基于人工蜂群算法优化的SVM管道风险评估  

An optimized SVM pipeline risk assessment method based on artificial bee colony algorithm

作　　者：艾月乔[1] 徐亮[1] 陶江华[1] 董润清 陈瑞波[1] AI Yueqiao;XU Liang;TAO Jianghua;DONG Runqing;CHEN Ruibo(PetroChina Pipeline Company;Petroleum Engineering School,Southwest Petroleum University)

机构地区：[1]中国石油管道公司 [2]西南石油大学石油与天然气工程学院

出　　处：《油气储运》2019年第5期510-515,共6页Oil & Gas Storage and Transportation

基　　金：中国石油管道公司科技攻关项目"光缆地表快速定位技术研究";20170203

摘　　要：管道风险评估是管道风险管理的重要组成部分,其目的是通过对风险的调查和分析,识别可能导致管道事故的重要因素,使得管道风险管理更加科学化。为了对管道日常运行状态风险进行准确评估,提出了一种利用人工蜂群算法优化的支持向量机(Support Vector Machine,SVM)管道安全风险评估方法:建立管道风险评估模型,从工艺运行角度收集成品油管道、正反输原油管道、掺混输送原油管道的工艺运行特征,并形成样本特征集合。对这4种类型管道的特征集合进行试验验证,结果表明:在小样本情况下,采用基于人工蜂群算法优化的SVM管道风险评估方法正确率较高,并具有良好的普适性,能够根据管道实际运行状态给出正确的风险评估结果。(图3,表1,参21)Pipeline risk assessment is one important part of pipeline risk management, and it is aimed at identifying the important factors that possibly lead to pipeline accidents by means of risk investigation and analysis to make the pipeline risk management more scientific. In this paper, an optimized SVM(Support Vector Machine) pipeline risk assessment method based on artificial bee colony algorithm was proposed in order to carry out accurate assessment on pipeline risks in the state of daily operation. In this method, the pipeline risk assessment model is established, the technological operation characteristics of products pipelines, forward/backward crude oil pipelines and mixed-transportation crude oil pipelines are collected from the viewpoint of technological operation, and the sample characteristic set is built up. Then, 4 types of pipeline characteristic sets were tested. It is indicated that in the case of small sample, the optimized SVM pipeline risk assessment method based on artificial bee colony algorithm has higher accuracy and good applicability, and it can provide the accurate risk assessment result according to the actual running state of pipelines.

关 键 词：油气管道 运行风险 管道失效特征 人工蜂群算法 SVM 

分 类 号：TE832[石油与天然气工程—油气储运工程]