摘要
由于高速公路车流具有时空演变性,故有必要研究其时空特性以建立更为准确合理的车流荷载模型.采用皮尔逊相关理论和奇异谱分析等方法研究了多个地区动态称重系统数据,根据高速公路车流平均车间距的40 m分界点,划分了两类不同的时空模型;定义了车流时间平均车重,并以此分析了车流系统平均车重与车速、车间距等参数的时空相关性及演变特征;进一步采用组合SSA-LSTM模型对车流系统的时间平均车重进行分解重构与预测.结果表明:车流荷载具有24 h周期性特征;组合SSA-LSTM模型相比于单一LSTM模型,其平均绝对误差和均方根误差分别降低了57.1%和54.7%,预测精度有较大提升,能有效预测车流系统的整体荷载变化,从而为评估车流荷载的长期效应提供依据.
Because of the spatiotemporal evolution of expressway traffic flow,it is necessary to study its spatiotemporal characteristics to establish a more accurate and reasonable traffic flow load model.Pearson correlation theory and singular spectrum analysis are used to study the dynamic weighing system data of several regions.According to the 40m dividing point of the average vehicle spacing of the expressway traffic flow,two different space-time models are divided;The time-average vehicle weight of vehicle flow is defined,and the time-space correlation and evolution characteristics between the average vehicle weight of vehicle flow system and parameters such as vehicle speed and vehicle spacing are analyzed.The combined SSA-LSTM model is further used to decompose,reconstruct and predict the time-average vehicle weight of the vehicle flow system.The results show that the traffic flow load has the periodic characteristic of 24 hours.Compared with the single LSTM model,the average absolute error and root mean square error of the combined SSA-LSTM model are reduced by 57.1%and 54.7%respectively,and the prediction accuracy is greatly improved,which can effectively predict the overall load change of the traffic flow system,thus providing a basis for evaluating the long-term effect of the traffic flow load.
作者
赵少杰
陈锟
ZHAO Shaojie;CHEN Kun(School of Civil Engineering,Xiangtan University,Xiangtan 411105,China)
出处
《湘潭大学学报（自然科学版）》
CAS
2024年第4期40-50,共11页
Journal of Xiangtan University(Natural Science Edition)
基金
湖南省教育厅项目(19C1765)
湖南省自然科学基金面上项目(2021JJ30681)。
关键词
高速公路
车流荷载
时空特征
相关性
奇异谱
LSTM
expressway
traffic load
space time characteristics
relevance
singular spectrum
LSTM
