Eksperimentell studie og Computational Fluid Dynamics (CFD) simulering av hvordan utenpåmonterte fotovoltaiske installasjoner påvirker branndynamikken på en realistisk norsk 30° skrå takkonstruksjon

Abstract

In combination with increased population growth, and that the global community has had a global overconsumption of the Earth's resources over a long period, humanity must act differently to reverse the negative trend when it comes to global warming. As part of a new pattern of action, photovoltaic (PV) installations, which are expected to increase by 13% each year from 2020 to 2030 [1], are one of several important measures when it comes to supplying the increased population with renewable energy. In this context, there is relatively little, but increasing research in the field of PV installations and fire safety [2]. In collaboration with RISE Fire Research [3], this thesis has conducted 6 medium and 2 full-scale fire experiments on a 30° sloped roof construction with externally mounted PV modules (BAPV). The roof construction was carried out in accordance with the classification BRoof (t2) [4], which is a pre-accepted solution in the Norwegian Building codes (VTEK17) [5]. The roof was constructed from the bottom up; OSB, chipboard, and a bitumen-based roofing membrane. With a distance of 12cm, BAPVs made of stainless steel were mounted parallel to the roof surface. The experiments were conducted outdoors at the RISE Fire Research facility in Trondheim in the autumn of 2021. The project was funded by the Norwegian Directorate for Civil Protection (DSB) and Norwegian Building Authority (DiBK). The motivation behind the medium-scale experiments was to research the size of a standardized initial fire, which could mirror a realistic brand fire, and which entailed fire development and propagation on the roof surface. The medium-scale experiments were carried out with different sizes of initial fire with both the presence and absence of a PV module. The results from the experiments showed an increase in damage extent and heat transfer inward in the roof construction when using an BAPV compared to experiments without a PV module, especially when UL B-wood crib [6] was used as the initial fire. For this reason, UL B-crib was also used as the initial fire in both full-scale experiments. Both full-scale experiments gave approximate results with fire propagation in the cavity to the BAPVs, all the way up and past the ridge. Both temperature measurements and damage extent showed that the fire had the greatest propagation in the middle of the roof construction, and with a tendency to propagate towards the right side of the construction. Most likely this had something to do with wind direction. Temperature measurements under the chipboard were relatively low, an indication that there is no immediate danger of fire spread inward in the roof construction. To recreate the experiments described above, Fire Dynamics Simulator (FDS) was used to simulate a specific small-scale experiment (test setup T4) previously conducted separately by RISE Fire Research [3]. The roof construction in the small-scale experiment had a similar construction and PV module as for the medium and full-scale experiments, but the distance between the PV module and the roof surface was 6 cm, the initial fire was an EN-wood crib [4], and a fan supplied constant 2m/s wind at the roof eaves. It required less computing power to simulate a complete test setup in this scale compared to the medium and full-scale experiments. Lack of data on the material properties of the roofing membrane and the effect of the initial fire necessitated experiments in the lab at Western Norway University of Applied Sciences (HVL), where, among other things, cone calorimeter and thermogravimetric analysis were used. The results of the FDS simulation failed in a satisfactory manner to recreate data from the small-scale experiment. Further work with the FDS model, especially with the pyrolysis model of the roofing membrane, is necessary to approach the results in the RISE Fire Research experiment.