摘要
接触网是铁路牵引供电系统的薄弱环节,对接触网缺陷的历史数据进行挖掘,有利于对接触网历史运行情况和未来缺陷发生趋势进行预测,为实现接触网计划性维修和预防性检修提供参考。考虑到接触网缺陷既取决于当前状态,又具有长时序依赖的特点,本文提出一种基于长短期记忆神经网络(LSTM)的接触网缺陷趋势预测方法,为接触网单项缺陷分项进行中短期预测。通过实例检验,运用该方法能够实现对缺陷趋势的准确、可靠预测。
OCS is the weak point of railway traction power supply system.Mining the historical data of OCS defects is conducive to predict the historical operation and future defect trend of OCS,and provide reference for realizing planned OCS maintenance and preventive maintenance.With consideration that the OCS defects not only depend on the current state,but also have the characteristics of long-term time sequence dependence,a trend prediction method of OCS defects based on long-term and short-term memory neural network(LSTM)is proposed to predict the medium-term and short-term OCS defects.Through the example test,this method can realize the accurate and reliable prediction of defect trend.
作者
薛逸凡
高仕斌
XUE Yifan;GAO Shibin
出处
《电气化铁道》
2021年第6期64-68,共5页
Electric Railway
