基于LSTM的气液两相流液相流量测量方法  

Liquid Phase Flow Measurement Method of Gas-liquid Two-phase Flow Based on LSTM

作　　者：仝卫国[1] 曾世超 李芝翔 朱赓宏 TONG Wei-guo;ZENG Shi-chao;LI Zhi-xiang;ZHU Geng-hong(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《仪表技术与传感器》2021年第11期94-98,共5页Instrument Technique and Sensor

摘　　要：气液两相流参数测量的准确性和实时性对火力发电、化工、石油等工业生产具有重要意义。为提升气液两相流液相流量测量速度和精度,通过对电阻层析成像(electrical resistance tomography,ERT)技术和长短时记忆(long and short time memory,LSTM)神经网络的研究,提出了一种基于LSTM神经网络的气液两相流流量测量方法。利用16电极ERT系统分别采集了弹状流、泡状流两种流型对应的阵列电阻值,并将其作为LSTM神经网络的输入,电磁流量计采集的液相流量值作为理想输出,通过网络训练得到了2种流型下的液相流量测量模型。仿真实验结果表明:利用该方法在弹状流和泡状流下的均方根误差分别为5.359 m^(3)/h和4.088 m^(3)/h,平均引用误差低于3%,测量精度和速度优于BP神经网络,能够实现对两种流型下液相体积流量的准确测量。The accuracy and real-time measurement of gas-liquid two-phase flow parameters are of great significance to thermal power generation,chemical industry,petroleum and other industrial production.In order to improve the measurement speed and accuracy of gas-liquid two-phase flow,a gas-liquid two-phase flow measurement method based on long and short time memory(LSTM)neural network was proposed through ERT and LSTM neural network.Ray resistance values of slug flow and bubble flow were collected by a 16-electrode ERT system,which were used as the input of LSTM neural network,and the liquid flow values collected by an electromagnetic flow meter were used as the ideal output.The liquid flow measurement models of the two flow types were obtained by network training.The simulation results show that the root mean square errors in slug flow and bubble flow are respectively 5.359 m^(3)/h and 4.088 m^(3)/h,the average reference error is less than 3%,the measurement accuracy and speed are better than that of BP neural network,which can realize the accurate measurement of liquid volume flow under two flow patterns.

关 键 词：电阻层析成像 长短时记忆网络 两相流 流量测量 均方误差 

分 类 号：TB937[一般工业技术—计量学]