高速公路基本路段实时交通状态判别方法的研究及应用

发布时间：2018-01-13 13:30

本文关键词：高速公路基本路段实时交通状态判别方法的研究及应用　出处：《长安大学》2015年硕士论文　论文类型：学位论文

【摘要】：目前高速公路交通状态的判别标准大多以固定阈值比较或绝对标准为主,忽略了不同时空背景下道路环境、天气状况等客观因素的影响。为此,本文以高速公路基本路段为研究对象,依据交通流具有时间序列相似性,利用海量的交通流表征参数运行数据,构建出交通状态划分的相对标准及实时交通状态判别决策模型。对高速公路交通流的控制与管理,具有重要的研究意义与应用价值。本文首先研究了交通流运行和拥挤特性,提取交通状态表征参数,同时针对交通状态判别采用绝对标准的缺陷,提出了基于模糊聚类的高速公路基本路段实时交通状态划分的相对标准的思路。其次,基于模糊C均值聚类算法对历史交通流运行数据进行聚类分析,将获取的交通状态聚类中心作为划分的相对标准,再根据欧式距离判别实测数据的交通状态。针对模糊C均值聚类判别算法中初始聚类中心选取具有随机性使得算法不稳定、易陷入局部最优的问题,引入遗传算法对初始聚类中心选取进行优化,增强交通状态聚类分析的可靠性。同时,鉴于欧式距离决策模型计算时间复杂度较大,构建了基于核极限学习机的实时交通状态判别决策模型。从分类性能和计算时间复杂度两个角度,将该方法与支持向量机模型进行对比分析。最后,以PeMS实测的交通流表征参数运行数据为基础,对本文所构建的高速公路基本路段实时交通状态判别方法进行了仿真分析。结果表明,基于遗传算法的模糊C均值聚类算法稳定性好,收敛速度快;基于核极限学习机的实时交通状态决策模型在保证分类精度的基础上,大大节省了时间成本,具有较好的实时性。
[Abstract]:At present, the criterion of freeway traffic state is mostly fixed threshold comparison or absolute standard, ignoring the influence of road environment, weather condition and other objective factors in different time and space background. This paper takes the basic section of highway as the research object, according to the traffic flow has the time series similarity, uses massive traffic flow to represent the parameter running data. The relative standard of traffic state division and the decision model of real-time traffic state discrimination are constructed. The control and management of expressway traffic flow are also given. It has important research significance and application value. Firstly, this paper studies the traffic flow and congestion characteristics, extracts the parameters of traffic state representation, and at the same time uses the absolute standard to judge the traffic state. This paper puts forward the idea of relative standard of real-time traffic state partition of the basic section of highway based on fuzzy clustering. Secondly, the paper analyzes the running data of historical traffic flow based on fuzzy C-means clustering algorithm. The traffic state clustering center is regarded as the relative criterion of the partition. Then according to the Euclidean distance to judge the traffic state of the measured data, the random selection of the initial clustering center in the fuzzy C-means clustering discriminant algorithm makes the algorithm unstable and easy to fall into the local optimal problem. Genetic algorithm is introduced to optimize the selection of initial clustering centers to enhance the reliability of traffic state clustering analysis. At the same time, the computational time complexity of Euclidean distance decision model is large. A real-time traffic state discriminant decision model based on kernel extreme learning machine (KLMs) is constructed. From the perspective of classification performance and computational time complexity, the method is compared with the SVM model. Finally, the proposed method is compared with the SVM model. Based on the operation data of the traffic flow parameters measured by PeMS, the real-time traffic state discrimination method of the basic section of expressway is simulated and analyzed. The results show that the method is effective. The fuzzy C-means clustering algorithm based on genetic algorithm has good stability and fast convergence speed. The real-time traffic state decision model based on kernel limit learning machine can greatly save the time cost and have better real-time performance on the basis of ensuring the classification accuracy.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【参考文献】