基于小波神经网络的油气水三相流相含率的软测量方法  

Soft Measurement Method for Phase Volume Fraction of Oil-Gas-Water Three-Phase Flow Based on Wavelet Network

作　　者：周云龙[1] 袁俊文[2] 

机构地区：[1]东北电力大学能源与动力工程学院 [2]东北电力大学自动化工程学院,吉林吉林132012

出　　处：《化学反应工程与工艺》2010年第3期218-222,共5页Chemical Reaction Engineering and Technology

基　　金：吉林省科技发展项目(20040513)

摘　　要：利用高速摄影仪对垂直上升管中油气水三相流的流动进行了动态图像的拍摄，提取每一帧图像的灰度均值组成灰度时间序列，并从时间序列中提取了能反映油气水三相流流动特性的统计和分形特征量，将这些特征量作为人工神经网络的输入量。在水的体积流量为1．32～12．15m3／h，油的体积流量为0．01～0．43m3／h，空气的体积流量为0．75～2．5m3／h条件下，采用小波神经网络作为相含率预测模型，结果表明，小波神经网络的预测值与测试值非常吻合，含气率预测最大误差为3．57％，含水率最大误差为3．3％，较好地实现了油气水三相流相含率的预测，为油气水三相流相含率测量提供了一种有效的软测量方法。Shooting the dynamic image of the oil-gas-water three-phase flow with the application of high-speed camera was in the vertical upward pipe. Statistics and fractal features reflecting flow characteristics of oil-air-water three-phase flow were extracted from time series of image gray-scale fluctuating signals, which consisted of gray averages extracted from each frame image. All extracted features were taken as t he input of artificial neural network. Under the conditions of water flow rate for 1.32-12.15 m3/h, oil flow rate for 0.01-0.43 m3/h and air flow rate for 0.75-2.5 m3/h, the model of wavelet neural network was used to prediction of phase volume fraction. The results showed that gas fraction was predicted with an error of 3.57% , and water fraction with an error of 3.3%, good phase volume fraction prediction results of oil-gas-water three-phase flow were obtained. This study provided an effective way to measure the phase volume fraction of three-phase flow by soft measurement.

关 键 词：油气水三相流 动态图像 相含率 软测量 

分 类 号：O359.1[理学—流体力学]