摘要： 有毒气体的扩散预测在应急响应中起着重要作用。现有的计算流体力学(CFD)方法存在计算耗时长等问题，无法快速进行毒害气体扩散预测。提出了一种利用深度学习技术进行有毒气体扩散预测的方法。根据有毒气体扩散原理，设计基于GRU的神经网络模型，实现快速、有效的气体扩散浓度的预测。将本文的方法在经典的公开数据集草原牧场数据集上进行验证，实验结果表明本文方法可实现较高精度的气体扩散浓度的预测，并且优于基于BP神经网络模型的方法。

中图分类号： TP18
文献标识码： A
DOI： 10.19358/j.issn.2096-5133.2021.04.007
引用格式： 陈立，陈贤富. 基于GRU神经网络的有毒气体扩散预测方法[J].信息技术与网络安全，2021，40(4)：42-45.

Prediction method of toxic gas diffusion based on gated recycle unit(GRU)

Abstract： The prediction of toxic gas diffusion plays an important role in emergency response. Existing computational fluid dynamics(CFD) methods have problems such as time-consuming calculations and cannot quickly predict the diffusion of toxic gases. This paper proposes a method for predicting the diffusion of toxic gases using deep learning technology. According to the principle of toxic gas diffusion, a neural network model based on GRU is designed to realize fast and effective gas diffusion concentration prediction. The method is verified on the classic open data set of grassland and pasture data set. The experimental results show that the method can achieve high-precision gas diffusion concentration prediction and be better than the method based on BP neural network model.

Key words : toxic gas diffusion；concentration prediction；GRU；deep learning