基于改进GM(1,1)和SVM的轨道电路故障最优组合预测模型研究  

Research on fault optimal combination prediction model of ZPW-2000A track circuit based on improved metabolic GM(1,1) and SVM

作　　者：黄斌[1] HUANG Bin(Liuzhou Railway Vocational and Technical College,Liuzhou 545616,China)

机构地区：[1]柳州铁道职业技术学院

出　　处：《铁道科学与工程学报》2019年第11期2852-2858,共7页Journal of Railway Science and Engineering

基　　金：广西高校中青年教师基础能力提升项目(2017KY1239);广西高等学校千名中青年骨干教师培育计划资助项目;广西高等学校高水平创新团队及卓越学者计划

摘　　要：针对传统电务部门对ZPW-2000A型轨道电路维护中所沿用的“故障修”及“定时修”在保证行车安全、提高运营效率及经济性等方面的不足,采用改进GM(1,1)和支持向量机(SVM)模型进行最优化权重组合,对轨道电路进行故障预测。首先,对GM(1,1)模型进行滑动平均法改进,将已知轨道电路的状态序列分别送入改进GM(1,1)和SVM模型进行序列预测。其次,分别对2种预测算法进行最优组合。最后,以2个案例进行模型的测试,验证了组合模型方法的可行性与有效性。In view of the shortcomings of the“fault repair”and“timing repair”used by the traditional electrical department in the maintenance of ZPW-2000A track circuit in ensuring traffic safety,improving operation efficiency and economy.This paper adopted the improved GM(1,1)and support vector machine(SVM)model to carry out the optimal weight combination for the fault prediction of track circuit.Firstly,the GM(1,1)model was modified by the sliding average method,and the known state sequence of the track circuit was fed into the improved GM(1,1)and SVM models for sequence prediction.Secondly,the optimal combination of the two prediction algorithms was build.Finally,two cases were used to test the model to verify the feasibility and effectiveness of the combined model method.

关 键 词：轨道电路 改进GM(1 1)模型 支持向量机 故障预测 

分 类 号：U283.2[交通运输工程—交通信息工程及控制]