This website uses cookies to store information on your computer. Some of these cookies are used for visitor analysis, others are essential to making our site function properly and improve the user experience. By using this site, you consent to the placement of these cookies. Click Accept to consent and dismiss this message or Deny to leave this website. Read our Privacy Statement for more.
Fire Safety of PV Modules on Roofs: Critical Gap for Hazardous Flame Spread

By Farah Binte Mohd Faudzi, International Masters in Fire Safety Engineering (IMFSE); Jens Steemann Kristensen, School of Engineering, BRE Centre for Fire Safety Engineering, The University of Edinburgh, Guest researcher at: Department of Civil Engineering, The Technical University of Denmark; Grunde Jomaas, School of Engineering, BRE Centre for Fire Safety Engineering, The University of Edinburgh

Solar photovoltaic (PV) power generation has increased exponentially over the last decade, from 13 GW in 2008 to 505 GW in 2018.1,2 As part of this trend, rooftop PV power systems have become commonplace as corporations seek to reduce energy expenses and meet energy efficiency targets by mounting PV installations on the roofs of their warehouses.3 Consequently, there has been a corresponding global increase of fire incidents involving rooftop PV systems, which raises concerns regarding the fire risk associated with them.4,5

The risks related to PV systems are caused by two compounded changes to the roof construction: A) an increased probability of ignition, and B) changed fire dynamics in the gap between the roof and the PV modules, which often lead to more severe consequences upon ignition events.

PV installations introduces a substantial power generating direct current (DC) system that increases the probability of ignition of fire due to a myriad of possible malfunctions throughout its lifetime; from installation errors to wear and tear as components exceed their stipulated lifetime.5 Furthermore, these installations are a hindrance to fire-fighting operations as the system may remain live even during a fire, and fire-fighting personnel is at risk to electric shocks. Also, the physical presence of the PV installation facilitates fire spread by introducing a significant increase in heat flux directed back to the roof due to re-radiation from the deflected flame on the underside of the PV module.6,7 Regardless of the source of ignition, the PV modules can facilitate the spread of fire, which is one of the central themes of current research within the field.

Past research focused on testing specific combinations of PV modules and roofing solutions, based on different classification standards such as UL 1703 and UL 790.8–11 These tests, which are based on a pass/fail criterion, give a reasonable prediction of the fire behaviour for a given system that has been ignited by a gas burner. However, they disregard the importance of understanding the complexity of the system behaviour associated with PV modules and the substrate they are placed on. Multiple aspects that are not covered by the specified test may render the test results irrelevant in a real fire scenario.

The fire dynamics associated with a system consisting of PV modules, roof construction and mounting system requires an understanding of the fundamental principles related to the configuration. With a variety of materials used for both the PV modules and roof construction, decoupling of the material properties and its role in fire spread would enable researches to reveal the fundamental processes that drive fire spread along the gap between a PV module and the roof. Therefore, a fundamental analysis of the influence of the gap between the incident non-combustible surface (i.e. PV module) and the combustible surface (i.e. roof surface) is of particular interest as it would allow for characterisation of the fire spread within the gap.


A series of experiments were undertaken to address the knowledge gap outlined in the introduction. In these, PMMA was used as a surrogate for the roofing material as its properties are well known and it is a fuel whose burning behaviour is well studied and reproducible.12,13 As combustible components represent less than 10 % of PV modules by mass, it was assumed to be almost non-combustible 14 and a stainless steel plate painted with a high heat resistant black paint was used as a surrogate. The use of a non-combustible surrogate material also ensured that there was no undefined, additional fuel load in the experiments. Both surrogate materials were assumed to be black bodies, as the stainless steel plate was painted black and the black PMMA had a transmission of 0. Depending on their geographical location, PV arrays are installed with various inclinations to optimise the energy generation. As previous research revealed that there is a correlation between a reduced gap distance and an increased heat flux 6, the smallest gap distance is expected to be most critical from a fire safety point of view. Therefore, a horizontal panel was considered in the experimental set-up. The gap distance (H) between the surfaces was varied between 10 and 25 cm to understand its influence on fire spread within the gap.

Ignition was achieved by igniting a 20 mm wide strip of methanol-wetted ceramic paper that was evenly laced along the width at the one end of the PMMA sample. Compared to ignition by the gas burner method used in test standards, the method adopted for these experiments was able to produce a small fire that is more representative of initial fires below a PV module. In addition, it ensured that the growth phase, if any, of the fire could be observed.

Figure 1 shows the set-up that was used to study the flame spread along a 2 mm thick horizontal PMMA (300 mm wide and 700 mm long) with a horizontal parallel stainless-steel panel above for varying gap distance. These experiments were then compared to a free-burning scenario without the presence of the surrogate panel.

Figure 1 Experimental set-up utilising representative surrogate materials and based on a representative orientation for characterisation of flame spread within the gap


The location of the flame front for the various panel height configurations are plotted in Figure 2. The resultant flame spread rate can be derived from the gradient of the plots for the tests of various panel heights.

Figure 2 Location of the flame front as a function of time for various panel heights

In the free burning scenario, i.e. without a panel, (Figure 3a), a constant flame spread rate was observed for a small flame moving along the length of the PMMA plate. For all gap distances of 20 cm or more (Figure 3b), a constant, though slightly higher, flame spread rate was recorded, and the flame was somewhat larger. In these experiments, the presence of the panel caused a hot gas layer to form below the panel, and this also heated the panel. The combined effect of the hot gas layer and the hot panel results in pre-heating of the sample ahead of the flame front, thus increasing the width of the pyrolysis zone. There was a corresponding increase in flame height though it did not exceed the gap distance and no flame impingement on the panel was observed. Overall, these results imply that the effect of the panel on flame spread rates is limited at such gap distances.

However, when the gap distance was reduced to 17 cm or less, there was a significant change in the flame spread behaviour. Although the flame spread rate was initially similar to that of the free burning scenario, the fire continued to grow until the flame impinged the panel, causing the flame to deflect on the underside of the panel (Figure 3c, Figure 3d). Upon flame deflection, the flame spread rate increased rapidly to a resultant flame spread rate that was more than 10 times larger than that observed in the free burning scenario situation. The deflected flame imposed a higher heat flux on the sample and pre-heated the sample ahead of the flame front not only with greater intensity but also across an extended distance away from the flame front corresponding the length of the deflected flame. Thus, the effect of the panel on flame spread rates was evident as the flame spread characteristics transited from a slow-growing stage to a rapid flame spread rate upon flame impingement on the panel. These observations indicate that there is a critical gap distance below which fires appear to be mild at the start of ignition but with sufficient time to grow, a rapid transition to a significant flame spread rate could occur, resulting in an uncontrolled fire that could incur extensive damages.

(A) No Panel (Time: 3814 s)

(B) Gap distance: 20 cm (Time: 2383 s)

(C) Gap distance: 17 cm (Time: 1051 s

(D) Gap distance: 10 cm (Time: 633 s)

Figure 3 Photographs from the instant the flame front reached of 50 cm. The white arrow indicates the direction of flame propagation. Note the increase in intensity as the gap was reduced.


These preliminary experiments provide an enhanced understanding of fire growth and flame spread in horizontal gaps and its relationship to the gap distances. The results highlight the limitation on current test procedures for PV modules on roofs, which are primarily concerned with material combustibility, and thus disregards the characteristics of the gap between the two surfaces and its implications related to fire dynamics.

Consequently, the findings also indicate that the flame spread behaviour within the gap is strongly affected by minute difference in the gap distance. In the current experiments, a critical gap height was found, above which the risk of extended damage is low. This suggests that it could be used as a mitigation measure and factored in PV installation design in conjunction with other drivers such as aesthetics and wind load that seeks to limit the gap distance. By optimising the gap distance, an economical design solution could be achieved without compromising fire safety and without adding other, potentially expensive, fire safety measures.

The current preliminary experiments suggest that the significance of PV installations on fire safety is reliant on installation parameters such as the gap distance. Although the current set-up has established a critical gap distance between 17 and 20 cm, further studies on its applicability to actual installations should be carried out with the experimental matrix extended to include the influence of the material properties related to the PV modules and the roof construction. Further studies are also needed to examine the influence of the PV module, where the combustible backside membrane might have an additional influence on the flame spread rate. Similar experimental tests are required for the various roofing membranes, which, contrary to the PMMA used in the experiments presented herein, almost invariably contains flame retardants. Research on how other variables such as the initial fire size and wind effects could affect the flame spread behaviour and the resultant critical gap distance would be useful in developing a comprehensive study. With more refinements and variability in tests, the current proposed set-up has the potential in providing a close indicator of flame spread behaviour for a realistic scenario with PV modules especially if modelling could be used to predict the results for various different scenarios. As a result, generalised scenarios for a variety of PV installations configurations could be developed in order to determine critical gap distances for use in formulating guidelines for the design for PV installations on roofs.

Given the complexity of the fire dynamics problem related to PV modules on roofs, partially due to variability in how they are installed, the full range of possible scenarios would have to be addressed in order to mitigate the hazard imposed by the PV installations. However, through a fundamental approach of understanding fire propagation within horizontal gaps and the factors that influence it, the consequences of a potential fire could be better managed. This would not only reduce the damage caused by such fire but also safeguard the occupants and the fire fighters on scene.


The authors appreciate the support by the sponsors of the project, namely IKEA, Kingspan and Rockwool. In addition, the authors thank Department of Civil Engineering at the Technical University of Denmark (DTU) for the hospitality and the use of their lab for the experiments and use of facilities for these experiments.


  1. International Energy Agency (IEA). Trends in Photovoltaic Applications: Survey Report of Selected IEA Countries between 1992 and 2008. Accessed June 12, 2019. Published 2009.
  2. International Energy Agency (IEA). 2019 Snapshot of Global PV Markets - Task 1: Strategic PV Analysis and Outreach. Published 2019. Accessed June 12, 2019.
  3. Solar Energy Industries Association (SEIA). Solar Means Business: Tracking Corporate Solar Adoption in the U.S. Published 2018. Accessed February 20, 2019.
  4. Fiorentini L, Marmo L, Puccia V, Danzi E. Fires in Photovoltaic Systems: Lessons Learned from Fire Investigations in Italy. SFPE Fire Protection Engineering Magazine. 2015.
  5. Laukamp H, Bopp G, Grab R, et al. PV Fire Hazard - Analysis and Assessment of Fire Incidents. In: European Photovoltaic Solar Energy Conference. ; 2013.
  6. Kristensen JS, Merci B, Jomaas G. Fire‐induced Reradiation Underneath Photovoltaic Arrays on Flat Roofs. Fire and Materials. 2018;42(3):316-323.
  7. Kristensen JS, Jomaas G. Experimental Study of the Fire Behaviour on Flat Roof Constructions with Multiple Photovoltaic (PV) Panels. Fire Technology. 2018;54(6):1807-1828. doi:10.1007/s10694-018-0772-5
  8. Sipe J. Development of Fire Mitigation Solutions for Photovoltaic ( PV ) Systems Installed on Building Roofs – Ph . 1. Published 2016. Accessed June 10, 2019.
  9. Sherwood L, Backstorm B, Sloan D, Flueckiger C, Brooks B, Rosenthal A. Fire Classification Rating Testing of Stand-Off Mounted Photovoltaic Modules and Systems.; 2013.
  10. Backstrom B, Sloan D. Effect of Rack Mounted Photovoltaic Modules on the Fire Classification Rating of Roofing Assemblies Phase 2. Underwriters Laboratories Inc. (UL). Published 2012. Accessed February 20, 2019.
  11. Backstrom B, Tabaddor M. Effect of Rack Mounted Photovoltaic Modules on the Fire Classification Rating of Roofing Assemblies. Underwriters Laboratories Inc. (UL). Published 2010. Accessed February 20, 2019.
  12. Fernandez-Pello A, Williams FA. Laminar Flame Spread Over PMMA Surfaces. Symposium (International) on Combustion. 1975;15(1):217-231.
  13. Ray SR, Fernandez-Pello AC, Glassman I. A Study of the Heat Transfer Mechanisms in Horizontal Flame Propagation. Journal of Heat Transfer. 1980;102(2):357-363.
  14. Sander K, Schilling S, Wambach K, et al. Studie Zur Entwicklung Eines Rücknahme‐ Und Verwertungssystem Für Photovoltaische Produkte (in German). Ökopol – Institute for Environmental Strategies. Accessed June 12, 2019. Published 2007.

Corporate 100 Visionaries

About Us

© 2019 SFPE | All Rights Reserved
Privacy Policy