基于优化支持向量机的管道安全威胁事件识别  

Identifi cation of threats to pipeline based on optimized support vector machine

作　　者：刘路[1] 李新民[2] 何岚[3] 周琰[1] 孙巍[1] 田孝忠[1] 

机构地区：[1]中国石油管道科技研究中心油气管道输送安全国家工程实验室,河北廊坊065000 [2]中国石化管道储运公司黄岛油库,山东青岛266000 [3]中国石油天然气管道科学研究院,河北廊坊065000

出　　处：《油气储运》2014年第11期1225-1228,共4页Oil & Gas Storage and Transportation

摘　　要：如何有效识别人工挖掘、机械挖掘和车辆经过等油气管道安全威胁事件,对基于相干瑞利的管道光纤安全预警系统至关重要。提出一种基于优化支持向量机的管道安全威胁事件识别方法,对管道光纤安全预警系统采集到的管道沿线振动信号进行分析,提取各频段的归一化能量与信号持续时间作为特征向量,利用人工蜂群算法对支持向量机的惩罚因子和核函数参数进行优化,采用优化后的支持向量机对特征进行分类。在港枣成品油管道开展安全试验,获得了90.7%的分类正确率,证明了该方法的有效性和工程应用价值。Effective identification of threats to the oil/gas pipeline due to artificial excavation, mechanical excavation and vehicles passing is of vital importance to the pipeline optical fiber security precaution system based on coherent Rayleigh. This paper presents a pipeline threat identification method based on optimized support vector machine. The method is used to analyze the vibration signals along the pipeline collected by pipeline optical fiber safety precaution system, extract the normalization energy and signal duration of each frequency band as feature vector, use artificial bee colony algorithm to optimize the penalty factor and kernel function parameter of support vector machine and use the optimized support vector machine to classify features. The safety test was carried out on the GangZao Products Pipeline with the accuracy of classification up to 90.7%, which proves the effectiveness and engineering value of the proposed method.

关 键 词：油气管道 安全 支持向量机 人工蜂群算法 参数优化 

分 类 号：TN247[电子电信—物理电子学] TN911.6