主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2024, Vol. 43 ›› Issue (6): 820-825.

• • 上一篇    下一篇

基于多特征参数的GA-WOA-BP火灾概率预测模型研究

刘全义1,2, 吴孟洋1,2, 艾洪舟1,2, 朱培2,3   

  1. （1. 中国民用航空飞行学院 民航安全与工程学院，四川 广汉 618307；2. 中国民用航空飞行学院 民机火灾科学与安全工程四川省重点实验室，四川 广汉 618307；3. 南京航空航天大学 民航学院，江苏 南京 210000）
  • 出版日期:2024-06-15 发布日期:2024-06-15
  • 作者简介:刘全义（1987- ），男，河南郸城人，中国民用航空飞行学院民航安全与工程学院教授，硕士生导师，博士，主要从事飞机火灾智能监测预警，新一代机载灭火技术等方面的研究，四川省德阳市广汉市南昌路46号，618307。
  • 基金资助:
    国家自然科学基金（U2033206）；民机火灾科学与安全工程四川省重点实验室项目（MZ2022JB01，MZ2022KF08）；航空科学基金（ASFC-20200046117001）；四川省院校合作项目（2024YFHZ0027）；民航应急科学与技术重点实验室项目（NJ2022022，NJ2023025）

Research on GA-WOA-BP fire probability prediction model based on multi-feature parameters

Liu Quanyi1,2,Wu Mengyang1,2,Ai Hongzhou1,2, Zhu Pei2,3   

  1. (1. College of Civil Aviation Safety Engineering, Civil Aviation Flight University of China, Sichuan Guanghan 618307, China;2. Civil Aircraft Fire Science and Safety Engineering Key Laboratory of Sichuan, Civil Aviation Flight University of China, Sichuan Guanghan 618307, China;3. College of Civil Aviation, Nanjing University of Aeronautics and Astronautics, Jiangsu Nanjing 210000, China)
  • Online:2024-06-15 Published:2024-06-15

摘要： 为进一步提升火灾概率预测的准确率，针对BP神经网络在拟合过程中探测精度低、泛化能力差的问题，提出一种基于多特征参数的GA-WOA-BP火灾概率预测模型。首先通过试验采集了榉木、棉绳阴燃、明燃时的火灾特征参量，计算后得到了相应的火灾类型发生概率；其次通过遗传算法优化BP神经网络的隐藏层结构，鲸鱼优化算法优化BP神经网络的初始权重，构建了GA-WOA-BP模型，提高融合算法的拟合能力。最后，以多特征火灾参数作为模型输入，以不同类型火灾发生概率作为输出完成火灾概率的预测。结果表明，相比单纯BP神经网络，基于多特征参数的GA-WOA-BP火灾概率预测模型具有更好的预测性能，其评价指标RMSE、MAE、R2分别为0.020 22、0.014 33和0.992 31，能为火灾概率预测提供数据参考。

关键词: 多特征参数, 鲸鱼优化算法, 遗传算法, 火灾概率预测, BP神经网络

Abstract: In order to further improve the accuracy of fire probability prediction, a GA-WOA-BP fire probability prediction model based on multi-feature parameters is proposed to solve the problems of low detection accuracy and poor generalization ability in the fitting process of BP neural network. Firstly, the fire characteristic parameters of beech and cotton rope smoldering and open burning were collected by experiment, and the corresponding probability of fire type was obtained after calculation. Secondly, by using genetic algorithm to optimize the hidden layer structure of BP neural network and whale optimization algorithm to optimize the initial weight of BP neural network, GA-WOA-BP model was constructed to improve the fitting ability of fusion algorithm. Finally, multi-characteristic fire parameters are used as model input and different types of fire occurrence probabilities are used as output to predict the fire probability. The results show that compared with the simple BP neural network, the model has better prediction performance, and its evaluation indexes RMSE, MAE and R2 are 0.020 22, 0.014 33 and 0.992 31, respectively, which can provide data reference for fire probability prediction.

Key words: multi-feature parameter, whale optimization algorithm, genetic algorithm, fire probability prediction, BP neural network