基于深度学习算法的轨道电路分路不良预测方法  

Poor Prediction Method for Track Circuit Splitting Based on Deep Learning Algorithm

作　　者：孔朗和 郭仲斌 邵迪 陈景周 

机构地区：[1]深圳地铁集团有限公司,广东深圳518000 [2]广州市中海达测绘仪器有限公司,广东广州511400 [3]广州铁路职业技术学院,广东广州510430

出　　处：《工业控制计算机》2022年第11期62-64,共3页Industrial Control Computer

摘　　要：轨道电路分路不良是影响地铁运营安全的重要因素,对地铁运营保障有着至关重要的作用。当前的分路不良预警做法大概分为2大类:(1)采用轨道电路的电压值、车站的温度/湿度、钢轨长度、轨道电路类型、钢轨阻抗、道渣电阻来做预警;该做法往往存在数据量小且成本较高的缺陷;(2)采用传统机器学习方法(SVM),通过微机监测采集的开关量、模拟量数据、电压电流检测曲线来做预警。然而这类方法往往与实际板卡电压的测量有一定的误差,从而导致预测准确性不高。为了弥补以上方法的不足,提出了基于transformer和LSTM网络的深度学习架构来对轨迹电路分类进行预测。该模型充分利用transformer自注意力机制和多头注意力机制能够更好地捕抓电压微小的变化特征的优点,以及LSTM网络天生适合处理时序特征的特点,对轨道电路接收板的电压值进行预测。Poor shunting of track circuit is an important factor affecting the safety of subway operation,which plays a vital role in the guarantee of subway operation.The current early warning methods for poor shunting are roughly divided into two categories:(1)early warning is made by using the voltage value of track circuit,temperature and humidity of station,rail length,track circuit type,rail impedance and ballast resistance;This approach often faces the defects of small amount of data and high cost.(2)The traditional machine learning method(SVM)is used to do early warning through microcomputer monitoring the collected switching value,analog value data and voltage and current detection curve.However,this kind of method often has a certain error with the measurement of the actual board voltage,which leads to the lack of prediction accuracy.In order to make up for the shortcomings of the above methods,this paper proposes a deep learning architecture based on transformer and LSTM network to predict the trajectory circuit classification.The model makes full use of the advantages that transformer self attention mechanism and multi head attention mechanism can better capture the characteristics of small changes in voltage,and LSTM network is naturally suitable for processing timing characteristics to predict the voltage value of track circuit receiving board.

关 键 词：轨道电路 分路不良 解决方案 

分 类 号：U231.8[交通运输工程—道路与铁道工程] TP18[自动化与计算机技术—控制理论与控制工程]