随机OD需求下城市道路交通网络设计问题的BP神经网络算法研究

Research on back-propagation neural network algorithm for continuous transportation network design problem under stochastic OD demand

本文应用反向传播(back-propagation,BP)神经网络方法研究随机Origin-Destination(OD)需求的城市道路交通网络设计问题.假设OD交通需求量服从某一分布,首先通过蒙特卡罗模拟的方法得到训练样本对神经网络进行训练并用测试样本测试所得到的神经网络,然后运用训练好的神经网络预测了系统总出行时间.将Nine-node网络作为测试网络,比较分析了不同样本规模下BP神经网络与蒙特卡罗模拟方法得出的结果和OD交通需求服从不同分布时系统总出行时间的分布情况.结果表明,BP神经网络的方法能够较好预测系统总出行时间的分布,且与蒙特卡罗模拟计算相比,能够产生一个合理且较窄的预测区间;当OD交通需求服从均匀分布、正态分布和对数正态分布时,系统总出行时间分布较为相似,接近正态分布.

This paper studies back-propagation(BP) neural network for continuous transportation network design problem under stochastic Origin-Destination(OD) demand.Assuming that the demand between every OD follows some distribution,the paper trains the neural network by the obtained training samples via Monte Carlo simulation,and then applies the trained BP neural network to predict the total system travel time.Using Nine-node network as a test network,this paper compares distributions of total travel time by BP neural network and Monte Carlo simulation when OD traffic demand respectively follows uniform,normal and log-normal distributions.The results show that the BP neural network method can obtain a better distribution of the total system travel time when solving the transportation network design problem,and can yield a more accurate,narrower range than the traditional Monte Carlo simulation calculation.Besides,the distribution of total travel time is more like the normal distribution when OD traffic demand is respectively subject to the above three distributions.