| Abstract |
Nuclear power plants are high-risk sectors with stringent safety requirements. A fire incident could lead to critical equipment failure, compromise the safety systems of the plant, and increase the risk of nuclear leakage, thereby posing a threat to the safety of surrounding smart cities. Fires in nuclear power plants are low-probability events, which has resulted in a lack of training datasets, a major obstacle in the development of artificial neural network prediction models in traditional research. This paper addresses the fire safety issues in nuclear reactor plants by integrating numerical simulation with artificial intelligence. Using the fire dynamics software FDS, we conducted numerical simulations of cable fires in nuclear reactor plants, covering various fire scenarios and generating a large amount of three-dimensional spatial fire field data. This effectively trains neural network prediction models. Based on this, a fire detection algorithm using a Bi-LSTM was designed, achieving early fire identification and precise localization, thereby providing robust support for the safety management of smart cities. Additionally, an intelligent fire suppression control system was developed, enabling rational configuration and efficient utilization of firefighting equipment, further enhancing the ability of smart cities to respond to emergencies. By integrating the safety management of nuclear power plants with the overall safety framework of smart cities, we can not only ensure the safe operation of nuclear power plants but also effectively protect the lives and property of urban residents, promoting the sustainable development of smart cities. © 2025 The Authors. |
