Analysis of Fire Risk Associated with Photovoltaic Power Generation System

Because of increasing energy consumption and severe air pollution in China, solar photovoltaic power generation plants are being deployed rapidly. Owing to various factors such as technology, construction, and imperfection of construction standards, solar photovoltaic systems have certain fire risks..is paper focuses on the fire risks of building-integrated solar photovoltaic buildings, as well as temperature and heat flow density near a photovoltaic system in a fire. Based on FDS simulation results, the influence of different building fires on photovoltaic systems is analysed. It is found that the influence of fire on photovoltaic systems installed on a building with a flat roof is stronger than that on a system installed on a building with a sloping roof; the influence of fire on a photovoltaic system installed on a building with external wall thermal insulation is stronger than that on a system installed on a building without such insulation; and in the presence of a skylight, a photovoltaic system installed on a building with a sloping roof is more dangerous.


Introduction
Given the difficulty in reconciling the contradiction between the limitations of conventional energy sources such as oil and coal and the expanding energy demand, all countries in the world have been promoting the development of photovoltaic systems.As a result, the photovoltaic industry has grown rapidly under the support of various national policies [1].Specifically, photovoltaic buildings as a typical form of photovoltaic systems have developed rapidly [2,3].However, the development of photovoltaic buildings creates huge pressures in terms of firefighting, especially at the present stage, where technology is in a relatively nascent stage and relevant specifications are lacking.erefore, photovoltaic buildings are potential fire hazards.First, photovoltaic power generation systems may undergo spontaneous combustion.Second, photovoltaic systems installed in buildings are threatened by building fires.Finally, because current flows through photovoltaic systems, a fire in such systems is difficult to extinguish.For the above reasons, PV systems are frequently susceptible to fire [4,5].In view of this, it is important to focus on the security of photovoltaic systems, especially distributed photovoltaic systems and photovoltaic buildings.
e composition of photovoltaic building systems is shown in Figure 1.Such systems generally consist of a photovoltaic array (which is composed of low iron tempered, EVA FILM, solar power chip, and TPT dorsal membrane), combiner boxes, inverter, and transformer.e risk of spontaneous combustion of photovoltaic systems is caused by the hot spot effect.
DC arc and other reasons have been greatly mitigated with the maturity of photovoltaic technology [6,7].Photovoltaic arrays of photovoltaic power generation systems are mainly installed on the roof of a building, which can be threatened by building fire.Because there is no specific fire prevention code for photovoltaic buildings, exiting photovoltaic buildings or those under construction are designed in accordance with the existing codes for re protection.Compared with traditional buildings, the re load of photovoltaic buildings is higher, and such buildings are more susceptible to re owing to the addition of power generation systems.Fires of photovoltaic power generation systems are electrical res, and it would be di cult to e ectively cuto the power supply.Moreover, water, which is the conventional re ghting medium, cannot be used to extinguish res in photovoltaic buildings [8][9][10].Furthermore, at present, the research on roof res in China is inadequate [11,12].us far, in China, there are no relevant speci cations for roo ng construction, re protection, and standards for ame-retardant performance of roof waterproo ng and thermal insulation materials.Roof waterproo ng and thermal insulation materials tend to be rigid combustible materials such as polyurethane foam [13], and they are hidden dangers in the event of a roof re.To sum up, photovoltaic buildings can catch re easily.
Until now, great importance has been attached to the power generation e ciency of the photovoltaic systems, but few researchers have focused on the safety aspects of these systems, especially photovoltaic buildings.Once a building re starts, photovoltaic power generation systems will be exposed to great danger; for this reason, in the present study, the authors apply FDS to simulate indoor res, building roof re, and other types of re scenarios and analyse the threats posed by di erent types of building res to solar photovoltaic power generation systems by detecting the temperature eld and heat ow density of certain areas.

Theoretical Model
According to Yang et al. [4], two factors pose dangers to photovoltaic systems, namely, temperature and heat ow density.To determine temperature and heat ow density, the rst step in calculation is to set up a control equation.e control equation is a mathematical description of the law of physical conservation.

Heat Flow Density.
For the vast majority of FDS applications, the "mixed is burnt" assumption is adequate to model the reaction system, and the mean chemical source term for fuel can be modelled using the eddy dissipation concept (EDC) put forth by Magnussen and Hjertager [14,15]: where Z F and Z ... A are the lumped mass fractions of fuel and air, respectively, and s is the mass stoichiometric coe cient of air.e quantity τ mix is the timescale for mixing, which must be modelled.e EDC model, therefore, states that the rate of fuel consumption is proportional to both the local limiting reactant concentration and the local rate of mixing.
Heat release per unit volume is calculated by summing the products of the mass production rates of various species with their respective heats of formation:

Advances in Civil Engineering
To maintain code stability, it is occasionally necessary to impose an upper bound on the local heat release rate per unit volume.A scaling analysis of pool fires by Orloff and de Ris [16] suggests that the spatial average of the heat release rate of a fire is approximately 1200 kW/m 2 .FDS imposes a less restrictive upper bound on the local heat release rate per unit volume: e value of 200 kW/m 2 is an empirically derived upper bound on the heat release rate per unit area of a flame sheet and δx is the characteristic cell size (m).

Simulation and Analysis
Fire simulation involves solving unknowns (density, velocity, temperature, and pressure) in equations based on reasonable simplification and approximation of the control equations.ere are considerable differences between the temperatures and heat flow densities of different photovoltaic buildings.To ensure the safety of photovoltaic systems, eight types of photovoltaic building scenarios are considered, and the temperatures and heat flow densities of eight fire scenarios are solved.

Model Building.
e flat-and sloping-roofed building models are shown in Figures 2 and 3, respectively.e geometric size of the flat-roofed building is 16,000 × 6600 × 3650 mm, thickness of its concrete walls is 250 mm, thickness of the polyurethane wall insulation layer is 50 mm, thickness of the concrete roof structure layer is 100 mm, thickness of polyurethane roof waterproofing and thermal insulation layers is 50 mm, and the building houses a few pieces of room furniture.e geometric size of the sloping-roofed building is 16,000 × 6600 × 3650 mm, thickness of its concrete walls is 250 mm, thickness of the polyurethane wall insulation layer is 50 mm, thickness of its PVC ceiling is 20 mm, thickness of the sloping concrete roof is 50 mm, thickness of its polyurethane waterproofing and thermal insulation layers is 10 mm, and an additional protective layer of ceramic tile is laid on top [17,18].

Scene Setting.
In this paper, eight fire scenarios, presented in Table 1, were used to simulate indoor or outdoor fire in a flat-roofed or sloping-roofed structure.

Data Detection Device Settings.
Because most photovoltaic arrays are installed on the roof of buildings and the threat of building fire to photovoltaic systems depends   mainly on the temperature eld of the roof, heat ow and other factors, the authors setup six thermocouples and heat ow metres in the simulation scenarios to monitor the temperature eld and heat ow on the building roof.e thermocouples and heat ow density metres on the roof of the building were distributed evenly, and their distance to the roof was 20 cm.

Simulation Result Analysis.
According to the experiments of Yang et al. [4], when the heat ow radiation intensity changes from 28 to 45 kW/m 2 , the ignition times of solar cell panels change from 913 to 83 s.When the temperature and heat ow density on the roof area are high, the photovoltaic systems can be ignited easily.e simulation results of the eight re scenarios are stated below.
e authors simulated eight re scenarios.By using the theoretical model described above, we can obtain the temperature eld near the roof.e heat ow density of a roof on re can be determined by solving (3)-( 5).Due to space constraints, it is di cult to present the analysis results of the eight types of re scenarios separately.erefore, case 1 is taken as an example to analyse the in uence of a re on a at roof on photovoltaic systems, and case 4 is taken as an example to analyse the in uence of a re on a sloped roof on photovoltaic systems.Finally, the in uences of eight types of re scenarios on photovoltaic systems are analysed by comparing the highest roof temperatures and heat ows in each re scenario.
e monitoring data of case 1 are shown in Figures 4 and 5: As shown in Figure 4, the changes in roof temperature in case 1 can be divided into three stages: the rst stage is the phase from the start of the re to 270 s, where the roof temperature rises gradually to about 100 °C; the second stage is the phase from 270 s to 750 s, during which time the roof temperature is maintained at approximately 100 °C; and the third stage is the phase from 750 s to the end of the simulation, during which time the roof temperature rst rises sharply and then falls sharply.
e roof temperatures at di erent locations within the roof are di erent, and the highest roof temperature is about 700 °C.
From Figure 5, we know that the roo ng heat ows at di erent locations in case 1 have almost the same change tendency, that is, between 0 s and 750 s; the heat ows increase very slowly; and then, they peak with a dramatic rise, followed by a sharp drop to zero.ere are some di erences among the roo ng heat ows at di erent locations.e peak heat ow density is nearly 250 kW/m 2 .
When the simulation time is 750 s, the indoor re spreads out of the room through the doors, windows, skylights, and other openings and ignites the roof waterproo ng and insulation layers, and a naked light appears on the roof, which leads to a sharp increase in the temperature of heat ow on the roof.e temperatures of most parts of the roof were higher than 500 °C, and the heat ow densities were higher than 150 kW/m 2 .As a consequence, we can see the trends of changes in roof temperatures and heat ows described above.
In case 2, the roof waterproo ng and insulation layers burned out within 150 s, and roof temperature in the former 150 s rose rapidly, with the roof top temperature reaching 1000 °C; it decreased gradually thereafter.Heat ow density increased sharply at 150 s, and the maximum heat ow density on the roof was close to 300 kW/m 2 , which decreased to 0.
e monitoring data of case 3 are shown in Figures 6 and 7.As can be seen in Figure 6, the trend of change in roof temperature is the same.From beginning to the simulation time of 200 s, the roof temperature rises gradually and reaches its peak at 200 s, after which it remains stable.e highest temperatures of most areas of the roof are less than 100 °C.We can see from Figure 7 that the heat ow densities on the roof in case 3 are close to 0. e monitoring data of case 4 are shown in Figures 8 and 9. Figure 8 shows the trend of change in roo ng temperature in case 4. From 0 s to 750 s, the roof temperature rises; from 300 s to 420 s, the roof temperature rises rapidly; and at 750 s, the roof temperature peaks at 350 °C.
Figure 9 shows that the trend of heat ow density throughout the area around the roof is consistent.e change in heat ow density tends to zero.Speci cally, the heat ows of HRR1 and HRR4 start increasing at 420 s, peak at 720 s, and abruptly drop to 0.
At the simulation time of 320 s, the indoor re spreads out through windows and skylights, which led to an increase in the heat ow densities of the roof areas near the window (regions near HRR1 and HRR4) until 720 s, at which point the indoor re reaches its peak.However, the roof waterproo ng and insulation layers were not ignited yet, and there was no naked light on the building roof.erefore, in other areas of the roof, the heat ow densities are close to zero, and the temperatures are low.
e heat ow in case 5 has the same change trend as in case 4. e temperature in case 5 was higher than that in case 4, and the highest temperature was above 500 °C.ere was no naked light on the building roof.
In cases 6 and 7, there was no naked light on the roof, and the heat ow densities on the roof were close to 0.

Advances in Civil Engineering
Accordingly, the roof temperatures were lower than 50 °C in case 6, and the highest temperatures of most areas of the roof were lower than 350 °C in case 7.
In case 8, the heat ow densities of most areas of the roof were close to 0, and the waterproo ng and insulation layers were not ignited, as shown in Figure 10; the roof temperature was less than 140 °C.
Comparisons of the roof temperatures and heat ow densities of these cases are shown in Figures 11 and 12.
As shown in Figure 11, the roof temperatures in cases 1, 2, 4, and 5 were higher, especially, the roof temperature in case 2, which peaked at nearly 1000 °C.According to Figure 12, in case 2, the heat ow density on the roof increased rapidly and peaked at close to 300 kW/m 2 when the roof temperature reached its maximum value.is means that the roof was ignited in case 2. In addition, the phase from the beginning to the time when the temperature and heat ow reached their maximum values was the shortest, which means that the roof ignited quickly in case 2. From Figures 11 and 12, it is clear that when the roof was ignited in case 1, the roof temperature and heat ow density increased rapidly and peaked.As shown in the gures, although the roof temperatures are high in cases 4 and 5, the roof heat ow densities are low because the roof is not ignited, as stated in the analysis in the preceding part of the text.
e temperature near the window and skylight is high because the heat of the indoor re spread to the roof through the opening.In cases 3, 6, 7, and 8, the roof temperatures are lower than 100 °C, heat ow densities are 0, and the e ect of the re on the photovoltaic systems is diminished.
Compared to the roof temperatures and heat ow densities in cases 2 and 4, those in cases 1, 5, 6, and 7 are higher, and their in uences on the photovoltaic systems are stronger in the case of a re in a at-roofed building than that in the case of a sloped-roofed building.Based on a comparison of cases 1, 2, and 3 and a comparison of cases 4, 5, 6, 7, and 8, indoor re in a building ignites the roof more easily, and the possibility of the re engul ng photovoltaic systems is greater.A comparison of cases 4, 5, 6, and 7 shows that, in a sloping-roofed building with a skylight, an indoor re spreads more easily to the roof and ignites photovoltaic systems on the roof compared to that in a sloping-roofed building with no skylight.A comparison of cases 1 and 2 shows that a re of the external wall thermal insulation ignites the roof more easily.A comparison of cases 4 and 5 shows that when the external wall thermal insulation of building is on re, the roof temperature is higher, and the in uence of the external wall thermal insulation building re on photovoltaic systems is stronger.

Conclusions.
e values of roofing heat flow density and temperature depend on whether the roof is ignited and a skylight is present, magnitude of heat flow density, and magnitude of roof temperature.If roof temperature and heat flow density are high, the photovoltaic systems on the roof are at great risk.
(1) Compared with fires in sloping-roofed buildings, fires in flat-roofed buildings have stronger adverse effects on photovoltaic systems.When the exposed waterproofing and thermal insulation material in a flatroofed building fire is ignited, the roofing temperature increases more than 750 °C, roof heat flow density more than 260 kW/m 2 , and any photovoltaic systems on the roof are more likely to ignite.(2) In case of a fire on a sloping-roofed building, because there are protective layers such as ceramic tiles and there is no naked light on the roof, heat flow density is smaller than 1 kW/m 2 , and the possibility of photovoltaic systems igniting is small.However, the roof temperature is higher than 450 °C when there is a skylight, and the photovoltaic systems on a slopingroofed building with a skylight are more susceptible to ignition.(3) Relative to fires of the internal wall thermal insulation of buildings, fires of the external wall thermal insulation of buildings have a greater influence on photovoltaic systems.

Suggestions.
Photovoltaic buildings are severely impacted by building fires.With the vigorous development of the photovoltaic industry, along with the great importance accorded to photovoltaic power generation efficiency, we should focus on improving the safety of photovoltaic systems.First, we should take roof fires seriously, use noncombustible or flame-resistant materials for waterproofing and insulation materials to the extent possible, and issue related standards in a timely manner, such as fire safety rules specific to photovoltaic buildings.Moreover, capabilities for preventing and controlling fires of photovoltaic systems should be strengthened by providing professional training to the firefighting staff.

Figure 12 :Figure 10 :Figure 11 :
Figure 12: Comparison of maximum heat ow in eight cases.