南昌市短时交通流预测方法研究

发布时间：2018-08-24 19:46

【摘要】：在智能交通系统中,交通流预测是一个重要的组成部分。实施交通信号控制、路径诱导、事故检测等的重要前提是快速而准确的短时交通流预测。但是,交通系统具有高度的复杂性、非线性和不确定性等特征,其是由人、车、路等多种客体组成的复杂系统,对交通流做出实时、准确地预测是目前智能交通领域研究的热点和难点之一。然而,由于短时交通流信息量大,不确定性噪声信号干扰强,加之城市道路网络的拓扑结构复杂,致使怎样实现城市道路短时交通流预测一直阻碍着智能交通的长远发展。为了解决上述问题,已经提出了许多预测方法,但因未考虑不确定干扰信号或城市路网复杂性对短时交通流所造成的影响,导致预测结果的实时性与准确性都不尽理想。本文采用Mallat算法,对短时交通流信号进行小波分解与重构,旨在滤掉短时交通流的强干扰噪声信号,此法可以提高短时交通流信息预处理速度及精度。针对交通流数据的复杂性和非线性特性,本文引入神经网络理论,利用其对非线性问题的良好处理能力来进行短时交通流的预测,是交通流预测的有效方法。综上,针对城市道路交通流量的时变性、复杂性以及非线性特点,为提高短时交通流预测的精准度,该文提出了一种基于小波去噪和自适应遗传算法优化BP神经网络的短时交通流组合预测模型。利用小波变换可以将交通流分解成具有不同频率的多个平滑子序列,然后对各个子序列分别进行预测,此种方式能有效解决被预测交通流的时变性、复杂性以及非线性问题,同时自适应遗传算法具有全局搜索能力,能够很好的解决神经网络的缺陷——易陷入局部极小值。并将预测结果与小波神经网络方法以及遗传神经网络方法的预测结果进行了比较。结果表明,本文所建立模型的平均绝对误差、平均绝对百分比误差以及均方根误差均较小,且拟合度EC值较大,由此较好地说明了该模型在短时交通流预测上的有效性、精确性。
[Abstract]:Traffic flow prediction is an important part of intelligent transportation system. The important premise of traffic signal control, route guidance and accident detection is fast and accurate short-time traffic flow prediction. However, the traffic system has the characteristics of high complexity, nonlinearity and uncertainty. It is a complex system composed of people, cars, roads and other objects, making real-time traffic flow. Accurate prediction is one of the hotspots and difficulties in the field of intelligent transportation. However, because of the large amount of traffic flow information, the strong disturbance of uncertain noise signal and the complex topology of urban road network, how to realize the short-term traffic flow prediction of urban road has been hindering the long-term development of intelligent transportation. In order to solve these problems, many forecasting methods have been put forward, but the real-time and accuracy of the prediction results are not ideal because of not considering the influence of uncertain interference signals or the complexity of urban road network on the short-term traffic flow. In this paper, Mallat algorithm is used to decompose and reconstruct the short time traffic flow signal with wavelet transform in order to filter out the strong interference noise signal of the short time traffic flow. This method can improve the speed and precision of the short time traffic flow information preprocessing. In view of the complexity and nonlinear characteristics of traffic flow data, the neural network theory is introduced in this paper. It is an effective method to predict traffic flow in short time by using its good ability to deal with nonlinear problems. In summary, in order to improve the accuracy of short-term traffic flow prediction, in view of the time-varying, complex and nonlinear characteristics of urban road traffic flow, This paper presents a short-term traffic flow combination prediction model based on wavelet denoising and adaptive genetic algorithm to optimize BP neural network. Using wavelet transform, traffic flow can be decomposed into multiple smooth subsequences with different frequencies, and each subsequence can be predicted separately. This method can effectively solve the time-varying, complex and nonlinear problems of the predicted traffic flow. At the same time, the adaptive genetic algorithm has the ability of global searching, and it can solve the defect of neural network, which is easy to fall into the local minimum. The prediction results are compared with those of wavelet neural network method and genetic neural network method. The results show that the average absolute error, mean absolute percentage error and root mean square error of the model are small, and the EC value of the fitting degree is large, which shows the validity of the model in the short-term traffic flow prediction. Accuracy.
【学位授予单位】：华东交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.14

【参考文献】