考虑区域占有率的RTCN路内停车泊位预测模型  

RTCN curb parking space prediction model considering regional occupancy

作　　者：徐建闽[1] 朱琳聪 马莹莹[1] XU Jian-min;ZHU Lin-cong;MA Ying-ying(School of Civil Engineering and Transportation,South China University of Technology,Guangzhou 510640,China)

机构地区：[1]华南理工大学土木与交通学院,广东广州510640

出　　处：《广西大学学报（自然科学版）》2022年第1期112-123,共12页Journal of Guangxi University（Natural Science Edition）

基　　金：国家自然科学基金项目(52072129,61873098);广东省科技计划项目(2016A030305001);中央高校基本科研业务费专项资金(2018KZ17)。

摘　　要：针对汽车停车过程中反复寻泊产生无效交通量这一问题,以向公众提供准确的实时及预测的停车位信息为目标,考虑到当前卷积神经网络(CNN)和递归神经网络(RNN)在空闲泊位短时预测的研究中存在的缺陷,同时为了解决路内停车样本数过少的问题,将RAdam算子引入TCN模型中,其中,TCN模型用于提取空闲泊位的时间特征,RAdam算子用于解决停车样本数过少的不足。为了解决空闲泊位数预测空间特征提取,在分析区域占有率和空闲泊位数相关关系的基础上,提出考虑区域占有率的RTCN短时空闲泊位数预测模型,最后以广东省深圳市南山区的路内停车路段为例进行分析。结果表明:RTCN模型的训练时间远远低于LSTM和GRU模型,并且均方根误差和平均绝对误差也低于LSTM和GRU模型,将区域占有率引入RTCN模型后,预测精度得到进一步提升,本文提出的模型不仅有效降低了预测误差还提高了模型的训练速度。Regarding the problem of ineffective transportation caused by repeated parking searches in the process of car parking,the goal is to provide the public with accurate real-time and predicted parking space information.Considering the shortcomings of the current CNN(convolutional neural network)and RNN(recurrent neural network)in the short-term prediction of free parking spaces,it is proposed to use a TCN(temporal convolutional network)which combines the advantages of CNN and RNN.At the same time,in order to solve the problem that the parking samples of the road parking is limited,RAdam algorithm is introduced into TCN then RTCN(RAdam temporal convolutional network)is obtained.Finally,in order to extract the space feature of the free parking spaces prediction,after analysis of the correlation between the area occupancy rate and the free parking spaces,based on the RTCN and considering the area occupancy rate,a short-term prediction model is proposed and is used to predict the number of free parking spaces.The results show that the training time of the RTCN model,the root mean square error and mean absolute error all are lower than that of the LSTM and GRU.Once the regional occupancy is introduced into the RTCN model,the prediction accuracy is further improved,indicating that the this method significantly improves the accuracy and training speed of the model.

关 键 词：空闲泊位数预测 TCN神经网络 RAdam 区域占有率 

分 类 号：TU473.1[建筑科学—结构工程]