基于GRNN神经网络的地铁疏散预警及对策研究

针对国内地铁车站客流无序性和突发性的现状,提出基于广义回归神经网络GRNN的地铁车站客流预警模型。以南京地铁全线网某时段客流数据为输入样本,运用GRNN神经网络进行训练与测试,得出预测数据并对比实际数据进行误差分析。结果表明:预测数据拟合,精度可行。将预测数据与南京地铁实时客流预警系统相结合,提出突发性大客流应急情况下的运营服务对策措施,为地铁运营管理单位避免突发大客流造成人员踩踏、恐慌等事故提供参考。

The passenger flow warning of metro station based on GRNN method and countermeasure research

Abstract In view of the domestic metro station situation, the passenger flow is disorderly and sudden. Based on the regression neural network(GRNN), the passenger flow warning model was introduced. Taking the ACC data of Nanjing Metro as example, using the GRNN model for data training and testing, the error analysis of predicted data and actual data was compared. Results show that the predicted data is fitting, the accuracy is feasible. Combining the predicted data with the real-time passenger flow waring system of Nanjing metro, the countermeasures were derived. Providing a reference for metro operation and management company to avoid stampede accident which is caused by sudden passenger flow.