基于Weibull分布和SVR的城轨列车自动进站停车精度预测方法

Weibull Distribution and SVR Based Prediction Method for Parking Error of Metro Train

通过分析城轨列车自动停车精度分布变化,研究停车精度预测问题,提出一种基于支持向量回归的列车自动进站停车精度预测方法。对停车精度数据进行分期预处理,进而采用Weibull分布拟合各期数据,由此得到两个分布参数的时间序列,通过支持向量回归算法对参数序列构建预测模型,实现对进站停车精度分布预测。采用北京地铁某条线一列车的自动进站停车数据对提出的模型进行仿真验证,结果表明预测分布与真实分布相似性均值可达0.9533,验证了提出的预测方法的有效性,为城轨列车自动进站停车精度变化提供了一种科学、高效的预测方法。

This paper studied the problem of prediction of the parking error of metro train by analyzing the trend of parking error distribution,and then proposed a support vector regression(SVR)-based parking error prediction method.First,the data set of parking error was divided into groups and was preprocessed,and the data for each single group was fitted by using the Weibull distribution.Thus,the time series of two distribution parameters were obtained.Next,the support vector regression algorithm was used to construct a prediction model for the parameter sequence to realize the prediction of the parking error distribution.Finally,the parking error data from one metro train in Beijing Subway were employed to validate the proposed model.The results indicate that its performance is promising with the similarity of 0.9533 between the real and predictive distribution,which verifies the effectiveness of the proposed predictive method and provides a scientific and efficient mean for the prediction of parking error.