Research On Dynamic Allocation Model Of On-street Parking Spaces Based On Short-term Prediction Of Remaining Parking Spaces

In order to solve the problem of unbalanced utilization of parking resources,the rational allocation and management of on-street parking resources is very important.However,the existing parking allocation model does not fully consider the impact of the real-time change of parking demand and supply on parking space allocation,which leads to the unsatisfactory effect of the final implementation.Therefore,dynamic parking allocation is of great value for optimizing the utilization of road parking resources and improving traffic conditions.Firstly,this thesis takes on-street parking lots as the research object,analyzes the influence of weather and working days on the number of remaining parking spaces,and studies the correlation of parking occupancy rates between adjacent and non-adjacent on-street parking lots.Then,it deeply analyzes the parking characteristics of seven typical on-street parking lots,obtains the peak demand period of typical on-street parking lots according to the analysis,and sums up the combination of parking space sharing in off-street parking lots by combining the correlation analysis results of parking occupancy rates among typical on-street parking lots.Secondly,a prediction model of the remaining parking spaces based on the combination of autoregressive moving average model(ARIMA)and convolution long-short time neural network(ConvLSTM)is proposed.According to Pearson correlation coefficient,the data of remaining parking spaces are analyzed,and the characteristic variables with strong correlation are selected as the input parameters of the model,including the number of remaining parking spaces,temperature,wind speed and working day type.The trend of the data is removed by ARIMA model to make it stable,the temporal and spatial characteristics of the data are extracted by ConvLSTM model,and three parking spaces are selected for training and verification in ARIMA-ConvLSTM model and other baseline models.The experimental results show that ARIMA-ConvLSTM model has higher accuracy in predicting the number of remaining parking spaces,which provides a reference for driver selection and platform dynamic allocation.Finally,a dynamic parking allocation method based on parking demand prediction is proposed.A dynamic parking allocation model based on neural network is established,which considers three influencing factors:walking distance,parking cost and parking index.The experimental results show that the dynamic parking space allocation model based on neural network can not only allocate parking spaces in real time,but also improve the utilization rate of parking spaces.The example shows that the performance score of the dynamic parking allocation model keeps above 80%at the time interval of 2-20 minutes,which proves that the model can effectively improve the utilization rate of resources and make the allocation more balanced.In this thesis,aiming at solving the low utilization efficiency of on-street parking resources,combining with the parking space sharing strategy,comprehensively considering the parking management platform and the subjective perceived utility of the parking people,a dynamic parking space resource matching method is put forward to alleviate traffic congestion and resource waste.