有杆抽油系统工况诊断的数学模型研究  

Study on the Working Condition Diagnosis Method for Suck Rod Pumping System Based on Wavelet Moment Features

作　　者：刘卓[1] 罗明良[1] 陆雪皎[1] 范坤坤 刘飞[1] 

机构地区：[1]中国石油大学(华东)石油工程学院,山东青岛266580

出　　处：《数学的实践与认识》2013年第15期41-48,共8页Mathematics in Practice and Theory

基　　金：中国石油大学(华东)自主创新科研计划项目(14CX06029A)

摘　　要：小波矩特征法是一种既能反映图像的全局信息,又能反映局域信息,并且具有旋转、平移和缩放不变形的图像特征提取方法,在模式识别中能大大提高对近似目标识别的能力.为保证油井正常生产,精确快速地诊断出有杆抽油系统的工况,基于小波矩特征和自组织竞争神经网络(SCNN)建立了一种泵功图诊断模型.利用Matlab将悬点示功图转化为泵功图,然后基于小波矩特征法提取泵功图的小波矩特征并优选出合适的小波矩特征量,使用SCNN识别诊断.仿真研究表明,模型能够可靠的诊断有杆抽油系统的工况,实现简单,准确率高,适用于现场诊断,具有一定的工程应用前景.For more precise and faster diagnosis of suck rod pumping system, maintaining normal production, build a new working condition diagnosis model base on self-organizing competitive neural network SCNN and wavelet moment feature. Transform dynamometer cards into pump work indicating diagrams based on Matlab, then extract the wavelet features of pump work indicating diagrams and prefer the suitable wavelet moments based on moment wavelet, then diagnosis by SCNN. The model designed could diagnose accurately and has characteristics of simple realization, high accuracy rate from data obtained experimentally. It＇s convenience to field diagnostics and can be applied to engineering practice.

关 键 词：小波矩特征 自组织竞争神经网络 泵功图 故障诊断 模式识别 

分 类 号：TE933[石油与天然气工程—石油机械设备] O242.1[理学—计算数学]