Optimal Model Of Locating Multi-Layer Charging Stations Based On DBSCAN Algorithm

Popularization of electric vehicles(EVs)is a promising approach towards realizing environmentally friendly transportation.While the traditional oil-fueled vehicle consumes a lot of energy,and automobile exhaust is the main source of air pollution,electric vehicles have the advantages of energy saving and emission reduction,which can be an important means to solve environmental problems.However,due to the current battery capacity limitations of electric vehicles,users’ "mileage anxiety" and inconvenient charging procedure,user experience has become a vital factor limiting the acceptance of electric vehicles.Nonetheless,most existing charging station siting methods are concentrate on grid distribution,road distance,charging station construction cost,etc.The potential charging behavior patterns and requirements of electric vehicle users are usually neglected,as a result,the charging infrastructure network does not serve the electric vehicle users very well.This paper focuses on the construction and the scale recommendation of electric vehicle infrastructure.Through the analysis of the spatial and temporal distribution of vehicle trajectory,as well as the mining and clustering of user’ parking points,the implied user travel behavior is obtained,which may find a new way to solve the problem of charging station deployment.The main research contributions of this paper are as follows:(1)Complete the pre-processing of the taxi GPS trajectory,the calculation of the stay point and the acquisition of the POI data for fusion.Firstly,according to the goal of this paper,a set of data cleaning rules is customized.Next,grouped by the vehicle ID,the cleaned trajectory data is divided into a track set and a sub-track set.Then,according to the definition of the stay point,the track set is calculated to obtain a series of stay points including the stay duration and the latitude and longitude information.Finally,fusion the POI data set and the stay point data set using a specific rule.(2)Research on a charging station demand estimation model based on parking point set and interest point distribution.In this paper,the target area is meshed.According to the length of stay,the stay points are divided into two types: short-term stay point and long-term stay point.Combined with the charging demand caused by POI data,a linear charging demand estimation model and a scale recommendation method are established.By the way,the spatial distribution of charging demand under existing traffic data can be identified.Rather than the complete trajectory set,the model focuses on the dwell events in the trajectory set,which avoiding processing the complex vehicle travel paths,reducing data redundancy and model complexity.(3)A multy-layer DBSCAN clustering method and R-constrained optimization process are proposed.Through the density analysis of the trajectory data,the improved DBSCAN algorithm divides the data area into three layers,clusters the three layers sequentially using different parameters.The center point of cluster can cover the most charging demand and is used as a candidate location for charging station locating.The charging service radius R is used to constrain the distance between charging stations in the area,and the clustering result is optimized to make it conform to the real world rationality.(4)Applying the model to the vehicle trajectory data of the main city of Chongqing,Based on the demand estimation model,this paper calculates the heat value matrix of the main urban area of Chongqing,and uses the improved multi-layer DBSCAN algorithm and R-constrained optimization to obtain the charging scheme for the charging station in the area.Finally,according to the difference of charging demand in each area,the power supply scale of each charging station is given.A large number of experimental analyses verify the availability of the model.