基于LSTM的直升机洒水落点分布预测  

Prediction of water-dropping point distribution based on LSTM from helicopter

作　　者：刘均[1] 王熇蒙 伊笑 艾敏敏 LIU Jun;WANG Hemeng;YI Xiao;AI Minmin(School of Physics and Electronic Engineering,Northeast Petroleum University,Daqing 163318,China)

机构地区：[1]东北石油大学物理与电子工程学院,大庆163318

出　　处：《火灾科学(中英文)》2024年第4期243-251,共9页Fire Safety Science

基　　金：黑龙江省自然科学基金项目(LH2023A002)。

摘　　要：针对传统灭火飞机洒水模型复杂性高、解算时间长的问题,为提高洒水灭火效率,提出了基于LSTM的直升机洒水落点分布预测方法。首先根据伯努利原理计算吊桶洒水速度随时间的变化,再根据射流原理和Legendre等人的研究,最终得到水体的地面分布。用边缘线上的6个点表示有效灭火区,引入LSTM网络,使用大量的水体状态参数6点坐标进行网络训练,建立神经网络模型,预测洒水地面分布。仿真测试表明,LSTM神经网络模型预测准确性高,且相较于数值积分法计算和BP模型预测都能更快地预测出地面水体分布,为之后的灭火方案制定提供了参考。A prediction method for the distribution of helicopter sprinkler drop points based on LSTM was proposed to improve sprinkling extinguishing efficiency.Firstly,the bucket sprinkling speed variation with time was calculated according to the Bernoulli principle.Then,the ground distribution of the water body was obtained according to the jet principle and the research of Legendre et al.The effective fire extinguishing area is represented by 6 points on the edge line,and the LSTM network is introduced.A neural network model is established by training the network with a large number of 6-point coordinates of water state parameters to predict the distribution of sprinkler water.The simulation test shows that the LSTM neural network model has high prediction accuracy and can predict the distribution of surface water faster than the numerical integration method calculation and BP model prediction time,which provides a reference for the formulation of fire extinguishing plans in the future.

关 键 词：LSTM网络 预测 洒水 有效灭火区 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] X915.5[自动化与计算机技术—计算机科学与技术]