Fire risk associated with photovoltaic installations on flat roof constructions: experimental analysis of fire spread in semi-enclosures
Kristensen, Jens Steemann
An increase of photovoltaic (PV) installations facilitated by a significant cost reduction give rise to the utilization of commercial flat roof constructions, as unexploited surfaces elevated above ground level are ideal for building applied PV (BAPV) systems. However, an increased number of fires related to the PV systems have been reported; thus, it is essential to understand the fire-related risk. The thesis aims to provide an improved understanding of the fire-related risk associated with BAPV systems on commercial flat roof constructions. In addition to the main focus on the fire dynamics of the system consisting of the PV module, the roof construction, and the initial fire between the two surfaces, supplementary studies are conducted to understand the frequency of PV-related fires. As such, the objectives of the thesis are i) to quantify the frequency and origin of PV-related fires. ii) To design and build a novel experimental set-up to enable measurement of relevant parameters, to iii) quantify the consequences of the semi-enclosure's geometry and facilitate analysis with a basis in fundamental flame spread theory. iv) To detach the geometry from the material parameters in the system of fire dynamics to enable, v) the selection of relevant parameters facilitating flame spread. Based on the outcome of those objectives, the final objective is vi) to discuss, suggest and examine possible mitigation solutions. A fault tree analysis of all available data on PV-related fires estimated an annual frequency of 28.9 fires per gigawatt. However, in 49% of the analysed fires, the initial source of ignition was unknown or not related to the PV system. As such, it was concluded that the risk associated with BAPV systems was not only related to the inherent likelihood of the direct current PV system acting as an ignition source. The modification of the roof construction also changed the fire dynamic scenario and thus the consequences of an ignition. The existence of a critical gap height was defined for one-dimensional flame spread on thermally thin polymethyl methacrylate (PMMA) in a semi-enclosure below a horizontal non-combustible stainless-steel board. If the gap height was above the critical gap height, a constant flame spread rate (FSR) occurred after ignition. In contrast, an enhancing heat feedback loop caused an increase of the FSR, which accelerated rapidly when the flame deflected below the horizontal barrier if the gap height was below the critical gap height. The relationship between the location of the flame front, the temperature of the PMMA, and heat flux towards the fuel ahead of the flame front corresponded well with the fundamental flame spread theories by F.A. Williams and J. Quintiere. The importance of the critical gap height was evident when the PMMA was replaced with a PVC-based roofing membrane compliant with EN 13501-5 BROOF(t4), where it defined the difference between flame spread and no flame spread outside the domain of the ignition source. If the flame front detached from the ignition source, it propagated below all of the PV module, which corresponded with results from published large scale experiments and genuine PV-related fires. To quantify the severity of fires in a semi-enclosure, experiments were conducted with half a square meter of roofing membrane below a PV module, which caused a peak heat release rate of 90 kW and a maximum heat flux of more than 20 kW/m2 at a distance of 15 cm from the flame front. Similar to previous large-scale experiments, the fire did not propagate outside the semi-enclosure. Tree feasible mitigation solutions were discussed: i) increase of gap height, ii) increase of distance between PV arrays, and iii) protection against downward flame spread. For experimental examination of the final case, the use of both 60 mm polyisocyanurate (PIR) insulation or 50 mm mineral wool insulation provided sufficient protection of a subjacent layer of flame retarded expanded polystyrene (EPS), as the EPS was not ignited, although sections melted in the majority of the experiments. Altogether the thesis provides a new perspective on the fire-related risk associated with the novel introduction of the PV technology into the built environment. It provides an increased understanding of flame spread in semi-enclosures and evaluates potential mitigation solutions. Besides the direct outcome of the thesis, it will form a basis for further exploration of the topic.