基于BP神经网络的高速公路车流量预测研究

发布时间：2019-04-27 11:53

【摘要】：随着我国改革开放不断深入，人们客运和货运需求不断上升，对建成的高速公路的通行能力提出了更高的要求，考虑环境和成本费用等问题，盲目的扩建高速公路是不可取的，高速公路的建设应该从一味的量增转变到合理规划、有效益的增长上来，这样可以减少不必要的投资。因此就要求企业对原有建成的高速公路车流量进行准确的预测，把预测结果作为交通规划决策的依据和企业未来收益预测的依据。高速公路车流量的预测属于一种长期车流量预测，而且容易受社会环境各方面的影响，，为了提升预测的准确性，必须选择一种对环境适应性更强的预测模型。神经网络模型不仅具有实行大规模的并行处理的优点，可以在同时分析大量相关因素的情况下保证系统能以更快的速度输出可靠结果，还具有非线性映射特性，这就大大增强了神经网络模型适应环境的能力。因此，运用神经网络模型可以对高速公路车流量进行比较准确的预测。本文对现阶段高速公路车流量预测方法进行了系统的梳理，总结不同预测模型存在的优缺点，构建了基于BP神经网络高速公路车流量预测模型，结合高速公路车流量数据的特点，对高速公路车流量样本数据预处理方法和BP神经网络预测模型的激励函数进行了改进，确定了预测模型中各参数的初始值的方法，同时提出了新建或改扩建高速公路对预测项目影响的定量化方法，从而结合影响程度定量化的结果对神经网络模型的预测值进行改进，提高了预测结果的精度。本文研究的主要结论有：第一，相比于其它预测模型，神经网络拥有更多的优势，它可以融合定性和定量两类数据，并且拥有很好的容错性和鲁棒性，能对非线性函数有很强的映射能力，最后保证系统的大规模并行处理能力，提高输出结果的速度和准确性。第二，BP网络模型的结构设计和各参数选取尽量避免模型自身的缺陷，并结合所要研究的预测项目特点进行细致的分析。第三，路网中如果有新建或者改建高速公路，就会改变原来的路网结构，对原有的高速公路就会产生很大的影响，转移一部分原来公路上的车流量，这会使预测模型的预测结果出现偏差，因此为了增加预测结果的准确性，必须对新建或改建高速公路的影响程度定量化进行相应的研究。
[Abstract]:With the deepening of China's reform and opening-up and the increasing demand for passenger and freight transport, it is not advisable to expand the expressway blindly, considering the problems of environment and cost, and putting forward a higher demand for the capacity of the built highway. Highway construction should be transformed from volume increase to rational planning and effective growth, so that unnecessary investment can be reduced. Therefore, the enterprise is required to accurately predict the traffic volume of the original highway, and take the forecast result as the basis of traffic planning and decision-making and the basis of the future profit forecast of the enterprise. The prediction of expressway traffic flow is a kind of long-term traffic flow prediction, and it is easy to be affected by various aspects of social environment. In order to improve the accuracy of prediction, it is necessary to choose a forecasting model which is more adaptable to the environment. The neural network model not only has the advantages of large-scale parallel processing, but also can ensure that the system can output reliable results faster, and it also has the characteristics of nonlinear mapping under the condition of analyzing a large number of related factors at the same time. This greatly enhances the ability of neural network model to adapt to the environment. Therefore, the neural network model can be used to predict the traffic volume of expressway accurately. This paper systematically combs the forecasting methods of expressway traffic flow at present, summarizes the advantages and disadvantages of different forecasting models, and constructs the forecasting model of expressway traffic flow based on BP neural network. Combined with the characteristics of expressway traffic data, the pretreatment method of sample data of expressway traffic flow and the excitation function of BP neural network prediction model are improved, and the initial values of each parameter in the prediction model are determined. At the same time, the method of quantifying the influence of newly built or expanded expressway on the forecast project is put forward, which improves the prediction value of the neural network model combined with the quantitative result of the influence degree, and improves the precision of the prediction result. The main conclusions of this paper are as follows: first, compared with other prediction models, neural network has more advantages, it can integrate qualitative and quantitative data, and has good fault tolerance and robustness. It can map the nonlinear function strongly. Finally, it can guarantee the large-scale parallel processing ability of the system, and improve the speed and accuracy of the output results. Secondly, the structure design of BP network model and the selection of each parameter avoid the defects of the model itself as far as possible, and the characteristics of the prediction items to be studied are analyzed in detail. Third, if there are new or rebuilt highways in the road network, it will change the structure of the original road network, which will have a great impact on the original highway and transfer a part of the traffic flow on the original highway. Therefore, in order to increase the accuracy of the prediction results, it is necessary to make a quantitative study on the influence degree of the newly built or rebuilt highways in order to increase the accuracy of the prediction results.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.14

【参考文献】