2015).
2015; Silvestre et al., 2013; Chine et al., 2014; Platon et al., 2015; However, these methods fail to automatically locate the specific faulty
Dhimish and Holmes, 2016; Garoudja et al., 2017; Yi and Etemadi, module in a string and identify the fault type. However, main short-
2017) in PV systems. An in-depth review study on fault detection and coming in automated fault diagnosis is its inability to locate the faulty
monitoring systems for PV installations is presented in (Triki-Lahiani modules in PV strings. Therefore, it is necessary to develop approaches
et al., 2018). This study provides an overview of the different mon- that automate fault detection and diagnosis (prediction of fault location
itoring techniques used to address the major system failures. Another and identification of the fault type) of any maloperation in the solar PV
study (Daliento et al., 2017) presents the recent advances in mon- system.
itoring, diagnosis and power forecasting for PV systems. Most of the This paper examines two types of unique line-line (short circuit)
monitoring and fault detection approaches use inverter level mon- faults found in a grid-connected PV array. The first one is intra-string
itoring for the ease of data collection and processing. However, most of line-line fault that occurs within a single string and another is cross-
these approaches are incapable of locating faults within a PV string due string line-line fault which occurs between two or more strings.
to a lack of visualisation. Many fault diagnostic techniques have been Previous literature mainly studied faults in PV arrays under high irra-
proposed which are based on simulations for performance analysis diance level in (Calais et al., 2008) and (Goss et al., 2011). In these
(Kurtz, et al., 2013), statistical analysis (Davarifar et al., 2013) using cases, the fault current is usually large enough to be cleared by pro-
current and voltage measurements that is reliant on accurate modelling tection devices easily. However, unlike ‘high irradiance’ conditions,
and processing to detect the fault. Another method (Takashima et al., faults in the PV array under low irradiance (or panels with degradation)
2008) uses time domain reflectometry (TDR) to detect breakdown point tend to have a low fault current (Dhoke and Mengede, 2017) that may
in PV array which can work in offline mode only. To detect the faults remain undetected by protection. Thus, the fault current never reaches
online, methods based on the difference between simulated and mea- the minimum trip level of OCPDs. As a result, the potentially dangerous
sured results are proposed in (Hu et al., 2013; Gokmen et al., 2013; line-line fault is never cleared and remain present in the PV array
Kang et al., 2012). These methods require climatic data such as radia- (Dhoke et al., 2018).
tion temperature and require a personal computer (PC) to be connected Table 1 highlights that the key novelty of the proposed approach is
all the time with commercially available software (e.g. MATLAB or its ability to automatically perform fault location up to module level
LabVIEW). The method using adjacent string comparison is proposed in and fault type identification that are achieved with standard string
(Davarifar et al., 2013) to detect faulty string. Power loss analysis current measurements. Whilst string level monitoring is not common in
method based on the evaluation of current and voltage indicators is most existing solar installations, it is becoming more and more wide-
proposed in (Chouder and Silvestre, 2010). The method proposed in spread as the significance of PV system operation and maintenance cost
(Hu et al., 2015) relies on optimized voltage sensors and aims to locate reduction is now acknowledged.
the faults, but the cost of the required system configuration is very high The overriding objective of the proposed approach is to develop
and the approach itself lacks an automatic approach. Most of these tools for automatic diagnostics of faulty underperforming solar panels.
approaches are validated on the PV system which may not provide The approach is specifically developed for intra-string and cross-string
feasibility of real time application. Clearly, the significance of advanced line-to-line faults that are considered as the most frequently occurring
fault diagnostics in large solar farms is now well understood and many faults in solar farms (Appiah et al., 2019; Flicker and Johnson, 2013;
fault diagnosis methods have been developed for the PV systems. Han et al., 2018; Johnson and Flicker, 2013) and is reliant on the
Table 1
Comparison of the proposed method with some of the existing methods for fault detection, identification and location.
Method Monitoring level Fault detection Fault identification Fault location at module Reference
level
Parameter comparison String ✓ × × (Pei and Hao, 2019; Silvestre et al., 2014)
String comparison String ✓ × × (Spataru et al., 2015; Zhao et al., 2020)
Outlier detection String ✓ × × (Ding et al., 2018)
Power loss analysis String ✓ × × (Guerriero et al., 2017; Cristaldi et al.,
2015)
Residual generation and regression String ✓ ✓ ✓ Proposed method
expression
865
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
I1 I2 I3 In
Ia
11 21 31 n1
Inverter
F4
Grid
Va
12 22 32 n2 utility
F2 F1
GFPD
F3
PV module
1m 2m 3m nm
OCPD
866
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
Fig. 4. (a) String current outliers during intra-string fault, (b) representation of intra-string faults.
greater than the threshold. This variation in residual values can be used and satisfaction of the following two conditions:
to ascertain occurrence of faults. This same concept is presented in Condition (1):
(Flicker and Johnson, 2013) and is summarised below for completion.
In order to formulate the proposed residual-based fault detection ∃ i ∈ [1, n]such that ∀ j ∈ [1, n]
approach, let us calculate the mean string current I¯a using the string where, ri or r j ⩽ εTh
current measurement j≠i (5)
n
∑ Ii Condition (2):
i=1
I¯a = ∃ i, j ∈ [1, n] ri > εTh and r j > εTh
n (1)
where, ri = |Ii − I¯a | and r j = |I j − I¯a | (6)
where n is the number of parallel strings in an array. Based on I¯a , re-
sidual as a fault indicator signal (ri ) for each ith string can be calculated Criteria for fault identification:
as
ri = |Ii − I¯a |, ∀ i = [1, n] (2) ▪ Satisfaction of Condition (1) implies the fault type as an intra-string
fault.
In order to detect occurrence of the fault as an outlier, letεTh re- ▪ Satisfaction of Condition (2) implies the fault type as a cross-string fault.
presents a predefined fault detection threshold which governs the
boundary around ri to ascertain the existence of an outlier within the An intra-string fault is expected to result in one outlier while a cross
string current measurements. Based on (2) and the choice of εTh , fault string fault is likely to result in two or more outliers. In intra-string
detection is governed by the following law: faults, only one string can be affected by lowering the string current and
the rest of the other string currents must be normal. On other hand, in
|ri| ⩽ εTh ⇒ No fault
If ⎧ cross-string line-line faults, more than one string can be affected, and
⎨
⎩ |ri | > εTh ⇒ Fault present in ith string (3) the relative string's current varies based on the voltage of the fault lo-
cation between two strings. Based on the number of string outliers and
I µ̂ Th can be estimated using the following expression:
their currents, faults can be identified as an intra-string or as a cross-
εTh = sup ∥ri ∥2 ⎫ string fault.
under no fault condition
∀ i ∈ [1, n] ⎭ ⎬ (4) As an example of line-line faults, Fig. plots the string currents under
intra-string faults within string 1 at various module locations (Lx), x
In (4), ‘sup’ represents the supremum. Accordingly, εTh is based on ∈ [1,4] for a 4 × 4 array. For each fault location (Lx), respective current
the calculation of the greatest lower bound on the values of string magnitudes (I1) of string 1 are changing. It is clear from Fig. 4(a) that
current measurements under no-fault conditions. Examples of εTh cal- the string current magnitude is heavily dependent on the location of the
culation are shown in the simulations and experimental results sections. fault within the faulty string. Furthermore, the change in current
magnitude is seen in string 1 only while the rest of the string currents
3.2. Identification of the fault type remain the same, indicating that an intra-string fault affects the current
output of the faulty string only. Each fault location represents bypassing
The objective of the development of a fault identification procedure one PV module (see Fig. 4(b)) which results in the loss of that module’s
is to ascertain whether the fault is within a single string or across two or voltage. Since the faulted string is connected in parallel with other non-
more strings. This section presents a novel method for fault type faulted strings, the voltage across the string 1 modules increases to
identification. The fault identification method is based on the definition maintain the same voltage across all the strings. It is also observed that
867
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
I1
L5
11 21 31 41
L4
String 1 String 2
10
12 22 32 42
5
String current (I)
0 L3
-5
Outliers between two strings 13 23 33 43
-10
-15 L2
-20
0 1 2 3 4 5 6 14 24 34 44
(a) (b)
Fig. 5. String current outliers during cross-string fault between strings 1 and 2, (b) location of cross-string faults (fault location on string 2 is kept fixed while the
locations on string 1 are changed).
string 1 current magnitudes represent negative values which are due to module(s). Let fault locations at the ith and jth faulty strings be de-
the back-feed current that flows from healthy strings to faulty strings to noted as Li and Lj. The corresponding current measurements are Ii
balance the string voltages. and Ij. Based on the measurements, the relationship established
In cross-string line-line faults, string current of two or more strings using relative location and current is given by
deviates from the expected current levels. Different relative fault loca-
ΔI = Ii − I j (7)
tions are accompanied with different potential differences that result in
different current differentials for different relative fault locations on ΔL = Li − Lj (8)
different faulty strings. This phenomenon is shown in Fig. 5(a) which
shows string current levels with a cross-string fault relative to fault
As the fault location between the two strings vary and the associated
location differential. As seen in Fig. 5(a), the difference in string current
string current varies as well. As a result, if the ΔI is known, the fault
magnitudes is governed by ΔL magnitude, where ΔL is the deviation in
location of one of the strings is estimated and the fault location in
currents of faulty and non-faulty strings. For instance, the currents of
the second faulty string can be obtained using the Eq. (8). The
strings 2 are away from the normal current levels as opposed to only
procedure to calculate cross-string fault location is given in the steps
one string for intra-string faults. The change in string currents during
below:
cross-string faults are caused by the potential difference created by the
fault location. Due to that, upon the occurrence of a cross-string fault
1. Establish a relationship between ΔL and ΔI
one string comprises a greater number of PV modules than the other
string. Consequently, current flows from lower potential to higher po- ΔL = f (ΔI )
tential location which results in a change in string currents.
Fig. 5 shows the relationship between fault current magnitude and In cross-string fault, one of the outliers has a larger current than
fault location in an event of a cross-string fault between strings 1 and 2. others. The string current (Ii) changes with fault location (Li) at a given
As illustrated in Fig. 5(b), the fault on string 2 is kept fixed while the irradiation. From this observation, it is understood that the difference of
locations on string 1 are varied from L1 to L4. The corresponding results faulted string current (ΔI) is proportional to the difference in fault lo-
of the analysis are presented in Fig. 5(a). At fault location L1, string 1 cations (ΔL) given as per Eqs. (7) and (8). Therefore, this relationship
current is higher than string 2. Conversely, at location L2 - L4 the cur- can be used to formulate the fault location approach. The ΔI vs ΔL
rent of string 1 is greater than string 2. These contradictory changes characteristics are plotted for various fault combinations in 4 × 4 array
happen due to the point of fault location within the two strings. As at a given irradiance level. Fig. 6 shows ΔI vs ΔL characteristics at the
current follows the path from higher potential to lower potential cre- irradiance 1000 W/m2.
ated by the fault, the respective string current changes to high/low with 2. Using string current measurements calculate ΔL
respect to each other or vice versa. However, at L2, both string currents From Fig. 6, it is observed that ΔI and ΔL have a nearly linear re-
are equal due to the fault location being at the same location (same lationship especially in the irradiance range between 700 W/m2 to
potential). 1000 W/m2. At high irradiance levels, the linear relationship exists
regardless of the fault locations. Consequently, if the string currents are
3.3. Determination of the fault locations (Lx) known, the module mismatch percentage or number of bypassed
modules (ΔL) can be estimated using the following expression,
Once the fault is detected and identified, the next step is to de- Δ L= k × Δ I+ b (9)
termine the location of faults in one or more strings. This section
Table 2 shows the estimated values of slope (k) and intercept (b) for
presents a novel approach to estimate the location (Lx) of faulty
plotted lines in Fig. 7 at the respective irradiation. In Eq. (9), the
868
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
Table 2 given PV array the values of m and n are fixed. Furthermore, the tem-
Estimated values of slope and intercept for Fig. 7. perature (Ta) is assumed constant considering it has a weak effect on
Irradiation (W/m2) m B current, hence on the fault location. Therefore, for a given PV array the
expression (10) can be reduced to the following functional form:
1000 −0.1434 −0.0093
900 −0.1713 −0.0032 i = f (G, Ii )
L (11)
800 −0.20310 0.0491
700 −0.2415 0.1480 Our objective is to obtain a regression function f (G, Ii ) such that the
600 −0.2982 0.2779
precise location of fault within a string can be estimated once the fault
500 −0.3826 0.4419
400 −0.6379 0.4935
has occurred. For the derivation of a function f (G, Ii ) , the range of G is
taken between 400 W/m2 − 1000 W/m2. The choice of this range is
practically meaningful because in Queensland nearly 75% − 90% of
the time on a clear sky day the irradiance levels are within 400 W/m2
− 1000 W/m2.
To obtain the regression function for Eq. (11), a 3-dimensional data
set between Li-GHI-Ii is obtained through a series of simulations per-
formed using a simulation model developed in MATLAB/Simulink
(Marthworks, 2017). Various array configurations are implemented in
Simulink including 4 × 4, 6 × 3 and 15 × 4. Fig. 7 shows the data set
obtained for the 15 × 4 configuration. The data set is obtained by
running the simulation sweep for various fault locations (between 1 and
15) and solar irradiance (between 400 W/m2 − 1000 W/m2 in the steps
of 100 W/m2) combinations and the corresponding string (fault) cur-
rent levels.
In order to obtain L i as per Eq. (11), the dataset is fitted into a
regression model using MATLAB’s polynomial surface fit function
(Dhoke and Mengede, 2017) to obtain a regression expression in the
following form:
Fig. 7. Effect on string current with fault location and irradiation (Dhoke et al.,
2019). i = P1 + (P2 × GHI ) + (P3 × Ii ) + (P4 × GHI 2) + (P5 × GHI × Ii )+
L
(P6 × Ii 2) + (P7 × GHI 2 × Ii ) + (P8 × GHI × Ii 2) + (P9 × Ii3)
expression is obtained for 4 × 4 array. A similar expression can be (12)
obtained for any array size through ΔI and ΔL characteristics.
3. Estimate the fault location on one of the strings Using the simulation generated dataset, we estimate the values for
Once ΔL is calculated in step 2, estimation of the fault in either of parameters (P1 to P9). A cubic fit is selected to estimate coefficients of a
the faulty strings suffices to estimate the fault location on both the polynomial that fits a set of data in a least-squares sense. For the 15 × 4
faulty strings. Let us suppose that fault location on the ith string (Lî ) is to array configuration corresponding to Fig. 6, a cubic fit was found to be
be esimated. Since the current in string is dependent on the solar ra- the most suitable.
diation and voltage is dependent on the temperature, it is expected that Using the regression expression in Eq. (12), fault location can be
the fault location, irradiance, array size (string and module numbers) calculated at a given irradiance. The fault characteristics show that
and ambient temperature govern fault current. Therefore, the like- string currents during 600 W/m2 – 1000 W/m2 are nearly linear.
lyhood location of the fault within a string can be expressed as follows Therefore, the accuracy of fault location is best for irradiance levels
(Appiah et al., 2019): greater that 600 W/m2. Nonetheless, this has minimal effect on the
applicability of the proposed approach as most of the time on clear
i = f¯ (G, Ii , m , n, Ta)
L (10) sunny days irradiance remains above 600 W/m2.
4. Calculation of the location on any additional faulty string
i denote the estimate fault location within the ith string. For a
where L Once the fault is identified, the location for intra-sting fault (Li) can
869
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
870
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
Table 4
Verification for line-line faults for 4 × 4 PV array.
Table 5
Fault verification on 6 × 3 PV array.
Table 6
Fault validation on 15 × 4 PV array.
871
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
Table 7 Table 7 confirms the effectiveness of the proposed method under dif-
Intra-string line-line fault detection and location with variation in load. ferent load conditions.
GHI (W/ Load Detection Location
m2) (kW)
5. Experimental validation
Residual εTh Lx(actual) x(estimated)
L
Fig. 9. (a) Picture of the experimental setup, (b) Connection diagram of PV system used in fault analysis.
872
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
Table 8
Experimental results of fault detection, identification and location for different irradiance in 4 × 4 array.
Table 9
Experimental results of fault detection, identification and location for different irradiance in 6 × 3 array.
verified using various intra-string and cross-string line-line faults that values of the fault locations are rounded off to the nearest whole
are created between strings 1 and 2. The intra-string fault analysis is numbers to get the true reflection of actual fault locations.
undertaken at 700 W/m2, 850 W/m2 and 950 W/m2 irradiance levels
and cross-string fault analysis is undertaken at 750 W/m2 and 800 W/
m2 irradiance levels respectively. The threshold is calculated using Eq. 6. Conclusions
(4) for each of the irradiation during fault free conditions. The results
observed from experiments are given in Tables 8 and 9. The paper presents an approach to automatically detect, identify
The experimental validation result successfully confirms the ability and locate faulty under-performing PV modules in solar farms. The
of the proposed approach for accurate fault detection, identification of proposed approach is based on characterisation of string currents under
the fault type and locating the faults within the strings. The estimated various fault conditions thereby resulting in a distinct fault current
signature specific to occurring faults. For instance, the paper shows that
873
A. Dhoke, et al. Solar Energy 206 (2020) 864–874
pattern on string current variations under line to line faults relative to Power Eng. 5 (04), 230.
no-fault string conditions are quite distinct for intra-string and cross- De Lia, F., Pistochini, P., Bonfiglio, L., Vivoli, F., Castello, A., 2013. Training on PV: Focus
on Maintenance, Fault Research and Plant performace optimization. 27th European
string line to line faults. In the event of a cross-string line to line faults, Photovoltaic Solar Energy Conference and Exposition, Paris.
two string currents will display abnormal residuals and the relative Dhimish, M., Holmes, V., 2016. Fault detection algorithm for grid-connected photovoltaic
plants. Sol. Energy 137, 236–245.
magnitudes of string currents corresponding to faulty strings are de- Dhoke, A., Mengede, A., 2017. Challenges to overcurrent protection devices in PV array
pendent on the locations of faults on each faulty string. The experi- during winter and low irradiation conditions in Australia. IEEE Innovative Smart Grid
mental studies reveal that the levels of string current variations bear a Technologies - Asia (ISGT-Asia), pp. 1–6.
Dhoke, A., Sharma, R., Saha, T.K., 2016. Condition monitoring of a large-scale PV power
strong correlation with the fault type and fault location within one or plant in Australia. In: 2016 IEEE Power and Energy Society General Meeting
more strings. The paper demonstrates these correlations through si- (PESGM), Boston, MA (United States), pp. 1–5.
Dhoke, A., Sharma, R., Saha, T.K., 2018. PV module degradation analysis and impact on
mulation and experimental studies, and exploits these correlations to
settings of overcurrent protection devices. Sol. Energy 160, 360–367.
automate fault diagnosis through the available string current mea- Dhoke, A., Sharma, R., Saha, T.K., 2019. An approach for fault detection and location in
surements. Both simulation and experimental validations are performed solar PV systems. Sol. Energy 194, 197–208.
Ding, H., Ding, K., Zhang, J., Wang, Y., Gao, L., Li, Y., Chen, F., Shao, Z., Lai, W., 2018.
to evaluate the proposed method. The results analytically and com- Local outlier factor-based fault detection and evaluation of photovoltaic system. Sol.
prehensively demonstrate the potential of the proposed approach for Energy 164, 139–148.
accurate and reliable fault diagnosis. Flicker, J.D., Johnson, J.D., 2013. Electrical simulations of series and parallel PV arc-
faults. IEEE 39th Photovoltaic Specialists Conference PVSC, 3165–3172.
Although the proposed method is specifically targeting fault diag- Fraunhofer ISE, “Photovoltaics Report 2019”. Retrieved from www.ise.fraunhofer.de.
nosis for two types of line-line faults (intra-string and cross-string Garoudja, E., Harrou, F., Sun, Y., Kara, K., Chouder, A., Silvestre, S., 2017. Statistical fault
detection in photovoltaic systems. Sol. Energy 150, 485–499.
faults) that generally occur in PV arrays, the proposed approach high- Gokmen, N., Karatepe, E., Silvestre, S., Celik, B., Ortega, P., 2013. An efficient fault di-
lights that the relative characteristics of string currents under various agnosis method for PV systems based on operating voltage-window. Energy Convers.
fault currents can be used to automatically perform fault diagnosis in Manage. 73, 350–360.
Goss, B., Reading, C., Gottschalg, R., 2011. A review of overcurrent protection methods
solar farms. Future work will involve consideration and incorporation for solar photovoltaic DC circuits.
of the effects of gradual module degradation on the proposed approach. Guerriero, P., Piegari, L., Rizzo, R., Daliento, S., 2017. Mismatch based diagnosis of PV
fields relying on monitored string currents. Int. J. Photoenergy.
Han, F.C., Chen, Z.C., Wu, L.J., Yu, J.L., Long, C., Lin, P.J., Cheng, S.Y., Su, F.P., 2018.
Declaration of Competing Interest Online module-level fault detection of PV arrays using an improved two-stage hampel
identifier. IOP Conference Series: Earth and Environmental Science 188 (1), 012063.
Harrou, F., Nounou, M.N., Nounou, H.N., Madakyaru, M., 2015. PLS-based EWMA fault
The authors declare that they have no known competing financial detection strategy for process monitoring. J. Loss Prev. Process Ind. 36, 108–119.
interests or personal relationships that could have appeared to influ- Hu, Y., Gao, B., Song, X., Tian, G.Y., Li, K., He, X., 2013. Photovoltaic fault detection using
a parameter based model. Sol. Energy 96, 96–102.
ence the work reported in this paper. Hu, Y., Zhang, J., Cao, W., Wu, J., Tian, G.Y., Finney, S., Kirtley, J.L., 2015. Online two-
section PV array fault diagnosis with optimized voltage sensor locations. IEEE Trans.
Acknowledgements Ind. Electron. 62 (11), 7237–7246.
Jäger-Waldau, A., February 2020. Snapshot of photovoltaics. Energies 13 (4), 930.
Johnson, J.D., Flicker, J.D., 2013. Photovoltaic ground fault simulations using SPICE. No.
An International Patent Application No. PCT/AU2019/051203 is SAND2013-1394C. Sandia National Lab. (SNL-NM), Albuquerque, NM, USA.
Kang, B.K., Kim, S.T., Bae, S.H., Park, J.W., 2012. Diagnosis of output power lowering in a
submitted on the basis of the work presented in this paper (submission PV array by using the kalman-filter algorithm. IEEE Trans. Energy Convers. 27 (4),
date: 31/10/2019). This work was performed in part or in full using 885–894.
equipment and infrastructure funded by the Australian Federal Klise, G.T., Lavrova, O., Gooding, R., 2018. PV System Component Fault and Failure
Compilation and Analysis. Sandia National Lab. (SNL-NM), Albuquerque, NM (United
Government’s Department of Education AGL Solar PV Education States).
Investment Fund Research Infrastructure Project. The University of Kontges, M., Kurtz, S., Packard, C., Jahn, U., Berger, K.A., Kato, K., 2014. Review of
failures of photovoltaic modules (external final report IEA-PVPS), IEA PVPS task 13.
Queensland is the Lead Research Organization in partnership with AGL,
International Energy Agency (IEA).
First Solar and the University of New South Wales. The authors grate- Kurtz, S., et al., 2013. Analysis of photovoltaic system energy performance evaluation
fully acknowledge the Ministry of social justice and welfare, method. National Renewable Energy Lab.(NREL), Golden, CO (United States).
Marthworks, “Curve fitting Toolbox: User's Guide (r2017b)”. Retrieved on June 10, 2017
Maharashtra State, Government of India for Doctoral Fellowship (Amit from https://www.mathworks.com/help/pdf_doc/curvefit/curvefit.pdf, 2017.
Dhoke, No. DSW/Edu/2014-2015/D-IV/114). Pei, T., Hao, X., 2019. A fault detection method for photovoltaic systems based on voltage
and current observation and evaluation. Energies 12 (9), 1712.
Platon, R., Martel, J., Woodruff, N., Chau, T., 2015. Online fault detection in PV systems.
References Sustain. Energy, IEEE Trans. PP (99), 1–8.
PV in Australia Report 2017, Australian Photovoltaic Institute, August 2019. Available
online: https://iea-pvps.org/wp-content/uploads (accessed on 6 March 2020).
Akram, M.N., Lotfifard, S., 2015. Modeling and health monitoring of DC side of photo-
Silvestre, S., Chouder, A., Karatepe, E., 2013. Automatic fault detection in grid connected
voltaic array. Sustain. Energy, IEEE Trans. 6 (4), 1245–1253.
PV systems. Sol. Energy 94, 119–127.
Appiah, Albert Yaw, Zhang, Xinghua, Ayawli, Ben Beklisi Kwame, Kyeremeh, Frimpong,
Silvestre, S., da Silva, M.A., Chouder, A., Guasch, D., Karatepe, E., 2014. New procedure
2019. Review and performance evaluation of photovoltaic array fault detection and
for fault detection in grid connected PV systems based on the evaluation of current
diagnosis techniques. Int. J. Photoenergy 2019, 1–19. https://doi.org/10.1155/
and voltage indicators. Energy Convers. Manage. 86, 241–249.
2019/6953530.
Spataru, S., Dezso, S., Tamas, K., Remus, T., 2015. Monitoring and fault detection in
Calais, M., Wilmot, N., Ruscoe, A., Arteaga, O., Sharma, H., 2008. Over-current protection
photovoltaic systems based on inverter measured string IV curves. In: 31st European
in PV array installations. ISES-AP-3rd International Solar Energy Society Conference-
Photovoltaic Solar Energy Conference and Exhibition, pp. 1667–1674.
Asia Pacific Region (ISES-AP-08) Sydney.
Takashima, T., Yamaguchi, J., Ishida, M., 2008. Disconnection detection using earth
Chine, W., Mellit, A., Pavan, A.M., Kalogirou, S.A., 2014. Fault detection method for grid-
capacitance measurement in photovoltaic module string. Prog. Photovoltaics Res.
connected photovoltaic plants. Renewable Energy 66, 99–110.
Appl. 16 (8), 669–677.
Chouder, A., Silvestre, S., 2010. Automatic supervision and fault detection of PV systems
Triki-Lahiani, A., Bennani-Ben Abdelghani, A., Slama-Belkhodja, I., 2018. Fault detection
based on power losses analysis. Energy Convers. Manage. 51 (10), 1929–1937.
and monitoring systems for photovoltaic installations: A review. Renew. Sustain.
Chouder, A., Silvestre, S., Taghezouit, B., Karatepe, E., 2013. Monitoring, modelling and
Energy Rev. 82, 2680–2692.
simulation of PV systems using LabVIEW. Sol. Energy 91, 337–349.
Vargas, J.P., Goss, B., Gottschalg, R., 2015. Large scale PV systems under non-uniform
Cristaldi, L., Faifer, M., Leone, G., Vergura, S., 2015. Reference strings for statistical
and fault conditions. Sol. Energy 116, 303–313.
monitoring of the energy performance of photovoltaic fields. International
Woyte, A., Richter, M., Moser, D., Reich, N., Green, M., Mau, S., Beyer, H., 2014.
Conference on Clean Electrical Power (ICCEP), pp. 591–596.
Analytical monitoring of grid-connected photovoltaic systems - Good Practices for
S. Daliento, A., Chouder, P., Guerriero, A.M., Pavan, A., Mellit, R., Moeini, Tricoli, P.,
Monitoring and Performance Analysis. IEA PVPS T13–03, 14–15.
2017. Monitoring, diagnosis, and power forecasting for photovoltaic fields: a review.
Yi, Z., Etemadi, A.H., 2017. Fault detection for photovoltaic systems based on multi-
Int. J. Photoenergy 13, art. no. 1356851.
resolution signal decomposition and fuzzy inference systems. IEEE Trans. Smart Grid
Davarifar, M., Rabhi, A., El Hajjaji, A., Dahmane, M., 2013. New method for fault de-
8 (3), 1274–1283.
tection of PV panels in domestic applications. International Conference of Systems
Zhao, Y., Li, D., Lu, T., Lv, Q., Gu, N., Shang, L., 2020. Collaborative fault detection for
and Control (ICSC) 727–732.
large-scale photovoltaic systems. IEEE Trans. Sustainable Energy.
Davarifar, M., Rabhi, A., El Hajjaji, A., 2013. Comprehensive modulation and classifica-
tion of faults and analysis their effect in DC side of photovoltaic system. Energy
874