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 ›› 2024, Vol. 43 ›› Issue (4): 535-540.

• • 上一篇    下一篇

森林火灾救援直升机吊桶灭火任务调度优化研究

徐浩, 瞿菁菁, 王明惠, 朱新平   

  1. (中国民用航空飞行学院,四川 广汉 618307)
  • 出版日期:2024-04-15 发布日期:2024-04-15
  • 作者简介:徐 浩(1996- ),男,江苏泰兴人,中国民航飞行学院空中交通管理学院研究生,主要从事森林火灾通航救援飞行任务调度优化方面的研究,四川省德阳市广汉市南昌路四段46号,618307。
  • 基金资助:
    国家重点研发计划课题(2022YFB2602004);国家重点研发计划课题(2021YFB2601704);四川省科技计划项目(2021YFS0391);中央高校基本科研经费项目(ZHMH2022-008,J2023-047)

Research on optimization of task scheduling for forest fire rescue helicopter bucket firefighting

Xu Hao, Qu Jingjing, Wang Minghui, Zhu Xinping   

  1. (Civil Aviation Flight University of China, Sichuan Guanghan 618307, China)
  • Online:2024-04-15 Published:2024-04-15

摘要: 鉴于森林火灾突发性强、处置困难等特点及火灾救援中直升机任务调度效率有待提高的现状,为降低森林火灾造成的损失,研究直升机吊桶灭火任务调度优化的方法。结合火场面积、火情状况、救援力量等因素,建立火场空域栅格模型。针对森林火灾场景中火场动态变化的情况,提出一种基于深度强化学习(DQN)的动态任务调度优化方法,该方法通过对起火栅格的燃烧时间约束、灭火所需水量等属性进行实时量化,不断更新全局环境,形成动态决策依据,生成任务调度方案,提升直升机吊桶灭火的效率。仿真结果表明,与传统算法中表现较好的贪婪算法相比,本文方法能在火场动态变化的场景下,降低了13.9%的火灾损失率,有效减少了森林火灾造成的资源损失。

关键词: 任务调度, 深度强化学习, 应急救援, 动态决策, 直升机吊桶

Abstract: In view of the characteristics of forest fires, such as strong sudden occurrence and difficult disposal, and the current situation that helicopter task scheduling efficiency needs to be improved in fire rescue, in order to reduce the loss caused by forest fires, the optimization method of helicopter bucket fire extinguishing task scheduling was studied. Combined with fire area, fire situation, rescue force and other factors, a grid model of fire airspace was established. Aiming at the dynamic changes of fire sites in forest fire scenes, a dynamic task scheduling optimization method based on deep reinforcement learning (DQN) is proposed. The method continuously updates the global environment, forms a dynamic decision basis, and generates a task scheduling scheme by quantifying the burning time constraints of the fire grid and the water required for extinguishing the fire in real time. Improve the efficiency of helicopter bucket fire fighting. The simulation results show that compared with the greedy algorithm in the traditional algorithm, the proposed method can reduce the fire loss rate by 15.4% under the dynamic change of fire site, and effectively reduce the resource loss caused by forest fires.

Key words: task scheduling, deep reinforcement learning, emergency rescue, dynamic decision, helicopter bucket