【作者】 周宇飞；

【导师】 唐小明；

【作者基本信息】 中国林业科学研究院 ， 森林经理学， 2010， 博士

【摘要】 森林火灾是一种自然灾害,它会损坏极难恢复的森林资源,并给人们的生命财产造成严重的威胁。为了减少森林火灾的损失,世界各国都在对林火进行研究。这其中,使用计算机模拟森林火灾的蔓延,从而对火场发展的状况进行预测已经成为目前森林防火的重要研究方向之一。然而要对森林火灾这样复杂的系统行为进行精确分析和预测是很困难的。目前为止,人们仍然缺乏对林火蔓延过程的精确描述能力。将动态数据驱动技术应用在林火空间扩散模拟中,能够在现有林火蔓延模型的基础之上,动态地吸收实时的火场数据对模拟过程进行控制；同时对蔓延模拟的阶段性结果进行评价,选择最优模拟结果；并能通过对数次模拟的模拟结果分析和知识积累,不断优化现有蔓延模拟模型。这样将林火扩散模拟和林火野外实时监测有机结合起来,使模拟可以实时动态接收野外监测数据并做出响应,就可以输出更为准确的模拟预测结果,从而为科学的火场决策指挥提供可靠依据。本文致力于建立林火蔓延模拟和实际林火动态发展之间的相互协作的林火蔓延计算机模拟系统,解决建立该系统过程中所要用到的模型库和数据库建库、模型并行蔓延模拟、林火动态数据获取处理及调用、模拟结果空间位置精度评价及模型参数修正等一系列问题。文章首先论述了动态数据驱动林火蔓延自适应模拟技术的内涵,确定了动态数据驱动技术的林火蔓延模拟运行机制及模拟实施的具体流程；然后总结了动态数据驱动林火蔓延自适应模拟系统的关键技术主要包括动态数据驱动林火蔓延自适应模拟模型库及参数库建库技术、自适应蔓延模拟应用组件开发技术、林火蔓延模型模拟评价技术及林火蔓延模拟参数调整修正技术；最后分别对这些技术展开研究。在动态数据驱动林火蔓延自适应模拟库建库方面设计了一种统分结合的模拟模型库建立方法,即利用林火蔓延数学模型管理总表和针对特定林火蔓延数学模型描述表,并结合各模型算法动态数据链接库调用进行模型库建设；对林火蔓延模拟数据按照输入及存储方式进行了分类,提出了每种数据的获取与处理方法,并提出了一种文件系统和存储表格的管理方法实现林火数据的管理；设计建立了林火蔓延模拟评价库,阐述了林火蔓延模拟模型适宜性选择知识库的建立方法,并设计建立林火蔓延模拟空间精度评价库。在自适应林火蔓延模拟应用组件开发方面利用阶段时间划分建立了模型动态调用数据驱动的方法,通过阶段时间控制实现模拟过程中模型模拟数据动态与火场数据进行对接；利用多线程编程的方法实现多个林火蔓延数学模型进行并行模拟,并使用基于惠更斯原理的波动传播模型和林火蔓延椭圆模型实现林火蔓延模拟；依据椭圆波动模型提出了基于初始蔓延速度和最大蔓延速度的模型统一的速度表达方式,并指明了一些模拟相关的计算机算法；对模型调用的输入输出接口进行规定,并设计了林火蔓延模拟空间表达方法。在林火蔓延模型模拟评价及林火蔓延模拟参数调整修正方面提出了林火蔓延模拟空间精度评价方案,指明了利用火场面积叠置率和火场边缘相似程度的量化值来对林火模型模拟的模拟空间精度进行评价的方法；研究了模拟误差产生的主要来源,提出了利用阶段时间控制程序输入实现模型利用动态火场数据做自适应调整的方法,提出了利用多次模拟结果专家干预下的模拟参数调整及模型在模拟过程中依据既有环境做自适应调整达到优化模拟的方法。本文最后建立了一个动态数据驱动林火蔓延自适应模拟原型系统,以发生在河北丰宁和北京怀柔交界地区的一起森林火灾做为火场实验样本,并利用王正非模型和Rothermel模型这两种不同的林火蔓延模型并行这次火灾进行模拟,指明了动态数据驱动林火蔓延模拟的基本模拟实现过程。更多还原

【Abstract】 Forest fire is a natural disaster, it is an extremely serious threat to forest resources and people’s lives. In order to reduce the loss causing by forest fires, there are many people carrying on the study of wildfire all around the world. Using computer to simulate the spread of forest fires and taking the forecast of the forest fire behavior have altogether become an important research direction. However, it is difficult to analysis and forecast a complex system such as forest fire. There is still a lack of the capability of describing forest fire spreading accurately.Using dynamic data-driven technology in the space of forest fire spreading simulation can absorb the dynamic real-time data to control the simulation process based on the current forest fire spreading models. At the same time, it can evaluate the results of simulation stage, and select the optimal results. It can optimize the existing simulation model through the analysis and knowledge accumulation of several simulation results. The combine of forest fire spreading simulation and the real-time monitoring can output more accurate forecasting results and provide reliable decision-making for commanding.The thesis points out what establishing a computer system of forest fire spreading simulation that the actual fire data and fire simulation can cooperate is about. Then Solving a series of problems in the process of establishing the system such as building a model database, models parallel simulation, forest fire dynamic data acquisition, processing and calling, spatial accuracy assessment and model parameter modification.The thesis first discusses the content of dynamic data-driven forest fire spreading simulation technology, determining the operational mechanism and specific process of dynamic data-driven forest fire spreading simulation; Then, it sums up the key technologies of dynamic data-driven forest fire spreading system, including establishing simulation model database and data database, adaptive application component technology of fire spreading simulation, simulation evaluation techniques and parameter adjustment technique. Finally it shows the study of these technologies respectively.In order to build up a model database of forest fire spreading dynamic data-driven adaptive simulation, the thesis designed a model database establishing method with unity and flexibility. That is, we can use model management table, specific model description table and DLL for algorithm of each model to build the model database. We classify the data according to the data input and stored method, bring forward the methods for each kind of data acquisition and processing, and put forward a file and storage form management to achieve forest fire data management. At the same time, the thesis designs and establishes a simulation evaluation database, describes the method of the establishment of model appropriate selection knowledge database, design and build up the space accuracy evaluation database.In the application of forest fire adaptive spreading simulation component development, we used a phase-time control to realize the method of dynamic data-driven, called by phase-time control of analog data, to connect the simulation data and real fire data with phase-time control. Using multi-threaded programming method to realize the parallel simulation of multiple fire spread models, and using wave propagation model that based on Huygens principle and elliptical model to achieve forest fire spreading simulation. It also puts forward a number of related computer simulation algorithm and provides the model input and output interfaces and designs the space expression of forest fire spread simulation.Thesis puts forward a spatial accuracy evaluation program of fire model simulation evaluation and fire parameter adjustment, points out the method of model space precision evaluation by using the quantitative value of overlay area rate and fire edge similarity. Thesis studies the main source of simulation error, puts forward the approach of using phase-time control to implement model dynamic fire data input, puts forward using simulation again and again and experts intervention to adjust the simulation parameters. During the simulation, the model can optimize the simulation based on the environment change.Thesis has built up a prototype system of forest fire adaptive spreading simulation based on dynamic data-driven, using Wangzhengfei Model and Rothermel Model to simulate a fire that happened at the border of Fengning Hebei and Huairou Beijing, pointing out the realization of forest fire spreading dynamic simulation based on dynamic data-driven.更多还原