PSO优化的高速列车轴温灰色预测模型  

Bearing Temperature Grey Prediction Model of High-speed Train Based on PSO Optimization

作　　者：邓佳林 邹益胜[1] 黄治光 张继冬 张笑璐 王超[1] DENG Jialin;ZOU Yisheng;HUANG Zhiguang;ZHANG Jidong;ZHANG Xiaolu;WANG Chao(Institute of Advanced Design and Manufacturing, Southwest Jiao Tong University, Chengdu 610000 Sichuang, China;CRRC Qingdao Sifang Co., Ltd., Qingdao 266000 Shandong, China)

机构地区：[1]西南交通大学机械工程学院先进设计与制造研究所,成都610000 [2]中车青岛四方机车车辆股份有限公司,山东青岛266000

出　　处：《铁道机车车辆》2019年第4期31-35,共5页Railway Locomotive & Car

基　　金：国家重点研发计划——复杂环境下轨道车辆全生命周期能力保持与优化研究(2017YFB1201201-06)

摘　　要：轴承温度预测是保障高速列车安全运行的重要手段。考虑到GM(1,1)模型建模机理存在着一定的缺陷,以及对建模数据有一定的单调性要求,对呈现较大波动的数据序列预测结果不太理想的问题。在GM(1,1)预测模型的基础上,提出了一种粒子群算法(PSO)优化的灰色预测方法。利用多项式对GM(1,1)模型进行修正,重构灰色模型的时间响应序列,再利用粒子群算法对重构模型进行求解,并据此构建了一种高速列车轴温预测模型。以高速列车实际轴温数据对模型进行验证,验证结果表明:模型相较于GM(1,1)模型的预测精度有明显的提高,其中5min预测的平均绝对误差由6℃降低到5℃,降幅为16.7%。平均相对误差由9.1%降到了7.8%,降幅为14.3%;最大绝对误差由20℃降低到18.6℃,降幅为7%。预测误差的方差由24.6降低到了20.6,降幅为16.3%,表明误差分布更加集中。It’s an important way for the safe operation of high-speed trains to predict bearing temperature.Considering the drawbacks of GM(1,1)model modeling mechanism,the monotonicity requirements for modeling data,and the poor performance on fluctuating data sequences,an improved method of bearing temperature is presented based on GM(1,1)and Particle Swarm Optimization(PSO).In this paper,GM(1,1)model is modified by polynomial to reconstruct the time response sequence,then PSO algorithm is used to solve the reconstructed model,and a prediction model of high-speed train bearing temperature is constructed.Finally,the method is validated by the actual bearing temperature data of the high-speed train.The results show that the proposed method achieves better prediction accuracy:the average absolute error in 5 minute is reduced from 6 ℃to 5 ℃,a decrease of 16.7%;the average relative error is reduced from 9.1%to 7.8%,a decrease of 14.3%;the maximum absolute error is reduced from 20℃to 18.6℃,a decrease of 7%;the variance of the prediction error is reduced from 24.6 to 20.6,a decrease of 16.3%,indicating the error distribution is more centralized.

关 键 词：高速列车 灰色模型 PSO 轴温预测 

分 类 号：U266.2[机械工程—车辆工程] U260.4[交通运输工程—载运工具运用工程]