动态数据驱动的森林火灾仿真框架及其实现

提出了动态数据驱动的森林火灾仿真方法，在仿真中将实际系统的实测数据反馈到模型中，以提高模型的准确性。仿真运行了模型的多个样本，模型的输入是大量动态的输入而不是传统固定的输入，每一步仿真结束后采用统计的方法筛选出较优样本集作为下一步仿真各样本的初始条件。该方法包含以下关键部分：如何基于不准确模型获得较为可信的仿真结果、模型计算结果准确性度量方法、应用系统的模型和大量仿真样本运行带来的计算复杂度和计算量大的问题等。为解决上述问题，基于离散系统仿真规范提出并实现了一个动态数据驱动的森林火灾仿真框架。实验表明，上述方法将基于传感器测量到的真实火灾数据动态反馈至仿真系统中，获得了更可信的仿真结果。

ARCHITECTURE OF DYNAMIC DATA DRIVEN SIMULATION FOR WILDFIRE AND ITS REALIZATION

Dynamic data driven simulation (DDDS) is proposed to improve the model by incorporaing real data from the practical systems into the model. Instead of giving a static input, multiple possible sets of inputs are fed into the model. And the computational errors are corrected using statistical approaches. It involves a variety of aspects, including the uncertainty modeling, the measurement evaluation, the system model and the measurement model coupling,the computation complexity, and the performance issue. Authors intend to set up the architecture of DDDS for wildfire spread model, DEVS-FIRE, based on the discrete event specification (DEVS) formalism. The experimental results show that the framework can track the dynamically changing fire front based on fire sensor data, thus, it provides more accurate predictions.