建筑火灾分布式反演与预测方法研究  

Research on decentralized inversion and prediction methods of building fire

作　　者：寇璐瑶 张辉[2] 王欣芝 马一飞 KOU Luyao;ZHANG Hui;WANG Xinzhi;MA Yifei(Emergency Management Training Center,Party School of the Central Committee of C.P.C(National Academy of Governance),Beijing 100091,China;Institute of Public Safety Research,Tsinghua University,Beijing 100084,China;School of Computer Engineering and Science,Shanghai University,Shanghai 200444,China;Fire and Rescue Department of Shanxi Province,Taiyuan Shanxi 030001,China)

机构地区：[1]中共中央党校(国家行政学院)应急管理培训中心,北京100091 [2]清华大学公共安全研究院,北京100084 [3]上海大学计算机学院,上海200444 [4]山西省消防救援总队,山西太原030001

出　　处：《中国安全生产科学技术》2024年第5期5-12,共8页Journal of Safety Science and Technology

基　　金：中央党校(国家行政学院)校(院)级重大理论和现实问题研究专项“关于中国式现代化进程中的风险挑战及其应对研究”(2023ZDZX059)。

摘　　要：为提高建筑火灾感知与预测的效率和准确性,采用边缘计算这一分布式计算模式,增强建筑终端的智能分析能力,提出边缘分布式的火源参数反演与态势预测方法。研究结果表明:对动态火灾场景,分布式反演与预测方法均能以10~(-1) s量级输出较准确结果,并对全部34组常见误报场景实现判别,同时分布式模式提高反演与预测方法对不同建筑结构的可扩展性。研究结果可为建筑火灾应急救援与人员疏散提供决策支持。To improve the efficiency and accuracy of building fire detection and prediction,the decentralized mode of edge computing was used to enhance the intelligent analysis capability of building terminals,and the edge decentralized fire inversion and situation prediction method were proposed.The results show that for the dynamic fire scenarios,both the decentralized inversion and prediction methods can produce relatively accurate results at the order of 10-1 s and achieve the discrimination for all 34 groups of common false alarm scenarios.Meanwhile,the decentralized mode improves the scalability of inversion and prediction methods for different building structures.The research results can provide decision support for the emergency rescue and personnel evacuation of building fires.

关 键 词：建筑火灾 参数反演 态势预测 模拟仿真 

分 类 号：X932[环境科学与工程—安全科学]