As electric vehicles (EVs) have become increasingly widespread, the danger of fires stemming from the transportation of EVs on passenger roll-on/roll-off (PRO) ships has grown notably more significant. The article takes a 15000GT passenger ship as the research object, and uses FDS numerical simulation software to establish a fire simulation model for the transportation of electric vehicles on the passenger ship, and realizes the simulation and simulation analysis of the thermal runaway of electric vehicles. Subsequently, the YOLOv5 model was integrated with the attention mechanism to develop the YOLOv5 – ECA model. This newly – developed model is intended for assessing the quantity of targets at the fire scene of an electric vehicle transported on a passenger – roller ship. The results show that the PHRR of EVs in a fire can reach up to 29.75 MW in 600 s, which is only 0.84% different from the actual calculation results. The YOLOv5-ECA model achieves a mAP of 88.42% for EV target number detection, and the overall model volume is lower and the detection speed is faster. Therefore, relying on the effective combination of fire dynamics simulation and deep learning technology, the analysis of thermal runaway spread of electric vehicles carried by passenger ships can be realized, and the possible economic losses of electric vehicles under fire scenarios can be more accurately grasped to provide guidance for good fire prevention measures.
