ArticlePDF Available

A Review for Solar Panel Fire Accident Prevention in Large-Scale PV Applications


Abstract and Figures

Due to the wide applications of solar photovoltaic (PV) technology, safe operation and maintenance of the installed solar panels become more critical as there are potential menaces such as hot spot effects and DC arcs, which may cause fire accidents to the solar panels. In order to minimize the risks of fire accidents in large scale applications of solar panels, this review focuses on the latest techniques for reducing hot spot effects and DC arcs. The risk mitigation solutions mainly focus on two aspects: structure reconfiguration and faulty diagnosis algorithm. The first is to reduce the hot spot effect by adjusting the space between two PV modules in a PV array or relocate some PV modules. The second is to detect the DC arc fault before it causes fire. There are three types of arc detection techniques, including physical analysis, neural network analysis, and wavelet detection analysis. Through these detection methods, the faulty PV cells can be found in a timely manner thereby reducing the risk of PV fire. Based on the review, some precautions to prevent solar panel related fire accidents in large-scale solar PV plants that are located adjacent to residential and commercial areas.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
(i) Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
A Review for Solar Panel Fire Accident
Prevention in Large-Scale PV Applications
Zuyu Wu1, Yihua Hu1,2 (Senior Member, IEEE), Jennifer Wen3, Fubao Zhou4, Xianming Ye5
1Department of Electronics Engineering, University of York, York, YO10 5DD, UK
2Department of Electrical Engineering, China University of Mining and Technology, Jiangsu, 221008, China
3School of Engineering, University of Warwick, Coventry, CV4 7AL, UK
4Jiangsu Key Laboratory of Fire Safety in Urban Underground Space, China University of Mining and Technology, Jiangsu, 221116, China
5Department of Electrical, Electronic, and Computer Engineering, University of Pretoria, Pretoria, 0002, South Africa
Corresponding author: Yihua Hu (
ABSTRACT Due to the wide applications of solar photovoltaic (PV) technology, safe operation and
maintenance of the installed solar panels become more critical as there are potential menaces such as hot
spot effects and DC arcs, which may cause fire accidents to the solar panels. In order to minimize the risks
of fire accidents in large scale applications of solar panels, this review focuses on the latest techniques for
reducing hot spot effects and DC arcs. The risk mitigation solutions mainly focus on two aspects: structure
reconfiguration and faulty diagnosis algorithm. The first is to reduce the hot spot effect by adjusting the
space between two PV modules in a PV array or relocate some PV modules. The second is to detect the DC
arc fault before it causes fire. There are three types of arc detection techniques, including physical analysis,
neural network analysis, and wavelet detection analysis. Through these detection methods, the faulty PV
cells can be found in a timely manner thereby reducing the risk of PV fire. Based on the review, some
precautions to prevent solar panel related fire accidents in large-scale solar PV plants that are located
adjacent to residential and commercial areas.
INDEX TERMS Photovoltaics, fire accident, solar panel, hot-spot effect, aging
Solar photovoltaic (PV) panels have been widely applied to
harness solar power for its renewable and environment-
friendly features. However, the working environment of PV
panels is usually not pleasant in practice, leading to fast aging
and degradations of power generation, and even suffering
from risks of fire accidents. According to [1], there is a 2%
probability that a fire may occur to PV arrays each year with
0.6% of the fire accidents occurring in residential areas and
3.5% of them started from some rooftop PV modules.
When the solar panels catch a fire, it not only results in
power generation reduction but also causes secondary
damage such as toxic gas emission. As shown in Figure 1,
the constituent materials of a PV panel are mostly organics.
Energy released by glass fiber, ethylene-vinyl acetate and
polyethylene terephthalate (PET) compounds in making
epoxy resin printed circuit boards is 1.012, 0.54, 0.073 MJ,
respectively based on the data from Tewarson and Quintiere
[2]. Hydrogen compounds such as HF and HCL that are toxic
are produced during the fire accident of solar panels. In 2009,
1826 PV modules with a generation capacity of 383 kW solar
PV arrays were damaged in a fire accident in California,
USA [3]. In the same year, another 15 events of solar PV
module related fire accidents were reported in Netherlands
[4]. In 2012, a solar panel related fire occurred in a
warehouse in Goch, Germany, which caused a burning area
of about 4000 m2 [3].
FIGURE 1. The structure of a PV module
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
The root cause of the solar panel related fire accident is
usually associated with a deficit in the PV system. Previous
analysis of solar panel fire events indicated that the causes of
fire can be divided into two types, i.e. arc fault and
spontaneous combustion [5-6]. The main reasons of the arc
failure include poor quality of PV modules, installation errors
and DC arc ignition back board induced by junction and
combiner boxes. Some aging solar panels, especially those
with components not meeting their specified standards, can
spontaneously ignite under high temperatures and sunlight
due to chemical reactions and hot spot effects [7].
Solar panels can be made from crystalline silicon or
amorphous. At present, the materials used for PV cells vary
in different regions [8]. For example, according to Table I,
based on the characteristics of high melting point, low
density, and good high-light performance, the crystalline
silicon is suitable for the roof-top installation in residential
areas. To avoid fire accidents, some fireproof obstacles must
be installed between two modules, which effectively prevent
the spread of fire in a large-scale PV array. Practically, more
thin-film PV modules are used in urban areas. This, along
with other technologies such as highly efficient CdTe single-
junction cells can achieve the fastest response speed in the
visible range. For example, based on the mean spectral ratio,
which is the ratio of smoky and clear irradiance in Table I,
the value of CdTe is smaller than other PV cells. It is
illustrated that the effect of smoke on CdTe is the greatest.
Meanwhile, smoke in the near-infrared and infrared ranges
has the least effect of monocrystalline silicon cells in visible
range. It has the highest response due to the thin-film
technologies (e.g., copper indium gallium selenide (CIGS)
solar cells). These results have an impact on PV fire-prone
areas [9-10]. As for the protection from fire of ground PV
array for commercial use, the installation distance between
each module can be calculated according to different PV
modules materials. TABLE I
Melting point
2.32 g/cm3
2.32 g/cm3
spectral ratio
In the large-scale PV arrays, the power generation
mismatch accelerates the aging process of the solar panels
[11] due to non-uniform patterns of shading, irradiance, and
temperature of each panel. According to [12],
approximately 51% of the PV related fire accidents is
related to installation errors or poor quality of PV modules,
which further causes cable faults on PV modules. On the
contrary, the hot-spot effect is liable for a relatively lower
percentage of the solar panel fire accidents. Low
manufacturing quality of solar panels is a major contributor
to the solar panel fire accidents. In order to reduce the risks
of field solar panels related fire accidents, this review
summarizes the cause factors and some effective fire
prevention solutions to the field solar panels. There are two
main solutions to alleviate the hot spot effect in PV arrays,
namely restructuring PV modules and reconstruction of the
distribution of PV arrays. As aged PV modules are easier to
cause DC arc shock and damage, real-time fault detection
mechanisms are helpful for preventing such damages. In
addition, solar panels must be tested against strict
engineering standards to reduce the risks of fire damage
post installation.
In the following sections, a comprehensive review will be
provided for solar panel fire accidents in large-scale PV
applications. Section II illustrates the reasons of the solar
PV related fire accidents, which include hot-spot effect, DC
arc, and other conditions. In Section III, the methods for
reducing the probability of the solar PV related fire
accidents are discussed, which include structure
reconfiguration and fault diagnosis. Section IV presents the
According to the summaries of [2, 5-7, 12, 14-33], the main
causes of PV fires are shown in Figure 2. There are 36%
fire events due to installation errors, 15% accidents because
of quality of PV modules [12]. Most fire events were found
to be caused by DC arc [18-27] due to poor quality of PV
modules, lack of drainage of PV systems, aging of
combiner box, and aging of IGBTs in inverters. In addition,
the hot spot effect should not be overlooked [14-17].
FIGURE 2. Factors lead to PV module fire accidents
A. The Hot-spot Effect
In PV modules, series connected cells are usually used.
Some PV cells suffer from partial shades from surrounding
objects, such as fallen leaves, dust accumulation, and bird
drops while other PV modules do not, hot spots may be
produced due to non-uniform power generation status
amongst the PV cells. The hot spot effect occurs if the
temperature exceeds 5% above the standard temperature in
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
a period in the standard testing condition (STC, 1000 W/m2,
25 °C). Since the performance of PV cells is different in
several cases, some shaded PV cells have obvious defects.
The hot spot effect increases the local currents and voltages
of PV modules, which results in a local temperature rise on
the PV module, causing the modules to spontaneously
ignite. Figure 3 shows a PV fire accident, which was caused
by the hot spot effect.
FIGURE 3. Hot spot effect [13]
Under the STC condition, hot spot temperature of opaque
PV modules is higher than that of semitransparent PV
modules by 23°C, which drops with an increment as far as
the numbers and areas of hot spots are concerned.
Moreover, the efficiencies of PV modules have been
predicted in the one and two hot-spot situations. For one-
hot-spot situation, the efficiencies of opaque and
semitransparent PV modules are 10.41% and 10.62%,
respectively. In the two incidents involving hot spots, the
efficiencies of the opaque and semitransparent PV modules
are 10.41% and 10.54%, respectively [14]. Hu et al. [15]
compared different degrees of shading and found that the
minor size shading would cause the temperature of the PV
panel in the shaded part to rise rapidly to cause a fire. Hu et
al. [16] tried to conditions to obtain the condition of hot
spot effect comparing different shading conditions on PV
modules. They found that different levels of impacts result
from different environments. The experimental conditions
of the irradiance and surface temperature of PV panels are
(820 W/m2, 25 °C), (740 W/m2, 22 °C), and (690 W/m2, 24
°C), respectively. The shading comparison diagram is
shown in Figure 4. For the first shading tests, three PV
panels were connected in series with one of them covered
with an opaque material to simulate the partial shades. It
was recorded by the thermal imager that a hot spot was
observed at the location of the shade. During the period of
minor shading, the I-V curve was shifted dramatically. In
Figure 4, Vm’ is the voltage of an unhealthy module, and
Varray is the voltage of the PV array. Figure 4 (b) shows the
second shading test, where a PV module was partially
covered by tissue paper to create a partial shade on the solar
panel so that certain lighting can penetrate the paper and
reach the solar panel. In the experiment, the faulty power
unit was short-circuited by a bypass diode when it cannot
generate enough current to support the load, shown as the
shift in the I-V curve. Where If is the shaded module
current, and IH is the healthy module current. As for the
third shading test shown in Figure 4 (c), three PV panels
were covered to create a large size of shade. In this case, the
shaded PV areas were short-circuited through the bypass
diode and all solar energy was converted into heat, such as
the shift of If in the I-V curve. However, a healthy PV panel
can still convert the partial incoming solar energy into
electricity, thereby decreasing the panel temperature. The
comparative results shown in Table II illustrates that the
only significant temperature increase is presented for the
case with minor shading, which was prone to generate hot
spots in PV modules.
FIGURE 4. The types of PV shading. (a) 1st shading test (b) 2nd shading
test (c) 3rd shading test [16] TABLE II
2nd Test
(740 W/m2,
3rd Test
(690 W/m2,
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
Healthy panel
Simultaneously, Vasko et al. [17] observed the hot spots
situations on PV cells, which were forward biased by a
current power supply. After 30 mins heating, the
temperature layer became non-uniform, and the hot spots
were usually generated adjacent to bus bars. When the
forward current of a PV cell exceeds a certain threshold, hot
spots will occur under the forward bias conditions. The
forward current is higher than the short circuit current in a
healthy module because the short circuit current determines
the upper limit of the module size before the hot spot
formation becomes prohibitive. At the beginning of the
electrical and temperature measurements, the voltage on the
power supply was slightly different. Besides the formation
of hot spots, the low temperature transient also caused by
the initial heating and capacitive processes. With the hot
spots appearance, the PV output voltage remained virtually
the same, and the voltage and temperature of hot spot were
linearly interrelated. Assuming that all the healthy PV
modules in a PV array have the identical parameters
concerning effective solar illumination intensity S (kW/m2),
ambient temperature Ta (°C) and total heat exchange
coefficient Upv (W/m2·K). Fault diagnosis could be
achieved based on Eq. (3), which is derived by (1) and (2).
When the module faces a fault, the calculated Upv will be
different from that of a healthy module [15].
= + ( )
pv m m a
S V I U A T - T
[ - )] -[ - ]
- ( - ) - ( - )
   
mpp mpp mpp mpp
f f H mpp mpp
e m ref e m ref
mpp mpp f f
=( - )
m m a
where Tm is the PV module temperature (°C); TH is the
healthy PV module temperature (°C); TH’ is the faulty PV
module temperature (°C); Am is the PV module area (m2); E
presents the electrical output power of the PV module (W);
If is the current of the healthy module in fault string (A); Vf
is the voltage of the healthy module in fault string (V); Tref
is the reference temperature 25 °C; Vmpp and Impp are the
voltage (V) and current (A) reached at the maximum power
point, respectively. ηe is the efficiency of the PV module at
a Tm. For a silicon PV module, the efficient temperature
coefficient is µ=0.05%/°C.
In general, aging is accelerated if the PV panel is overheat
over a long time. In addition, studies in [16] and [18]
showed that when the solar irradiance is greater than
800W/m2, the temperature difference between the
maximum temperature of the hot spot and the average
temperature of the module is about 10 °C. If fewer than 5%
modules have a temperature difference of more than 10 °C,
the PV array’s power output remains stable. Therefore,
even if there are partial shades and PV cell performance
defects, the overheating part of the PV cell is not the load
necessarily, and the hot spot effect may not occur. Even a
hot spot effect occurs, its severity is also related to multiple
factors. Since the hot spot effect is caused by a short-
circuited PV cell, the current of the PV string produces a
reverse bias. To avoid excessive reverse bias, current
crystalline silicon components generally have two or three
diodes in parallel to prevent hot spots in PV cells.
B. Cables Aging Effect
The arc is the phenomenon of glow discharge produced by
the inter electrode electromotive force breakdown medium.
Circuit and device damages are both likely to cause an arc
failure. Once a DC arc occurs, PV panels will have a high
probability to burn. As is shown in Figure 5, the arc failures
of the PV system are divided into three kinds: series arc
fault, parallel arc fault, and ground arc fault [18]. The series
arc occurs mainly due to loose device interfaces, resulting
in a small spacing, and current breakdown spacing. The
parallel arcs usually occur between phase and neutral lines,
as well as between phase lines because of broken line
insulation. The ground arc refers to arc current flowing
from a live conductor into the earth, which is usually
caused by the failure of insulation in the high-voltage phase
FIGURE 5. Three types of PV arc failures
Some researchers have observed the significant damages of
PV panel fire accidents through experiments and proposed
the corresponding protection methods to prevent such
accidents. Liao et al. [2] compared the four burning
conditions of single-sided PV panels with the irradiance of
15, 20, 30, and 40 kW/m2, respectively. The experiment
setup is shown in Figure 6. A high-power bulb is used as a
predicted source to illuminate the front of the PV panel A,
and at this time the natural combustion scenario of the PV
panel is simulated. Then, PV panel B is ignited, and the
heat transfer phenomenon of the adjacent PV panel is
simulated. Finally, make the back of the C PV panel face
up, simulating the scenario that the PV panel is ignited by
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
the flame underneath it, when an arc fault fire accident
occurs. According to the experimental results, at 15 kW/m2
irradiance, the solar panel was on fire in 200s, but at 40
kW/m2 irradiance, the solar panel was on fire in 25s. The
PV panel is prone to fire accident when the irradiance
exceeds 26 kW/m2. This is a critical environmental
condition as it takes shorter than 50s to cause a fire accident
[19]. In [20-21], when setting 10~80 kW/m2 of applied
radiation intensity to simulate firing the flame radiant heat
flow, the heat flux on the surface of the sample can be up to
70 kW/m2 [22]. In the pre-experiment, it was found that the
radiant heat flow of 30~40 kW/m2 can ignite the sample
and be safe and controllable. Theoretically, the waste
produced after a completely combustion of PV panels are
carbon dioxide and water. However, because PET decays
during combustion, its chemical bonds will be randomly
reorganized. The carbon group of the PET molecular chain
on the oxygen atom first attracts the hydrogen atom, and
then the ester bond is broken down into acids and vinyl
esters transitioned through the six-member rings state, and
these cracked products are formed after some secondary
processes [2]. Therefore, the decomposition products of
PET combustion include CO, CO2, acetaldehyde, aromatic
acids, and vinyl esters. Besides, the outdoor oxidation is the
most significant problem of ethylene-vinyl acetate film,
which is caused by ultraviolet rays and humid hot O2.
Therefore, HF, HCL, SO2, HCN and other flammable and
toxic gases are generated after the final reaction. Among
them, the hydrogen produced by HF or HCL causes
secondary damages to PV panels.
The relationship between the time of the fire and the
radiation heat flow was obtained, which is
shown in (4) [23]. Besides, the fire caused by the arc fault
from TPT, which is the membrane of backboard of a PV
module. The fire starts rapidly and becomes more intense
from the membrane.
t k c T T
where t is the ignition time (s),
is the heat flux (W/m2),
k is thermal diffusivity (W/m·K), ρ is the air density
(kg/m3), c is specific heat capacity (J/kg·K), T is the
thermal degree (°C) and T0 is the reference thermal degree
FIGURE 6. Experimental setup to simulate a fire accident of solar panels
Moreover, the increase in resistance of the components,
heating, or arcing causes the components to burn out, which
causes a fire. If any of the joints is loose, it may cause a DC
arc, and consequently causes a fire [24-25]. If the connector
is not wrapped and protected properly to prevent infiltration
of sand and dust, contact resistance of the connector will
increase. When the ground wire is not connected, the
equipment such as the combiner box lacks effective ground
protection. Once there is a virtual connection or a lightning
strike, it will cause a short circuit to ground, which not only
degrades the power generation efficiency but also causes
serious consequences such as a burning of the combiner
box. As shown in Figure 7, explosion accidents during the
combustion period in PV arrays have a large impact on the
safety of operation and maintenance personnel. The
explosion mainly come from the IGBTs and capacitors
inside the inverter [26]. The power of a capacitor explosion
can penetrate a 2 mm thick steel plate. The possible reasons
for the destruction of the combiner box and DC cabinet
include unreliable grounding, low cable insulation
resistance, bad connector contact, or the wiring disorders,
etc. [27].
FIGURE 7. Damaged combiner box by fire [28]
C. Other Conditions
PV modules may also suffering from physical damages. For
instance, the cracks of PV modules are caused by the stress
or pressure. If the welding area of the module is too small,
it will easily cause the panel to rupture over a long time.
Cracking is the main cause of fault of PV modules. These
cracks are usually not visible to naked eyes and can only be
detected through specific testing methods. All PV modules
must have certain degrees of pressure resistance to prevent
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
from being crushed. The quality of material (the choice of
glass) and the manufacturing process are the main
determinants of the PV module quality. The main reasons
for the solar panel breakage are environment conditions,
construction and installation method.
The low vacuuming temperature and foreign matter that
enter the crack will generate bubbles, which will affect
delamination and seriously cause the module to be
completely scrapped, as shown in Figure 8. Component
delamination is a serious problem because it allows
moisture to penetrate, which will lead to catastrophic
failure. At this point, the broken components on the panel
need to be replaced. When moisture penetrates the
protective layer of the solar module and contacts within the
internal circuit, it seriously accelerates the degradation
process of a PV module, which eventually leads to
catastrophic consequences for the module and the entire PV
system [29]. Gluing is caused by poor quality products and
materials. Over time, the backplane sometimes turns yellow
or brown. This is a chemical reaction between the inferior
materials and sunlight. Once it begins to change color,
ethylene-vinyl acetate will continue to change from its
original state, inevitably causing damage to the material
Without good drainage measures on the roof-top, it is easy
to cause water accumulation throughout the year. It not
only leads to a decrease in PV efficiency, but also cause the
aging and corrosion of cables, which may lead to fire
accidents. For a ground PV array, the impact of rainwater
may cause soil erosion, landslides, etc., so that the PV
panels are seriously damaged [31].
FIGURE 8. PV module crack [32]
Quality of Solar panels must be guaranteed by proper
regulations. PV modules have to pass the test of UL 61730-
2 PV Module Safety Assessment Part 2: Test
Requirements” [33], with a fire rating of C (basic fire proof
rating). The components installed on buildings should at
least reach the rating of C, and the price of PV modules
with different fire proof ratings varies significantly.
Components installed on existing roofs should be subjected
to barrier tests and flame spread tests. Components used for
roofing materials should be subjected to other subsequent
test materials specified in UL 790 Standard for Standard
Test Methods for Fire Tests of Roof Coverings [34]. There
is no international standard for the combustion performance
testing methods and judgment rules of modules on different
buildings. The industry standard JG/T 492-2016General
Technical Requirements for Building Photovoltaic
Modules [35] stipulates that PV modules should meet the
flammability rating requirements of building materials or
building modules in alternative locations and meet the
requirements of GB8624 “Combustion Performance” [36].
Relevant regulations on building materials, products, and
their product classification, the fire resistance test methods
and measurement rules need to comply with the provisions
of GB15763.1 “Building Safety Glass Part 1: Fireproof
Glass”, GB/T 12513 Fire-resistant test method for glass-
encrusted components and GB/T 9978.1 Fire resistance
test method of building components” [37-39].
To sum up, based on the above-mentioned PV production
and installation standards, it can be found that the fire
safety of PV-building integration is related the design of PV
modules, and certification of the PV façade elements. The
combination of good quality PV modules with a design-safe
PV system can solve many of the safety issues observed so
Depending on different fire-causing factors in the PV array,
this section summarizes existing different solutions for
different factors. Existing approaches to avoid solar PV fire
accidents mainly include preventive actions. The preventive
actions include array recombination and detection
algorithm research. The studies [40-50] illustrate the
reconfiguration of PV modules or PV arrays, and the
studies [51-78] introduce algorithm to detect the faulty PV
FIGURE 9. Detection methods for PV fires accidents
A. Preventive maintenance action in PV array
In PV arrays, shades and dust accumulations are
unavoidable, which are also the biggest threats to the safety
of PV arrays. Therefore, some preventive maintenance
actions such as conducting a periodical cleaning can be
very effective in slowing the aging process of PV
components and mitigating the hot spot effect.
There are currently two styles of solar panel installation:
ground mounted and roof-top mounted. The surrounding
environmental conditions, equipment conditions, and
temperature changes of the project location need to be
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
concerned for the ground mounted PV arrays [36]. Due to
the influence of turbulent kinetic energy (TKE) among the
modules, the soiling on the surface of the module must be
uneven, resulting in the hot-spot effect and PV module fire
accidents. It is necessary to establish a flexible inspection
and cleaning mechanism or use a data collection system to
decide whether unplanned maintenance is necessary to
reduce the risk of fire in different environments. However,
if the distance between any two PV panels in the array is
too far or too close, the PV array’s generation capacity will
be reduced. As shown Figure 10, the spacing D between
two PV panels should be large enough to avoiding shading
effect, which is selected according to latitude, time angle,
etc. The latitude angle (φ) of the winter solstice is (-23.45
°), and the time angle (ω) corresponding to 9:00 am is 45 °
[41-42]. In this case, not only the optimal photoelectric
conversion efficiency can be guaranteed, but also the TKE
value can be obtained to avoid the dust deposition.
Therefore, calculating the distance between two panels
according to (5) can obtain the most suitable distances
between PV panels.
FIGURE 10. Dust deposition on a PV array
( 1)
cos tan[sin (sin sin cos cos cos )]
 
where D is the distance between obstacles (m), A is the
azimuth of the sun (°), φ represents the latitude (°), δ means
the declination (°), ω is the time angle (°) and H is the PV
array height difference (m).
Considering that the rooftop buildings are in close contact
with people, the following factors need to be noted: 1)
whether it can be avoided by string arrangement design or
equipment technology improvement to personnel injured by
the high voltage of the DC line in the event of a fire; 2) plan
the location of the roof upper and lower channels and
electrical equipment according to the meteorological data of
the project location to reduce the time of power-off;
Enough firefighting passages are provided to ensure rapid
passage during a rescue. At the same time, the roof array
distribution map is marked at the entrance of the bottom of
the passage, and the opening and closing points of the
power lines are marked. The marking should be easy to
identify and well-marked to prevent fires. It can be cut off
quickly; and 3) the module arrangement includes both
horizontal and vertical arrangements, and the corresponding
purlin arrangement also has two directions. When the
module is arranged horizontally, the purlins are arranged
vertically as shown in Figure 11. In this case, due to the
chimney effect, the fire spreads faster than arrays with
vertically arranged components [43-45].
Overall, strictly controlling the entry threshold of
construction units, paying attention to environmental risks
during the initial site selection, standardize cable
connection construction, and establishing a reasonable
operation and maintenance system and cycle according to
the actual conditions of different projects can effectively
reduce hidden dangers. By improving the technology and
considering the design and training of the roof owner and
local fire department, the impact of the fire can be reduced.
That is to say, through comprehensive management before,
during and after the accident, the loss can be minimized or
avoided [46].
FIGURE 11. Solution to prevent PV fires on roof-top PV array
The impact of dust reduction on PV panels is enormous,
both for the ground or rooftop mounted PV arrays.
Formulas (6) is used to estimate dust flux around the PV
array, and CFD simulation can accurately calculate the
annual dust drop and dust distribution of a PV array, and
thus can get a suitable cleaning cycle and cleaning method
for the local PV array. Proper cleaning can effectively
reduce the fire probability of PV arrays.
( / ) ( ( / ))
(1 )(1 )
In z z In z z
F aEc u
g u u
 
where E is the erosion factor, α is the sand blasting
efficiency, c is the empirical proportionality constant, g is
the gravitational acceleration (m/s2) , ρα is the air density
(kg/m3), u* is the friction velocity (m/s). κ is a constant
obtained empirically (about 0.35 for turbulent flow), and z0
is the roughness length (m).
Hot spots occur when the PV module is partially blocked,
and part of the solar cell string becomes a reverse bias and
dissipates energy in the form of heat. If the solar cell
consumes more power than the maximum power of the PV
cell, which maintain the maximum power under hot spot
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
conditions, the PV cell will be completely damaged and
open-circuit. To protect the series PV cell, the bypass
diodes are added on the PV cell string [47].
K. Kim proposed the first hot-spot mitigation technique that
using bypass diodes to reconfigure PV modules [48]. The
model structure is shown in Figure 12 (a). In the research,
K. Kim shaded 1 of a 24-cell string, and found that a bypass
diode imposes 0.5 V across the substring. However, there is
still current passing through the shaded PV cell. Actually,
the bypass diode can be treat as a load, which will not
generate power. By using Kirchhoff’s Voltage Law, the
reverse voltage in the circuit can be describe in Eq. (7).
Once hot spot is detected, there are two approaches to
mitigate the potential risks. For short PV string (2~3 cells),
the traditional bypass diodes are more effective on reducing
the probability of hot spots effect. For long PV string, a low
reverse-breakdown PV cell limits the power dissipation in
the hot spotting time. It is an effective prevention method if
the power dissipation can be managed without damaging
the cell.
( 1)
V N V V 
where VR is reverse voltage (V), VF is voltage drop (V), and
VD is the diode voltage (V). (N-1) means the voltage drop
from node 1 to node N.
The advantages of this technique is to reduce the
temperature of solar cells in hot spots. Meanwhile, the
probability of hot spots is also reduced for longer PV
Based on the traditional bypass diode, S. Daliento proposed
a modified bypass diode reconfiguration, namely, an ON-
OFF MOSFET for PV modules in a hot-spot scenario [49],
which is shown in Figure 12 (b). This method is applicable
to any PV module, which composed of series connected
cells. When the PV panel is partially shaded, this solution
can significantly reduce the hot spot temperature by
transferring the reverse voltage of the normal PV cells to
the MOSFET of series connected in each sub-panel. To
conclude, when the gate-source voltage (Vgs) is high, the
MOSFET is short circuited. When Vgs is low, there is a
significant drain-source voltage drop VDS of MOSFET. The
formula is shown below:
( 1)
V N V V V  
where VDS is the MOSFET drainsource voltage drop (V).
This method was verified by testing the reduction of hot
spots temperature of polycrystalline silicon and
monocrystalline silicon PV modules, which cooled down to
about 20 °C and 24 °C, respectively.
Based on the single ON-OFF MOSFET switch circuit, M.
Dhimish proposed a double MOSFET switch circuit, which
is more effective to mitigate the hot spots effect [50]. The
model is shown in Figure 12 (c). The switch 1 is connected
in series with the PV cells, and the general state is “on”.
When hot spot situation is occurred, switch 1 will open to
further alleviate the hot spot effect. The switch 2 is in
parallel connection with the PV cells, and the general state
is “off”. When the PV string is open, it will open to
circulate current. To ensure the health of the PV module,
switch 2 is controlled by 16F877A microcontroller and
activated twice every three hours. Because M. Dhimish
found that three hours is the maximum allowable duration
before the hot spot reappears in the PV cells, and the
number of activations is determined by thermal image
analysis. As for the 16F877A, it is a microcontroller-based
system that prevents hot-spot operation using open-circuit
PV modules. This method not only reduces the heat spot
temperature by 17 °C, but also increases the output power
by 3.8%.
Simultaneously, P. Guerriero proposed a new bypass diode
circuit, which is an evolution circuit from S. Daliento [49].
The diagram is shown in Figure 12 (d) [51]. In the circuit,
the drain-source voltage drop of MOSFET M1 supplies
power to the TLC555 digital oscillator, and its output
voltage drives MOSFET M2. Therefore, as long as M1
works normally, the oscillator is turned off, its output is
low, and M2 is also turned off. When a part of the PV cells
is blocked, the drainsource voltage drop of M1 increases,
and the oscillator turns on and begins to provide an output
signal that alternates between high and low. The output
signal remains high for approximately 97% of the time.
During this time interval, M2 is on, so M1 remains off.
Conversely, if there is no longer partial occlusion, M1 is
turned on, its drainsource voltage drop is decreased, and
the oscillator is turned off, returning to normal operating
This method can reduce the hot spot temperature to 50 °C
and increase the output power by 8% in a shadow-shaded
scenario. Different from others, this solution address the
rising in temperature of shaded cells completely.
Meanwhile, the oscillator will not generates more power on
bypass events, due to the oscillator is sleeping in the rest
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
FIGURE 12. Reconfiguration of PV string. (a) bypass diode circuit (b)
ON-OFF MOSFET circuit (c) 16F977A microcontroller circuit (d) TCL555
microcontroller circuit
By changing the structure of the PV string, as well as by
some controllers, the probability of hot-spot effect can be
effectively reduced. This method not only reduces the risk
of a PV array, but also increases the power of PV output.
B. Fault Diagnosis
In 2011, the U.S. Insurer Laboratory (UL) launched UL
Standard 1699B draft [53], which is the DC arc detection
standard of circuit safety outline of DC arc fault protection
for the PV systems [54]. At present, numerous methods
could detect the arc fault of PV systems: physical analysis
(clustering method) [55-58], Fast Fourier Transform
(frequency domain analysis) [59-63], time domain analysis
[64-67], wavelet detection (multi-resolution analysis) [68-
77], and Artificial Intelligence method (neural networks,
support vector machines, fuzzy logic systems, etc.) [78-
In the event of an arc failure, the heat, arc, noise, or
electromagnetic signals will be emitted. The physical
analysis is based on the physical properties of sound, light,
and radiation are detected by cluster method. As for the
famous and widely use physic-based model, Myer arc
model is suitable for low current arcs [55], which assumes
that thermal causes power loss, and the formula is shown
( 1)
g dt P
where g is arc conductance (S), iarc is arc current (A), P is
the static cooling power (W), and τ is the arc time constant
determined empirically (s).
In addition, Peng et al. used fuzzy logic to indicate
clustering to detect arc failure [56]. The mold maximum
value of the electromagnetic radiation signal of the fault arc
after noise reduction is selected as the fault criterion. In
[57], the Hilbert antenna is used to measure the
electromagnetic radiation signal of the DC arc, the
frequency of the electromagnetic radiation signal, the pulse
interval, and the pulse cluster duration as the basis for the
failure. Physical-based detection methods install devices in
local locations in the system, making it easier to locate fault
locations [58]. However, because these models involve
many parameters, the operation is complex and is not easy
to be implemented.
Fourier Transform is a classical frequency domain-based
method, and it is recommended to carry out fault detection
in the frequency band of 1 to 100 kHz [59-60]. The time of
the Fast Fourier Transform (FFT) detection method is less
than 16 ms. It effectively disconnects the arc from the
inverter in the DC micro grid. While, this algorithm may
not effective at the converter startup. In this case, the time
domain changes dramatically and the size of the high-
frequency content in the frequency domain increases like an
arc failure leading to unnecessary tripping [61-62].
The FFT transformation of single current mutation and
electromagnetic radiation waveform is carried out and its
spectral characteristics are analyzed. The spectral
characteristics of current and electromagnetic radiation
signals are similar, with the largest frequencies as 13MHz.
The electromagnetic radiation field is proportional to the
current rise rate, at the beginning of the current steep rise
edge, the inductor of the arc is close to 0. Estimated
maximum amplitude of arc electromagnetic radiation
spectrum is based on (10) [63]:
where ε is the dielectric constant of the air (F/m), and ρ is
the arc resistivity (kg/m3).
According to (8), the frequency with the largest amplitude
in the electromagnetic radiation spectrum is only related to
the arc resistance and the dielectric constant in the air. The
resistivity of arcs generated by different inter-polar
distances and electrode diameters may vary, and the
frequency of electromagnetic radiation in DC arcs may be
different. Therefore, the pulse interval, characteristic
frequency, and duration of the arc electromagnetic radiation
signal can detect DC arc failure as feature parameters.
The advantage of time domain analysis is intuitive and
accurate. The time domain representation of the system
output can be obtained from differential equations or
transfer functions. In [64-65], the accuracy rate of fault
detection in a PV module detected by Minimum Covariance
Determinant (MCD) estimator under STC is 98%, and the
false alarm rate is 0.01%. This method is to operate the
voltage and current of different PV modules into the MCD
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
estimator at the same time instant. Then, the distribution of
the I-V curve to the centerline of each PV module can be
used to detect arc faults. The MCD estimator can be
determined as (11).
(( ) ( ))
i i S S i S
med x C x
are estimates of sample mean and
covariance matrix computed using the MCD estimator, and
xi is a data subset.
In [66], Schimpf et al. used Finite Impulse Response (FIR)
estimator to detect the arc fault. The idea of this method is
that when the arc detector is integrated into the PV module,
the detector can only measure and monitor the PV current
and the PV voltage. Due to the need for shunt resistors, Hall
sensors or current transformers, the only signal used as the
arc detector input is the PV voltage. The arc voltages
measured on the PV module various significantly according
to their position in the system. The operation of FIR
estimator fault detection is that first passing the input signal
through a bandpass filter whose cut-off frequencies are 1
kHz and 7.5 kHz. The estimator then compares the current
signal value to the previous value, and when the difference
is 0, the system is fault-free.
In [67], Yao et al. found that the selection of time window
length will impact the current waveform pattern. The
research shows that time domain analysis, although simple,
is very effective in identifying arc failures. Because it has
long enough time to ensure the randomness of the test.
At present, wavelet analysis is the mainstream detection
method, which is gradually multi-scale refinement of signal
functions through telescopic translation operation, and
finally reach the high frequency time segmentation, low
frequency subdivision, so as to focus on any details of the
signal [68-71]. According to the fault signal, it sets the
motion threshold of the fault alarm device in normal state
and in different value range, thus solving the difficult
problem of Fourier transformation. Wu et al. [72] selected
the db4 wavelet for wavelet decomposition, selected the
energy value of the wavelet high-frequency component as
the fault standard, and used the reliable value between the
normal state and the fault state as the fault alarm threshold.
Meanwhile, Lu et al. [73] selected the standard deviation as
the characteristic in the time domain, took the energy of
each band after the db5 wavelet decomposition as the
frequency domain feature, constructed the feature plane,
and divided the fault critical line within the feature plane to
detect the arc. The maximum signal and wave detail are
determined by experiments. The variance and model values
of the numbers are the three time-frequency domain
standards, and time domain-based measurements are
proposed. Mix the condition with the arc fault of the
frequency domain, and the judgment of this method has a
single method with high precision and reliability, which
further reduces the error rate and suppression rate of the
detection method [74]. The accuracy of wavelet
decomposition fault detection is 100% [67][75].
According to the basic principle of time-domain emission
method [76], the relative position of the fault point and the
measuring point can be calculated as:
where v is the wave speed in the cable (m/s); τ is the signal
of time-delay value in the fault.
For a row wave, if the distance of propagation L along the
cable within a cycle time T, the propagation speed of the
wave is v, then, it can be obtained that:
When the transmission line loss is very small or the test
signal is high frequency, the wave speed can be derived as:
 
 
 
where c is the speed of light, which is 3 x 108 m/s; µr is the
relative magnetic guide coefficient of the medium around
the cable at high frequencies; εr is the relative dielectric
constant of the medium around the cable at high
According to (14), the transmission speed of the pulse wave
in the cable is not related to the structure, length, conductor
material. It only depends on the relative magnetic
conductivity and relative dielectric constant of the cable
insulated medium. For cables made of different conductor
materials, the insulation medium is the same and the signal
travels at the same speed inside it.
This method fills the blank of arc fault detection and
positioning on the DC bus in the PV system, and effectively
prevents accidents caused by arc failure. Because the
detection signal of this method has sharp self-correlation, it
can have the good anti-jamming ability and high accuracy
in the on-line detection and positioning of DC bus arc fault
In recent years, artificial neural networks (ANN), support
vector machines (SVM), fuzzy logic, and other intelligence
algorithms have replaced thresholds to decide whether there
is arc fault.
The ANN aims to obtain the model through learning, and
use the model to predict the desired target value. In the field
of arc detection, the position of DC arc can be detected by
using the data of neural network. He et al. in [78] uses an
RBF neural network to judge arc fault, but it is easy to local
optimization and slow training. The study [79] uses a
genetic algorithm optimized BP neural network to judge arc
fault. The ANN method is fast and accurate for arc
detection [80].
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
The arc detection neural network model is shown in Figure
13, uses a three-tier structure [81], where P is the input
matrix; i, j, and k represent the number of nodes at each
layer respectively; wij is the weight between the implied
layer j node and the output layer i node, and wjk is the
weight between the node k of the output layer and the node
j of the implied layer. The implied layer activation function
selects the S-type activation function, and the output layer
activation function selects the linear activation function.
(a) Input layer: Input layer nodes are related to the
number of input data. The input to the model is the
12th to 31st harmonics after the FFT, so the
junction of the input layer is 20.
(b) Implied layer: Implied layer nodes are not fixed
and can be adjusted as needed. Currently, there is
no universal way to determine the number of
implied layer nodes. If the number of nodes is too
small, the network performance is poor or cannot
be trained, if too much selection, although the
error can be reduced, but will increase the network
training time, easy to fall into the local minimum
point and not reach the optimal solution. The
determination of the number of implied layer
nodes is obtained by the formula (15) [82].
(c) Output layer: The output layer only needs one
node, the output with 0 and 1 respectively to
represent the arc-free and arc-less.
n n n
 
where n is the implied layer junction, n1 is the input layer
junction, n0 is the output node, and β is the constant
between 1 and 10.
According to (16) and combined with the results of a large
number of experiments, it is found that the training effect is
best when the implied layer node points take 14.
FIGURE 13. Arc detection neural network model [81]
In addition, the studies [83] and [84] use the SVM
algorithm to extract the mean current and high-frequency
components from the time-frequency domain. Fault criteria
is used to train the model, and the obtained model can
classify whether an arc fault occurs.
SVM is a better supervised learning algorithm. This
algorithm is used to solve the separation hyperplane
problem that can divide the training data set normally and
has a very large geometric interval. As shown in the Figure
14, all of the “circle” means training data, among them, the
red circle is the support vector. “Wx+b=0” means the
separation hyperplane. Actually, there are countless
hyperplanes corresponding to linearly separable data sets.
Among them, the separation hyperplane with the largest
geometric interval is unique. Compared with ANN, SVM
searches the global minimum data during training, while
ANN will only search the local minimum data. And the
performance of SVM is highly related to the quality of
training data.
FIGURE 14. An diagram of SVM trained samples [74]
In [85] and [86], the authors used fuzzy logic system to
detect the arc fault in the PV array. The accuracy of this
method is increased up to 98.8%. The operation of the arc
detection system based on fuzzy logic is: First, input the
initial signal to the fuzzification process. Then, use
predefined rules to classify arc faults and normal operation.
It should be mentioned that the rules in fuzzy systems are
designed based on the fault modes and mechanisms.
C. Discussion
The method of fire prevention and detection of PV Arrays
can be summarized as the optimal distance method (ground
mounted PV array), obstacle-adding method (roof-top
mounted PV array), and reconfiguration of PV components,
physical analysis, frequency domain analysis, time domain
analysis, wavelet detection, and the artificial intelligence
algorithm. The advantages and disadvantages of these
methods are shown in Table III. Due to the increasing fault
cases, there are many data-base can be used in the future.
Therefore, the artificial intelligence methods will be
concerned popular in the future.
Based on these methods, the isolation device can be added
to PV arrays with fireproof materials, and the alarm system
can be set up according to the intelligent algorithm to
identify the DC arc failure, thus minimizing the probability
of a PV fire. In addition, the safety training of the
firefighters is essential due to the large amount of toxic
gases produced by PV combustion [87].
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
Adopted technique
Distance of each PV
panel (ground mounted
PV modules)
Design the optimal
spacing between each
Avoid the hot-spot
effect while ensuring
maximum power
capacity of the PV
Error in the amount of
dust accumulation on the
surface of PV panels
Not mention
Obstacle of each PV
panel (roof-top mounted
PV modules)
The baffle is used to
block the air flow
between each panel
Blocking the airflow
between PV panels
reduces the flame burn
Increased roof load, and
the rescue of firefighters
made it more difficult
Not mention
Structure analysis
Add bypass diode or
MOSFET in the circuit
Reducing hot-spot
effect and improve the
power efficiency
Increased cost of PV
Not mention
Physical analysis
Detect arc faults
through physical
High accuracy on small
collections of data with
less than 200 data
These models involve
many parameters, the
operation is complex and
is not easy to implement
in simulation
Frequency domain
Detect arc faults by
using FFT in the
frequency domain
Fast and high universal
This algorithm may not
work properly at the
inverter or converter
startup. It is easily to
Below 90%
Time domain analysis
Detect arc faults
through estimators in
the time domain
Intuitive, high
accuracy, and easy to
Constrained by time
Wavelet detection
Detect the arc faults
through the wavelet at
time-frequency domain
Effective and directly
Limited by vibration
diagnostic and analytical
instruments, resolution,
and analysis software
Artificial intelligence
Detect the position of
DC arc by using ANN,
SVM, and fuzzy logic
High accuracy, easy,
and convenient
Need a huge data-base
About 99%
The safety of PV power generation and PV arrays is
receiving increasing attention, especially the need to reduce
the possibility of fire and timely maintenance. The hot spot
effect and aging of PV panels were found responsible in
previous fire accidents can be caused by the dust density
around the PV array, the ambient temperature, and the
material structure of the PV array. Preventive solutions to
the fire accident can be distinguished into solar panel
reconfiguration and fire fault detection algorithm. The
advantages of reconfiguration of PV modules include
reducing hot spot and improving power efficiency.
Meanwhile, the advantage of the fire fault detection
algorithm is to detect faulty position accurately.
In order to reduce the probability of PV fire accident, there
are technical specifications to comply. Firstly, the PV
module needs to pass the UL 790 Safety Standard for
Roofing Material Fire Test combustion and flame spread
test. Secondly, the inverter should be designed without
fuses to avoid fire caused by DC side faults. The inverter
internal transformer, PCB board and other internal
components prone to high temperature should be made of
non-combustible or non-combustible materials. Thirdly, the
internal components of the junction box, control equipment,
and power distribution equipment should be made of non-
combustible materials. Fourthly, all cables are required
flame retardant coating and made of low smoke, and low
toxicity materials. Fifthly, fire-proof sealing measures
should be applied to holes, such as cable inlets and outlets
of power distribution equipment in houses, equipment inlet
holes, cable inlets and outlets of junction boxes, cable
penetration holes, cable trenches, and cable trench
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
In addition to research on the mechanism and prevention of
PV fires, it is also necessary to consider fire safety issues of
PV-building integration. In order to improve the safety of
fire prevention and extinguishing of PV systems, it is basal
to conduct fire risk investigation and hazard assessment.
Test and evaluate the combustion properties and fire
resistance of PV modules. Secondly, considering the impact
on building safety, it is advised to conduct a comprehensive
risk assessment for potential failure units of PV building
integration. Design fire separation facilities and use
fireproof materials to reduce losses caused by fire
accidents. Thirdly, realize the management
intelligentization of electrical fire monitoring and early
warning, and strengthen the investigation of hidden fire
hazards of the equipment. Specifically, the fire prevention
and control system can automatically identify and eliminate
fire risks. For example, set up an appropriate automatic fire
alarm system, intelligent protection against DC arc, and
intelligent blocking components. Finally, it is also critical
to strengthen the daily fire supervision and management,
and regularly hold the fire safety training on PV power
This study is supported by the UK Royal Academy of
Engineering under the Grant No.: IAPP17/18.
[1] P. D. Moskowitz, V. M. Fthenakis, “Toxic materials released from
photovoltaic modules during fires: Health risks”, Solar Cells, Volume 29,
Issue 1, 1990, Pages 63-71, ISSN 0379-6787.
[2] B. Liao, L. Yang, X. Ju, Y. Peng, Y. Gao, Experimental study on
burning and toxicity hazards of a PET laminated photovoltaic panel,
Solar Energy Materials and Solar Cells, Volume 206, 2020, 110295, ISSN
[3] J. Sipe, "Development of Fire Mitigation Solutions for Photovoltaic
(PV) Systems Installed on Building Roofs", July 2016.
tml, 20/02/2013.
[5] G. Manzini, P. Gramazio, S. Guastella, C. Liciotti, G. L. Baffoni, “The
Fire Risk in Photovoltaic Installations - Test Protocols For Fire Behavior
of PV Modules, Energy Procedia, Volume 82, 2015, Pages 752-758,
ISSN 1876-6102.
[6] S. S. Nair, A Survey report of the firefighters on fire hazards of PV
fire, 2018 IEEE International Conference on System, Computation,
Automation and Networking (ICSCA), July 2018.
[7] W. John, K. Sarah, How Can We Make PV Modules Safer”,
Conference Record of the IEEE Photovoltaic Specialists Conference. June
[8] Chr. Lamnatou, G. Notton, D. Chemisana, C. Cristofari, Storage
systems for building-integrated photovoltaic (BIPV) and building-
integrated photovoltaic/thermal (BIPVT) installations: Environmental
profile and other aspects, Science of The Total Environment, Volume 699,
2020, 134269, ISSN 0048-9697.
[9] M. Perry, A. Troccoli, “Impact of a fire burn on solar irradiance and
PV power, Solar Energy, Volume 114, 2015, Pages 167-173, ISSN 0038-
[10] G. N. Tiwari, R. K. Mishra, S. C. Solanki, Photovoltaic modules and
their applications: A review on thermal modelling, Applied Energy,
Volume 88, Issue 7, 2011, Pages 2287-2304, ISSN 0306-2619.
[11] D. Prince Winston, S. Kumaravel, B. Praveen Kumar, S.
Devakirubakaran, Performance improvement of solar PV array topologies
during various partial shading conditions, Solar Energy, Volume 196,
2020, Pages 228-242, ISSN 0038-092X.
[12] C. Coonick, Fire and Solar PV Systems Investigations and
Evidence”, breNational Solar Centre, 11th May 2018. P100874-1004,
Issue 2.9.
[14] P. Rajput, G.N. Tiwari, O.S. Sastry, “Thermal modelling and
experimental validation of hot spot in crystalline silicon photovoltaic
modules for real outdoor condition, Solar Energy, Volume 139, 2016,
Pages 569-580, ISSN 0038-092X.
[15] Y. Hu, W. Cao, J. Ma, S. J. Finney and D. Li, “Identifying PV
Module Mismatch Faults by a Thermography-Based Temperature
Distribution Analysis”, IEEE Transactions on Device and Materials
Reliability, vol. 14, no. 4, pp. 951-960, Dec. 2014, doi:
[16] Y. Hu, B. Gao, X. Song, G. Tian, K. Li, X. He, “Photovoltaic fault
detection using a parameter based model, Solar Energy, Volume 96,
2013, Pages 96-102, ISSN 0038-092X.
[17] A. C. Vasko, A. Vijh, V. G. Karpov, Hot spots spontaneously
emerging in thin film photovoltaics, Solar Energy, Volume 108, 2014,
Pages 264-273, ISSN 0038-092X.
[18] N. L. Georgijevic, M. V. Jankovic, S. Srdic and Z. Radakovic, “The
Detection of Series Arc Fault in Photovoltaic Systems Based on the Arc
Current Entropy”, IEEE Transactions on Power Electronics, vol. 31, no. 8,
pp. 5917-5930, Aug. 2016, doi: 10.1109/TPEL.2015.2489759.
[19] B. Lecouvet, M. Sclavons, S. Bourbigot, C. Bailly, Thermal and
flammability properties of polyethersulfone/halloysite nanocomposites
prepared by melt compounding, Polymer Degradation and Stability,
Volume 98, Issue 10, 2013, Pages 1993-2004, ISSN 0141-3910.
[20] A. Ito, T. Kashiwagi, Characterization of flame spread over PMMA
using holographic interferometry sample orientation effects, Combustion
and Flame, Volume 71, Issue 2, 1988, Pages 189-204, ISSN 0010-2180.
[21] P. A. Beaulieu, N. A. Dembsey, Effect of oxygen on flame heat flux
in horizontal and vertical orientations, Fire Safety Journal, Volume 43,
Issue 6, 2008, Pages 410-428, ISSN 0379-7112.
[22] M.B. Ayani, J.A. Esfahani, R. Mehrabian, Downward flame spread
over PMMA sheets in quiescent air: Experimental and theoretical studies,
Fire Safety Journal, Volume 41, Issue 2, 2006, Pages 164-169, SSN 0379-
[23] M. A. Delichatsios, Piloted ignition times, critical heat fluxes and
mass loss rates at reduced oxygen atmospheres, Fire Safety Journal,
Volume 40, Issue 3, 2005, Pages 197-212, ISSN 0379-7112.
[24] M.C. Falvo, S. Capparella, Safety issues in PV systems: Design
choices for a secure fault detection and for preventing fire risk, Case
Studies in Fire Safety, Volume 3, 2015, Pages 1-16, ISSN 2214-398X.
[25] K. Murata, T. Yagiura, K. Takeda, M. Tanaka, S. Kiyama, New type
of photovoltaic module integrated with roofing material (highly fire-
resistant PV tile), Solar Energy Materials and Solar Cells, Volume 75,
Issues 34, 2003, Pages 647-653.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
[26] K. Chen, C. Huang, J. He, Fault detection, classification and location
for transmission lines and distribution systems: a review on the methods”,
High Voltage, vol. 1, no. 1, pp. 25-33, 4 2016, doi:
[27] L. Chen, P. Wang, L. Xu, “Novel detection method for DC series arc
faults by using morphological filtering”, The Journal of China Universities
of Posts and Telecommunications, Volume 22, Issue 5, 2015, Pages 84-91,
ISSN 1005-8885.
[29] S. Lee, C. Wu, K. Hung, “A Study on Fire Risks to Firefighters in the
Building with Photovoltaic System”, Proceedings of the 2017 IEEE
International Conference on Applied System Innovation, 2017.
[30] J. K. Hastings, M. A. Juds, C. J. Luebke and B. Pahl, "A study of
ignition time for materials exposed to DC arcing in PV systems," 2011
37th IEEE Photovoltaic Specialists Conference, Seattle, WA, 2011, pp.
[31] R. Tommasini, E. Pons, F. Palamara, C. Turturici, P. Colella, “Risk of
electrocution during fire suppression activities involving photovoltaic
systems”, Fire Safety Journal, Volume 67, 2014, Pages 35-41, ISSN 0379-
[33] UL 61730-2, “PV Modul e Safety Assessment Part 2: Test
Requirements”, 2017.
[34] UL 790, “Safety Standards for Roof Fire Tests”, 2018.
[35] JG/T 492, “General Technical Requirements for Building
Photovoltaic Modules”, 2016.
[36] GB8624, “Combustion Performance”, 2006.
[37] Y. Zheng, Analysis on fire safety of integrated photovoltaic building
system. Fire Science and Technology. 2018. (in Chinese)
[38] L. Mazziotti, et al. “Fire risk related to the use of PV systems in
building facades”, Matec Web of Conferences, Volume 46, 2016, 05001,
Page 9.
[39] H. Häberlin, F. Volkenborn, M. Halfmann, et al. Determination of
fire safety risks at PV systems and development of risk minimization
measures, 2011, http://www.pv-
[40] Z. Zhang, J. Wu, L. Wang, F. Liu, P. Jia, L. Dai, Y. Lu, T. Bian, The
analysis on simulation and invalidation of hot-spot temperature
distribution in micro-defective crystalline silicon solar cells, Renewable
Energy, Volume 147, Part 1, 2020, Pages 2218-2228, ISSN 0960-1481.
[41] Z. Wu, Z. Zhou, A. Mohammed. Time-effective Dust Deposition
Analysis of PV Modules Based on Finite Element Simulation for
Candidate Site Determination”, IEEE Access, 2020, PP. 1-1.
[42] Z. Wu, W. Li, S. Kuka and M. Alkahtani, “Analysis of Dust
Deposition on PV Arrays by CFD Simulation,” IECON 2019 - 45th
Annual Conference of the IEEE Industrial Electronics Society, Lisbon,
Portugal, 2019, pp. 5439-5443.
[43] N. G. Dhere, N. S. Shiradkar, “Fire hazard and other safety concerns
of photovoltaic systems”, Journal of Photonics for Energy, 2012.
[44] F. Bayrak, H. F. Oztop, Effects of static and dynamic shading on
thermodynamic and electrical performance for photovoltaic panels,
Applied Thermal Engineering, Volume 169, 2020, 114900, ISSN 1359-
[45] X. Ju, et al. “Impact of flat roof–integrated solar photovoltaic
installation mode on building fire safety”, Fire and Materials, Volume 43,
Issue 7, 2019,
[46] P. Dunbabin, “Fire and Solar PV Systems – Recommendations for the
Fire and Rescue Services”, Scinence and Innovation, 2017.
[47] S. Silvestre, A. Boronat, A. Chouder, “Study of bypass diodes
configuration on PV modules”, Applied Energy, Volume 86, Issue 9, 2009,
Pages 1632-1640, ISSN 0306-2619.
[48] K. A. Kim, P. T. Krein, “Reexamination of Photovoltaic Hot Spotting
to Show Inadequacy of the Bypa ss Diode”, IEEE Journal of Photovoltaics,
vol. 5, no. 5, pp. 1435-1441, Sept. 2015, doi:
[49] S. Daliento, F. D. Napoli, P. Guerriero, V. d’Alessandro, “A modified
bypass circuit for improved hot spot reliability of solar panels subject to
partial shading”, Solar Energy, Volume 134, 2016, Pages 211-218, ISSN
[50] M. Dhimish, V. Holmes, P. Mather, M. Sibley, “Novel hot spot
mitigation technique to enhance photovoltaic solar panels output power
performance”, Solar Energy Materials and Solar Cells, Volume 179,
2018, Pages 72-79, ISSN 0927-0248.
[51] P. Guerriero, P. Tricoli, S. Daliento, “A bypass circuit for avoiding
the hot spot in PV modules”, Solar Energy, Volume 181, 2019, Pages 430-
438, ISSN 0038-092X.
[52] M. Dhimish, G. Badran, Current limiter circuit to avoid photovoltaic
mismatch conditions including hot-spots and shading, Renewable Energy,
Volume 145, 2020, Pages 2201-2216, ISSN 0960-1481.
[53] Outline of investigation for photovoltaic (PV) DC arc-fault circuit
protection Underwriters Laboratories (2011)
[54] G. Bao, X. Gao, R. Jiang, K. Huang, "A Novel Differential High-
Frequency Current Transformer Sensor for Series Arc Fault Detection",
Sensors, 2019.
[55] N. Gustavsson, “Evaluation and simulation of black -box arc models
for high-voltage circuit breakers”, Linköping, Sweden: Linköping
University, 2004.
[56] X. Peng, J. Lai, “Application of Clustering Analysis on Typical
Power Consumption Profile Analysis”. Power System Protection and
Control, 2014. (in Chinese)
[57] C. Yao, P. Qiao, P. Chen, Y. Chen, “Fractal antenna of 0.4~1 GHz for
UHF monitoring of partial discharge in electrical equipment”. High
Voltage Engineering, 2014, 40. 2285-2291. 10.13336/j.1003-
[58] M. C. Damas, R. T. Robiscoe, “Detection of radio -frequency signals
emitted by an arc-discharge”, Journal of Applied Physics, 1988, Volume
64, Issue 2 > 10.1063/1.341971.
[59] J. Johnson, S. Kuszmaul, W. Bower, D. Schoenwald, Using PV
module and line frequency response data to create robust arc fault
detectors”. Proceedings of the 26th european photovoltaic solar energy
conference, 2011. p. 374550.
[60] J. Johnson, D. Pahl, C. Luebke, T. Pier, T. Miller, J. Strauch, et al.
Photovoltaic DC arc fault detector testing at sandia national laboratories”.
Proceedings of the 37th IEEE photovoltaic specialist conference, 2011 p.
[61] S. Chae, J. Park, S. Oh, “Series DC Arc Fault Detection Algorithm for
DC Microgrids Using Relative Magnitude Comparison”, IEEE Journal of
Emerging and Selected Topics in Power Electronics, vol. 4, no. 4, pp.
1270-1278, Dec. 2016, doi: 10.1109/JESTPE.2016.2592186.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
[62] B. Novak, “Implementing arc detection in solar applications:
achieving compliance with the new UL 1699B Standard”. Texas
Instrument, 2012.
[63] S. Zhao, R. Zhang, J. Du, “Photovoltaic series Arc Fault Detection
Utilizing Wavelet Tranform”, Journal of Huaqiao University. 2017. (in
[64] H. Braun et al., “Signal processing for fault detection in photovoltaic
arrays”, 2012 IEEE International Conference on Acoustics, Speech and
Signal Processing (ICASSP), Kyoto, 2012, pp. 1681-1684, doi:
[65] S. Buddha et al., “Signal processing for photovoltaic applications”,
2012 IEEE International Conference on Emerging Signal Processing
Applications, Las Vegas, NV, 2012, pp. 115-118, doi:
[66] F. Schimpf and L. E. Norum, “Recognition of electric arcing in the
DC-wiring of photovoltaic systems”, INTELEC 2009 - 31st International
Telecommunications Energy Conference, Incheon, 2009, pp. 1-6, doi:
[67] X. Yao, L. Herrera, S. Ji, K. Zou, J. Wang, “Characteristic Study and
Time-Domain Discrete- Wavelet-Transform Based Hybrid Detection of
Series DC Arc Faults”, IEEE Transactions on Power Electronics, vol. 29,
no. 6, pp. 3103-3115, June 2014, doi: 10.1109/TPEL.2013.2273292.
[68] H. Zhu, Z. Wang, R. S. Balog, “Real time arc fault detection in PV
systems using wavelet decomposition”, 2016 IEEE 43rd Photovoltaic
Specialists Conference (PVSC), Portland, OR, 2016, pp. 1761-1766, doi:
[69] Z. Wang, R. S. Balog, “Arc fault and flash signal analysis in DC
distribution systems using wavelet transformation”. IEEE Transactions on
Smart Grid, 2015, 6(4): 1955-1963.
[70] H. Lan, R. Zhang, “Study on the feature extraction of fault arc sound
signal based on wavelet analysis”. Proceedings of the CSU-EPSA, 2008,
20(4): 57-62.
[71] M. Rabla, E. Tisserand, P, Schweitzer, “Arc fault analysis and
localisation by cross-correlation in 270 V DC”. Proceedings of IEEE Holm
Conference on Electrical Contacts. Newport, USA: IEEE, 2013: 1-6.
[72] C. Wu, W. Xu, Z. Li, L. Xu. “Arc Fault Type Identification and
Circuit Protection in Photovoltaic System”. Proceeding of the CSEE. 2017.
(in chinese)
[73] Q. Lu, T. Wang, B. He, T. Ru, D. Chen. "A new series arc fault
identification method based on wavelet transform", IECON 2017 - 43rd
Annual Conference of the IEEE Industrial Electronics Society, 2017.
[74] C. Wu, Y. Hu, Z. Li, W. Fei, “On-line Detection and Location of DC
Bus Arc Faults in PV Systems Based on SSTDR”. Proceeding of the
CSEE. 2020. (in chinese)
[75] Y. Guo, L. Wang, L. Wu, B. Jiang, “Wavelet packet analysis applied
in detection of low-voltage DC arc fault”, Proceedings of the 4th IEEE
conference on industrial electronics and applications, 2009 p. 40136.
[76] C. Luebke, T. Pier, B. Pahl, D. Breig, J. Zuercher, "Field test results
of DC arc fault detection on residential and utility scale PV arrays," 2011
37th IEEE Photovoltaic Specialists Conference, Seattle, WA, 2011, pp.
[77] C. Wu, X. Feng, T. Yuan, Z. Li, Y. Zhong, “Photovoltaic arc fault
detection method based on bp neural network”, Acta energiae solaris
sinica, 2016, 37. 2958-2964.
[78] Y. He, Q. Xu, S. Yang, B. Yu, “A Power Load Probability Density
Forecasting Method Based on RBF Neural Network Quantile Regression”.
Proceedings of the CSEE. 2013. (in Chinese)
[79] X. Ding, H. Zhu, H. Yin, Y. Wang, “DC arc fault detection based on
machine learning method”. Transducer and Microsystem Technologies,
2017. (in Chinese)
[80] J. A. Momoh, R. Button, “Design and analysis of aerospace arcing
faults using fast fourier transformation and artificial neural network”.
Proceeding of IEEE power and energy society general meeting, 2003, p.
[81] X. Lu, P. Lin, S. Cheng, Y. Lin, Z. Chen, L. Wu, Q. Zheng, “Fault
diagnosis for photovoltaic array based on convolutional neural network
and electrical time series graph”, Energy Conversion and Management,
Volume 196, 2019, Pages 950-965, ISSN 0196-8904.
[82] G. Zhang, “Intelligent control system and application”. China Electric
Power Press, 2007. (in Chinese)
[83] Z. Wang, R.S. Balog, “Arc fault and flash detection in photovoltaic
systems using wavelet transform and support vector machines”.
Proceedings of the IEEE 43rd photovoltaic specialists conference, 2016. p.
[84] K. Yang, R. Zhang, J. Yang, C. Liu, S. Chen, F. Zhang, “A novel arc
fault detector for early detection of electrical fires”. Sensor, 16 (2016), pp.
[85] B. Grichting, J. Goette and M. Jacomet, “Cascaded fuzzy logic based
arc fault detection in photovoltaic applications”, 2015 International
Conference on Clean Electrical Power (ICCEP), Taormina, 2015, pp. 178-
183, doi: 10.1109/ICCEP.2015.7177620.
[86] Mahmoud Dhimish, Violeta Holmes, Bruce Mehrdadi, Mark Dales,
Peter Mather, “Photovoltaic fault detection algorithm based on theoretical
curves modelling and fuzzy classification system”, Energy, Volume 140,
Part 1, 2017, Pages 276-290, ISSN 0360-5442.
[87] S. Martin, “Fire safety and solar electric and photovoltaic systems”,
Fire, Volume 107, Issue 1363, 2013, Page 33.
Zuyu Wu was born in Shannxi province in
P.R. China, on December 05, 1990. He received
the B.Eng. and the M.Eng degree in electrical
engineering from Newcastle University, Newcastle,
UK, in 2011 and 2015, respectively. Currently, he
is a PhD student in the University of York. His
research interests include photovoltaics soiling
management and MPPT of PV power generation.
Yihua Hu (M’13-SM’15) received the B.S.
degree in electrical motor drives in 2003, and the
Ph.D. degree in power electronics and drives in
2011 from China University of Mining and
Technology. Between 2011 and 2013, he was with
the College of Electrical Engineering, Zhejiang
University as a Postdoctoral Fellow. Between 2013
and 2015, he worked as a Research Associate at the
power electronics and motor drive group, the
University of Strathclyde. Currently, he is a
Lecturer at the Department of Electrical Engineering and Electronics,
University of Liverpool (UoL). He has published 65 papers in IEEE
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
Transactions journals. His research interests include renewable generation,
power electronics converters & control, electric vehicle, more electric
ship/aircraft, smart energy system and non-destructive test technology. He
is the associate editor of IET Renewable Power Generation, IET Intelligent
Transport Systems and Power Electronics and Drives
Jennifer Wen was born in Zhizhong County
of Sichuan Province, China. She studied
Mechanical Engineering at Shanghai Jiao Tong
University from 1980 to 1984. Upon graduation,
she won one of the only three fully funded
scholarships among more than 200 graduates in
her department for study abroad. This has
supported her to conduct PhD research on
condensation heat transfer at Queen Mary College,
University of London from 1985 to 1988. Her
professional career started in 1988 firstly with Computational Dynamics
Ltd (now CD-Adapco) where she worked on the development of STAR-
CD, a general purpose computational fluid dynamics (CFD) code. She left
CD-Adapco to join the Watson House Research Centre of the former
British Gas plc in 1991 working on the numerical study of multi-
dimensional laminar flames with detailed chemistry. Jennifer returned to
academia in 1993 initially with London South Bank University and moved
to Kingston University in 1998 as Reader and awarded the Professorial
title after 18 months. She established and led the world-class Centre for
Fire and Explosion Studies at Kingston University for over 10 years. She
was also the Director of Research for the former School of Engineering
and then Faculty of Engineering at Kingston University for over 10 years.
Fubao Zhou was born in July 1976, Nanjing,
Jiangsu Province. In June 2003, he received his
Ph.D. in Engineering from China Mining
University. Since 2011, he has been the Deputy
Director of the Internet of Things (Perceived Mine)
Research Center of China Mining University, the
Executive Dean of the School of Safety
Engineering, the Dean of the School of Safety
Engineering, and the Director of the Office of
Talent Work. In April 2019, he became a member
of the Standing Committee and Vice President of
the Party Committee of China Mining University.
He is mainly engaged in mine disaster prevention and utilization of
resources, occupational health and public safety research. He presided over
more than 30 scientific research projects such as the “13th Five-Year Plan”
Key Research and Development Program, the National 111 Program
Innovation and Intelligence Base, the National Science Foundation for
Distinguished Young People, the National Natural Science Foundation's
Key (Coal Joint) Project, and the Ministry of Education's Innovation Team
Development Plan. He achieved the second prize of national scientific and
technological achievements, and the first prize of the provincial and
ministerial scientific and technological progress. More than 50 patents for
authorized national inventions, 4 pieces of software copyright, 1 academic
monographs, and 2 editor-in-chief of teaching materials had been
Xianming Ye received his BEng and MEng
degrees in the Department of Automation, Wuhan
University, China in 2008 and 2010, respectively.
He completed his PhD degree in Electrical
Engineering from the University of Pretoria in
2015, where he conducted research on the design
of optimal measurement and verification strategies
for the national energy efficiency and demand side
management programme at the Center of New
Energy Systems. He is currently a Senior Lecturer
in the Department of Electrical, Electronic and Computer Engineering,
University of Pretoria, South Africa. He is also a Certified Measurement
and Verification Professional. His research expertise lies in the areas of
energy efficiency and demand-side management, building and industry
energy system modelling and optimisation, renewable energy, microgrids,
P2P energy sharing, battery management systems, and electric vehicles.
He is currently a guest editor of IEEE Access and associate editor of the
journal IET Renewable Power G eneration.”
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.3010212, IEEE Access
... These findings are closely related to improper installation practices as a significant contributing factor to PV fire hazards. 16,17 When PV modules catch fire, it results in a reduction of energy generation, 18 toxic gas releases, [18][19][20] property damage, 9,[18][19][20] and even casualties. 9,19 The two major fire safety concerns in BAPV are (a) the large DC system on a building that increased the probability of ignition 21,22 and (b) a change in fire dynamic scenario due to the alteration of existing roof construction by trapping heat near the roof, causing the fire to spread faster. ...
... These findings are closely related to improper installation practices as a significant contributing factor to PV fire hazards. 16,17 When PV modules catch fire, it results in a reduction of energy generation, 18 toxic gas releases, [18][19][20] property damage, 9,[18][19][20] and even casualties. 9,19 The two major fire safety concerns in BAPV are (a) the large DC system on a building that increased the probability of ignition 21,22 and (b) a change in fire dynamic scenario due to the alteration of existing roof construction by trapping heat near the roof, causing the fire to spread faster. ...
... These findings are closely related to improper installation practices as a significant contributing factor to PV fire hazards. 16,17 When PV modules catch fire, it results in a reduction of energy generation, 18 toxic gas releases, [18][19][20] property damage, 9,[18][19][20] and even casualties. 9,19 The two major fire safety concerns in BAPV are (a) the large DC system on a building that increased the probability of ignition 21,22 and (b) a change in fire dynamic scenario due to the alteration of existing roof construction by trapping heat near the roof, causing the fire to spread faster. ...
With the rapid growth of the worldwide photovoltaic (PV) installation, the number of fire incidents involving PV systems also shows an increasing trend. Several studies revealed that installing PV systems on the rooftop has introduced an additional fire risk to the building. Therefore, risk assessment is required to identify the possible cause of fire initiation involving PV systems and subsequently provide the solar industry with fire risk information regarding PV faults. A BowTie analysis of rooftop grid‐connected PV systems was conducted, where initiation of ignition was determined as the hazard and PV fires as the loss event. Four threats in the BowTie analysis were identified using fault tree analysis, that is, arc fault, ground fault, hotspot effect at PV modules, and overheating. Arc fault contributes the most to PV fire incidents, while poor installation of PV systems was found to be the primary underlying cause of all PV fault scenarios. The main factor is due to lack of fire safety knowledge and negligence behavior of the installers. The consequences of PV fires in the BowTie diagram were investigated from the event tree analysis. Twelve possible outcomes were identified and regrouped to five consequences, that is, respiratory poisons, electrical shock, fall from heights, asset damage, and fire propagation. The evaluation of the consequences of PV fire shows that electrical shock poses a very high risk to the surrounding people, including firefighters. Additional measures are proposed to reduce the impact of electric shock.
... The government of Japan has also issued a warning against rooftop PV fire risk as the country has chronicled 172 PV fire-related accidents from 2008 to 2017 (Emiliano, 2019). When solar panels catch fire, it does result in not only a reduction of energy generation but also the release of toxic gases, property damage and even death (Wu, Hu, Wen, Zhou & Ye, 2020). As a PV system is a subset of an electrical system, it is subjected to the two leading causes of electrical fire ignitionoverload and short circuit (Cancelliere, 2016;Mazziotti et al., 2016). ...
... Regardless of the fire origin, installing a PV electrical generation system on a building worsens the pre-existent fire risk level and increases the fire severity compared to a building without a PV system (Cancelliere, 2016). The primary causes reported contribute to PV fire accidents are the phenomena of hot spots in PV modules, overheating of PV components and the occurrence of direct current (DC) arc-fault at PV components (Armijo, Johnson, Hibbs, & Fresquez, 2014;Bataille et al., 2019;Cancelliere, 2016;Coonick et al., 2018;Fiorentini, Marmo, Danzi & Puccia, 2016;Mazziotti et al., 2016;Wu et al., 2020). These factors are directly associated with or ripple effects caused by errors during the design or poor practices during PV system installation. ...
... Numerous circumstances may stop the solar cell from generating power, for example, incompatibilities or misalignment of PV modules, damaged PV modules and unpredictable shading at the modules, such as accumulation of dirt, falling leaf, tall vegetation or bird droppings. In addition, poor connection quality during the installation of the PV system could produce excessive thermal energy within that might be sufficient to ignite a spontaneous fire under high temperature or even cause DC electrical arcing fault to any PV system components (Cancelliere, 2016;Coonick et al., 2018;Fiorentini et al., 2016;Mazziotti et al., 2016;Wu et al., 2020). Although incidents of DC arc-fault are considerably rare in PV system installation, there were incidents of DC arc-fault reported in the US and Germany, resulting in significant fire and damage to the PV system (Armijo, Johnson, Hibbs, & Fresquez, 2014). ...
Numerous photovoltaic (PV) fire incidents are caused by overheating of PV system components, direct current (DC) arc-fault or hot spot phenomenon. These causes happen mainly due to poor installation practices by the installers. Many PV system installation guides do not emphasise much on the fire hazard during installation. As the PV system is becoming increasingly popular nowadays, it is crucial to establish a specific and comprehensive guideline pertaining to fire safe installation of the system. Therefore, this paper aims to assess and incorporate such fire safety practices from all PV installation guidelines that are publicly accessible. A total of 40 PV installation publications have been systematically reviewed and classified into two categories – design consideration and installation stage. The analysis pointed out a compilation of fire safety practices during PV system installation focusing on residential rooftop applications from the reviewed publications. Although DC isolators have been reported as the top component of PV fire causes, many guidelines do not emphasise the fire hazards involved as well as the things that should and should not be done during installation. The inclusion of a fire safety checklist is suggested as part of the installation guideline.
... Currents flowing through an element with a lowered internal resistance significantly heat it up and cause the appearance of local overheating zones, the so-called "hot spots" (Dhimish et al., 2018). As a result, this can lead to the failure of the entire solar array as a whole and even to a fire (Wu et al., 2020). ...
Full-text available
Purpose The paper aims to substantiate optimization directions of resettable fuses parameters to protect solar arrays from overcurrent. Design/methodology/approach The method of modeling the electrophysical characteristics of resettable fuses is used. Findings Resettable fuses currently produced are of little use for protecting photovoltaic cells (PVC) in solar arrays from overcurrent. The volume fraction of the conductive filler should be about 0.15, near the percolation threshold. Thus, reducing the resistance by increasing the amount of filler is not possible. The matrix of the composite should consist of a material with a significant proportion of the crystalline phase to ensure a sharp increase in the composite's volume near the melting point. Using a polymer with a lower melting point instead of polyethylene can reduce the power required to switch a resettable fuses. Originality/value The possibility of using resettable fuses based on polymer composite materials with a positive temperature coefficient of resistance to protect photovoltaic solar cells from current overloads is considered. Modeling of the electrophysical characteristics of modern industrial fuses of this type based on polyethylene-nanocarbon composites has been carried out. The limits of their applicability for the protection of photovoltaic solar cells are analyzed. On the basis of the obtained results, the optimization directions of the resettable fuses parameters for use in the protection circuits of PVC of solar array are determined.
... Importantly, the occurrence of arc faults must be carefully considered to avoid fire accidents in PV systems. 24 Figure 2e shows the temperature maps of the investigated devices (replicas 1) at the maximum reverse bias reached by the devices at their rear side before the drop of their reverse current toward near-zero values (electrical disconnection). In PSC-A, the heat linearly propagated over the length of the active area, determining the fracture line of the device (see also Video S2). ...
Full-text available
Perovskite solar cells have reached certified power conversion efficiency over 25%, enabling the realization of efficient large-area modules and even solar farms. It is therefore essential to deal with technical aspects, including the reverse-bias operation and hot-spot effects, which are crucial for the practical implementation of any photovoltaic technology. Here, we analyze the reverse bias (from 2.5 to 30 V) and temperature behavior of mesoscopic cells through infrared thermal imaging coupled with current density measurements. We show that the occurrence of local heating (hot-spots) and arc faults, caused by local shunts, must be considered during cell and module designing.
... It is also necessary to keep the PV system in safe operating mode with proper designing, fault identification, and fault diagnostic system. In [25], discusses the proper design of the PV system. In this case, two equal power-generating PV system of 100MWp of each is analyzed in similar environmental conditions. ...
Full-text available
The performance of the solar photovoltaic system is affected by the unpredictable phenomenon of partial shading. This causes the mismatch losses that suppress the power generation of healthy PV modules in it. The objective of the proposed work in this paper is to bring out the maximum power from each PV module in the PV array by reducing the mismatch losses. A new array configuration method is proposed in this paper, which follows the screw pattern in the row formation. Each PV row is created with distinct PV modules from the rows of the conventional array configuration. This proposed work allows the PV system to operate with minimum mismatch losses by even shade dispersion over the PV array. The proper mathematical expression with all necessary constraints was derived for the array formation of the proposed work. The output analysis is been validated in the simulation of the 9X9 PV array in MATLAB/Simulink®. The mismatch loss generation and output power enhancement are measured and compared with the various conventional array configuration methods under six kinds of partial shading patterns. The proposed array configuration is 40% more efficient than the conventional series-parallel array configuration and also performs better than the total cross-tied and sudoku puzzle pattern methods. The shade dispersion rate of the proposed array configuration has highly reduced the mismatch losses in the PV system and hence, it improves the power output.
... A majority of the literature on PV-related fires focuses on fault detection, fire behaviour analysis, and the safety of installers and first responders [21,22,[24][25][26][27][28]. In a recent study, Wu et al. [29] presented a review on PV fire prevention techniques in which it was concluded that fault diagnosis and configuration of PV panels is key to fire prevention in large-scale PV systems. Currently, there is no model that can predict the number of fire incidents due to BAPV systems. ...
A fault tree analysis of fires related to photovoltaic (PV) systems was made with a focus of understanding the failure rate of the electric components. The failure rate of different components of these systems was calculated from data obtained from reports, research studies, and fire incident statistics of four countries. The results explain the significant causes of fire on the component level and various failure patterns resulting in PV-related fires. The qualitative analysis identified seven major events that led to incidents caused by a PV-related ignition source, with electrical arcing being the main cause of fires. This finding is highly related to the imprudent installation practices due to negligence and low awareness of the fire risk associated with PV systems by installers. The quantitative results show that 33% of the PV fire incidents are due to unknown or unrelated ignition sources, indicating that great focus should be given to mitigate the consequences caused by PV-related fires. The PV module, isolator, inverter, and connector are the major PV system components that are highly responsible for the ignition of PV-related fires, with the connector being the prime contributor in 17% of the PV-related fires. Finally, the quantitative analysis established an annual fire incident frequency of 0.0293 fires per MW. The results enable estimation of the number of fire incidents linked to the installed PV capacity, and the fault tree analyses highlight where improvements are most critical. Based on the results of the analyses, two questions are suggested for implementation in the post-incident reports of the national fire and rescue services.
... That way, they can get repairs or replacements of defective solar panels. Additionally, there are faults such as hot-spot, double-ground, and arcs that have the potential of causing a fire hazard in solar panels [139][140][141]. Therefore, for such kinds of incidents as well, there is a requirement for an insurance cover. ...
Full-text available
India is a leader when it comes to agriculture. A significant part of the country's population depends on agriculture for livelihood. However, many of them face challenges due to using unreliable farming techniques. Sometimes the challenges increase to the extent that they commit suicide. Besides, India is highly populated, and its population is steadily increasing, requiring its government to grow its GDP and increase its energy supply proportionately. This paper reviews integrating solar farming with agriculture, known as Agrivoltaics, as a Climate-Smart Agriculture (CSA) option for Indian farmers. This study is further supported by the Strength, Weaknesses, Opportunities, and Threats (SWOT) analysis of agrivoltaics. Using the SWOT analysis, this article presents how agrivoltaics can make agriculture sustainable and reliable. This paper identifies rural electrification, water conservation, yield improvement, sustainable income generation, and reduction in the usage of pesticides as the strengths of agrivoltaics. Similarly, the paper presents weaknesses, opportunities, and threats to agrivoltaics in India. The research concludes with the findings that agrivoltaics have the potential of meeting multiple objectives such as meeting global commitments, offering employment, providing economic stability, increasing clean energy production capacity, conserving natural resources, and succeeding in several others. The paper also includes a discussion about the findings, suggestions, and implications of adopting agrivoltaics on a large scale in India.
Conference Paper
Numerous studies have been conducted on the biogas generation process using different feedstock as mono-substrates. However, the direct digestion of mono-substrates presents several disadvantages associated with substrate properties. To overcome the disadvantages, co-digestion of two or more substrates is recommended as an attractive alternative to enhance yields and economic feasibility of biogas plants. However, the problem of low biogas generation and process instability remains challenging in co-digestion systems due to the variability in co�substrate properties, the characterization of which requires advanced analytical approaches. The International Water Association’s Task Group developed the Anaerobic Digestion Model No. 1 (ADM1), which is considered as the most advanced and extensible model for simulating an anaerobic digestion process, as has been noted by several studies. In this study, ADM1 was used to simulate anaerobic co-digestion of agricultural bio-wastes under different operating conditions to identify a suitable co-digestion strategy for locally available agricultural substrates (smallholder farming system cases). Anaerobic co-digestion of agricultural bio-wastes including cabbage residues, potato peel, rice straw, and maize straw with manure was assessed via ADM1, which evaluates the effects of substrate combination on biogas production. The model results showed that the co-digestion of agricultural substrates with cow manure is favourable for biogas production with greater than 50% methane content generated. The highest methane content (59.59%) was obtained when 70% manure was co-digested with 20% cabbage residue and 10% straw. This study provides a basis for predictive modelling of co-digestions, which can be utilized as an initial screening test.
Full-text available
For a photovoltaic (PV) power generation system, the shading effect of PV panels caused by dust deposition is extremely unfavorable. The deposition of dust results in a severe reduction of power generation output, since the efficiency of PV panels is affected by the shading irradiance and blocking the cooling. In this study, a numerical simulation method is proposed to model the dust accumulation on PV panels to detect the effects on PV power generation caused by different wind directions and wind speeds. Due to the high accuracy of numerical simulation, and the short calculation cycle, the proposed method provides a certain prediction for the soiling management of PV panels in the wind-sand environment. Through simulations and experiments, the impacts of dust accumulation on the performance of PV panels with different wind directions are studied in detail with the wind speed changing from 4.43 m/s to 6.48 m/s and the dust particle size of 10 μm to 100 μm, which are based on the environment of Liverpool, England in a year. Besides, for PV arrays, the turbulences of the dust distribution around the PV panels are also analyzed. The data collected from experiments and simulations are used to verify the effectiveness of the proposed strategy.
Full-text available
Fault arc detection is an important technology to ensure the safe operation of electrical equipment and prevent electrical fires. The high-frequency noise of the arc current is one of the typical arc characteristics of almost all loads. In order to accurately detect arc faults in a low-voltage alternating-current (AC) system, a novel differential high-frequency current transformer (D-HFCT) sensor for collecting high-frequency arc currents was proposed. The sensitivity and frequency band of the designed sensor were verified to ensure that the acquisition requirements of the high-frequency current were satisfied. A series arc fault simulation experiment system was built, and resistive, inductive, and non-linear load and high-power shielding load experiments were carried out. Experiments showed that the sensor output signal was close to zero in the non-arc state, and the sensor output response was a high-frequency glitch in the arc state. The results were consistent for different loads, and the discrimination between normal and fault states was obvious, which proved that the sensor is suitable for series arc fault detection.
An experimental examination has been performed on the performance assessment of different shading shapes on photovoltaic panels by using energy-exergy analysis methods in this study. The problem is important due to decreasing of performance of PV panel with shading effect. A non-transparent material and powder were used for static shading while the time-varying shading effect of a chimney for dynamic shading was applied on to the system. Novelty of the work is to use diagonal shading on energy and exergy efficiency. Moreover, thermal camera technique was used to observe the effects of temperature distribution as a novel technique for PV panels. The photovoltaic panel was kept outside for one month in terms of dusty shading. For other static shading, artificial shadows of three different triangular shapes (40*40 cm for Case I, 36*60 cm for Case II and 60*36 cm for Case III) were created.
The output power produced by solar Photovoltaic (PV) array is reduced drastically by partial shading effect. Various array formation and reconfiguration techniques were introduced by many researchers to mitigate partial shading effects in PV array. This paper proposes new PV Array Topologies (PVATs) to improve the performance during partial shading conditions (PSCs). Totally, eight shading patterns are considered in the analysis for seven types of array configurations. Based on the existing array configurations, six novel PVATs are proposed to address the partial shading effect. A 4 × 4, 4 kW solar PV array which consists of sixteen panel of each 250 W rating is considered in this paper. The proposed PVATs are simulated in MATLAB/Simulink® to assess the performance. The results obtained from the simulation are compared with the conventional PVATs and suitable topologies which give best performance during various PSCs are identified. The result comparison shows that the modified total cross-tied (TCT) configuration performs well to extract more power in most of the PSCs. For the Short and Wide PSC, the proposed TCT improves the output power of PVAT by 105% compared to the existing TCT topology. The proposed method is also validated experimentally using 2×2 TCT PV array topology and the output waveforms are presented in this paper. This research would be helpful for the PV power plant installers to identify a suitable array configuration.
Currently, photovoltaics have been used on a large scale for commercial and civilian use. Aging short circuit, fire and other reasons will bring great security risks. In this paper, an experimental study of burning and toxic hazards was carried out on a widely used, flammable photovoltaic panel with a sample size of 180 mm*180 mm at atmospheric conditions. Combustion experiments were performed on the early stage fire characteristics bench of State Key Laboratory of Fire Science in China. Several important combustion parameters were investigated by oxygen consumption method under four representative external thermal (15 kW/m²,20 kW/m²,30 kW/m²,40 kW/m²), including Ignition Time, Heat Release Rate, Mass Loss Rate and Total Heat of Combustion. The results of experimental combustion heat were consistent with the thermodynamic calculation data of various organic matter in the sample. Several dangerous toxic gases have been detected, such as sulfur dioxide, hydrogen fluoride, hydrogen cyanide and a small amount of VOCs, of which the concentration of sulfur dioxide is relatively high. In the case of higher external radiant heat, there is a higher risk.
According to the distribution of reverse biased leakage current of micro-defective solar cells, the mathematic models of the hot spot thermal power of photovoltaic (PV) modules are established. The uniform and non-uniform distribution of thermal power in hot spot solar cell are theoretically calculated and experimentally analyzed in different shading conditions. Based on the model, the hot spot temperature distribution of different defective cells is simulated by ANSYS software. Then a novel experimental platform is set up to validate the simulation results in outdoor conditions. The mechanism of hot spot failure and the worst hot spot condition are analyzed. According to the results of experiment and simulation, the relative deviation between simulation and experimental data is less than 10% through the model proposed in this paper. Furthermore, in the normal modules, according to the result of simulation and experiment, the worst case of hot spots happens when the heat power is maximum and the shading ratio is about 10%-20%. In the defective modules, it happens when the heat power is not maximum and the shading ratio is about 50%-60%
In recent years there has been an increasing interest in Building-Integrated Photovoltaic (BIPV) and Building-Integrated Photovoltaic/Thermal (BIPVT) systems since they produce clean energy and replace conventional building envelope materials. By taking into account that storage is a key factor in the effective use of renewable energy, the present article is an overview about storage systems which are appropriate for BIPV and BIPVT applications. The literature review shows that there are multiple storage solutions, based on different kinds of materials (batteries, Phase Change Material (PCM) components, etc.). In terms of BIPV and BIPVT with batteries or PCMs or water tanks as storage systems, most of the installations are non-concentrating, façade- or roof-integrated, water- or air-based (in the case of BIPVT) and include silicon-based PV cells, lead-acid or lithium-ion batteries, paraffin- or salt-based PCMs. Regarding parameters that affect the environmental profile of storage systems, in the case of batteries critical factors such as material manufacturing, accidental release of electrolytes, inhalation toxicity, flammable elements, degradation and end-of-life management play a pivotal role. Regarding PCMs, there are some materials that are corrosive and present fire-safety issues as well as high toxicity in terms of human health and ecosystems. Concerning water storage tanks, based on certain studies about tanks with volumes of 300 L and 600 L, their impacts range from 5.9 to 11.7 GJprim and from 0.3 to 1.0 t CO2.eq. Finally, it should be noted that additional storage options such as Trombe walls, pebble beds and nanotechnologies are critically discussed. The contribution of the present article to the existing literature is associated with the fact that it presents a critical review about storage devices in the case of BIPV and BIPVT applications, by placing emphasis on the environmental profile of certain storage materials.
Fault diagnosis of photovoltaic array plays an important role in operation and maintenance of PV power plant. The nonlinear characteristics of photovoltaic array and the Maximum Power Point Tracking technology in the inverter prevent conventional protection devices to trip under certain faults which reduces the system’s efficiency and increases the risks of fire hazards. In order to better diagnose photovoltaic array faults under Maximum Power Point Tracking conditions, the sequential data of transient in time domain under faults are analyzed and then applied as the input fault features in this work. Firstly, the sequential current and voltage of the photovoltaic array are transformed into a 2-Dimension electrical time series graph to visually represent the characteristics of sequential data. Secondly, a Convolutional Neural Network structure comprising nine convolutional layers, nine max-pooling layers, and a fully connected layer is proposed for the photovoltaic array fault diagnosis. The proposed model for photovoltaic array fault diagnosis integrates two main parts, namely the feature extraction and the classification. Thirdly, this model automatically extracts suitable features representation from raw electrical time series graph, which eliminates the need of using artificially established features of data and then employs for photovoltaic fault diagnosis. Moreover, the proposed Convolutional Neural Network based photovoltaic array fault diagnosis method only takes the array of voltage and current of the photovoltaic array as the input features and the reference panels used for normalization. The proposed approach of photovoltaic array fault diagnosis achieved over 99% average accuracy when applied to the case studies. The comparisons of the experimental results demonstrate that the proposed method is both effective and reliable.