林火蔓延中的数据同化方法研究

发布时间：2018-07-25 20:22

【摘要】：森林火灾严重破坏生态系统及人类的生命财产安全,因此研究林火蔓延机理,对林火蔓延发展进行准确地预测和模拟,可以有效地制定扑火决策,控制火场的持续蔓延。然而作为一种复杂系统,准确地预测林火蔓延依赖于很多因素,包括地理信息数据,可燃物数据、气象条件、高效准确的林火蔓延模型等。但由于林火蔓延的动态性和复杂性,很难准确地获得这些数据。本文以湖南地区的森林火灾为研究对象,在不考虑灭火的前提下,选择基于离散事件系统规范(Discrete Event System Specification,DEVS)的王正非模型(简称 DEVS-王正非模型)为林火蔓延的仿真模型,研究了林火蔓延仿真预测过程中的数据同化方法。本文的主要工作如下:(1)针对DEVS-王正非模型林火蔓延预测精确度较低的问题,采用分布式存储及计算框架,结合粒子滤波算法原理,搭建基于Hadoop的林火蔓延数据同化方法框架。(2)利用分布式文件存储技术对采集的林火蔓延仿真所需的模型数据进行存储管理,以更方便有效的形式供数据同化系统调用。(3)基于DEVS-王正非林火蔓延模型,以火场外围火线为林火蔓延的状态,结合分布式计算框架MapReduce和粒子滤波算法,将火线及其外围温度作为比较参数,设计了采样算法和测量密度的计算方法,实现了林火蔓延仿真预测过程中的数据同化。本文选择影响林火蔓延的主要气象因子“风”作为动态数据,分别以风速和风向作为模拟因子,根据提出的林火蔓延数据同化方法,对其进行实验仿真模拟。实验结果表明提出的林火蔓延数据同化方法,能够有效的矫正由于气象数据采集误差而导致的林火蔓延仿真预测的不准确,从而提高林火蔓延仿真模型的预测精度。
[Abstract]:Forest fire seriously damages the ecosystem and the safety of human life and property. Therefore, studying the mechanism of forest fire spreading and accurately predicting and simulating the development of forest fire spread can effectively make fire suppression decision and control the continuous spread of fire field. However, as a complex system, the accurate prediction of forest fire spread depends on many factors, including geographic information data, combustible data, meteorological conditions, efficient and accurate forest fire spread model and so on. However, due to the dynamic and complexity of forest fire spread, it is difficult to obtain these data accurately. In this paper, the forest fire in Hunan province is taken as the research object. Without considering the fire extinguishing, Wang Zheng non-model based on discrete event system specification (Discrete Event System specification / devs) is selected as the simulation model of forest fire spread. The method of data assimilation in simulation prediction of forest fire spread is studied. The main work of this paper is as follows: (1) aiming at the problem of low accuracy of fire spread prediction in DEVS- Wangzheng non-model forest fire, the distributed storage and computing framework is adopted, and the particle filter algorithm is combined with the principle of particle filter algorithm. The method frame of forest fire spread data assimilation based on Hadoop is built. (2) using distributed file storage technology to store and manage the model data needed for forest fire spread simulation. (3) based on DEVS- Wang Zhengfei forest fire spread model, the fire line around the fire field is taken as the state of forest fire spread, and the distributed computing framework MapReduce and particle filter algorithm are combined. Taking the fire line and its peripheral temperature as the comparison parameters, the sampling algorithm and the calculation method of measuring density are designed, and the data assimilation in the process of simulation prediction of forest fire spread is realized. In this paper, the main meteorological factor "wind", which affects the spread of forest fire, is selected as dynamic data and wind speed and wind direction as simulation factors respectively. According to the proposed assimilation method of forest fire spread data, the experimental simulation is carried out. The experimental results show that the proposed method of forest fire spread data assimilation can effectively correct the inaccuracy of simulation prediction of forest fire spread caused by meteorological data acquisition error, thus improving the prediction accuracy of forest fire spread simulation model.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S762;TP391.9

【相似文献】