基于KICA的石油管道泄漏检测方法研究

On Oil Pipeline Leak Detection Based on KICA

石油管道监测参数之间呈现非线性、非高斯性的特性,导致传统石油管道检测技术存在误报率较高的现象.文章提出一种核独立分量分析(KICA)的石油管道泄漏检测方法.首先通过核方法将压力、温度和流量参数非线性映射到具有更好线性结构的特征子空间;然后利用独立分量分析法提取核特征子空间的独立元信息,根据白化处理的得分方差对独立元进行排序,分别计算I2、SSPE两个统计量,由核密度估计管道正常运行条件下I2、SSPE的对应99%置信水平的控制限;最后通过在线检测I2、SSPE是否超限建立泄漏检测模型.以某一输送场站采集数据进行实验仿真,结果表明:基于核独立分量分析的石油管道检测方法误报率较低,为石油管道安全监测提供了新的研究方法.

Characteristics of the nonlinear and non-gaussian lead to the problem of high rate of false positives in traditional oil pipeline detection.A kernel independent component analysis method of oil pipeline leak detection has been put forward in this paper.Firstly,by kernel method pressure,temperature and flow parameters of nonlinear have been mapped to better the feature subspace of linear structure.Secondly,independent component analysis been used to extract information of independent component in kernel feature subspace.Thirdly,independent component has been sorted according to the variance of whiten score to obtain two statistic I2,SSPE,respectively depending on kernel density estimation with the corresponding control limit of 99% confidence interval.And finally,fault detection model has been established by judging whether I2 and SSPEare out of control limit or not.Taking data of a transportation for simulation,results show that oil pipeline detection method based on kernel independent component analysis low rate of false positives,which offers a novel method for oil pipeline safety monitoring.