| With the increasing problem of environmental pollution,more and more people are paying attention to green and environmentally friendly electric vehicles.The strategic placement of public charging stations is critical in advancing the growth of the electric vehicle industry as they play a vital role in providing energy to power these vehicles.Starting from user needs,establishing an optimized charging station network is beneficial for reducing distance anxiety among electric vehicle users.Reasonable site selection planning for Electric Vehicle Charging Station(EVCS)is also beneficial for improving the service quality and operational efficiency of existing charging facilities.Most of the existing research on EVCS location planning methods often focus on basic factors such as population density and traffic flow during charging demand analysis,but do not fully consider the impact of dynamic vehicle operating data,urban population activities,city functional areas,points of interest(POI),and other factors on location planning.In addition,these planning methods also neglect factors such as road conditions and electricity costs,and there are problems such as regional overlap or distances that are too far for the selection of charging stations.These all lead to EVCS planning that cannot meet the charging needs of electric vehicle users,and it is not conducive to providing users with better quality services.To address these issues,this paper proposes an EVCS location planning method based on real-time and spatial data of electric vehicles and geographical semantic information.By analyzing the spatiotemporal data of vehicles and mining and clustering surrounding points of interest,potential travel demands of users are obtained.Combining with the coverage range and distribution characteristics of EVCS facilities,considering the service range of candidate EVCS and user charging needs,an optimal location planning model is constructed to seek new ways to solve the problem of EVCS deployment.The specific research contents of this paper include:(1)Estimation of new charging stations based on unsupervised learning.Due to the variable and uneven distribution of points of interest(POI)in cities,which are affected by various factors such as geography,terrain,and urban development,unsupervised learning is used to discover multidimensional connections among the elements in POI distribution data.Using unsupervised clustering algorithms,this study combines spatiotemporal data with points of interest related to activities as the basis for cluster analysis,and proposes a model to describe the estimation of charging demand.The visualized results demonstrate the strong representation ability of the clustering algorithm,and the rationality of selecting the scale of charging stations is confirmed by experimental validation.Overall,this method is based on unsupervised learning and can discover multidimensional connections among elements in POI distribution data.By utilizing clustering algorithms,spatiotemporal data,and points of interest related to activities,this study proposes a charging demand estimation model with a strong representation ability,and demonstrates the rationality of selecting the scale of charging stations through experimental validation;(2)Construction of EVCS Location Model and Location Solution Based on Dual Archive Evolutionary Algorithm.First,unsupervised clustering analysis is used to estimate the potential charging demand based on spatiotemporal data and POI information.Second,a multi-objective optimization model is constructed using the estimated demand and size of the charging station as inputs,with constraints including the number of POIs covered,distance between charging stations,and electric vehicle access distance.The model also considers the uncovered POIs and avoids overlap with existing and other new charging station coverage areas.The bi-archive evolutionary algorithm is then used to evaluate and update population information based on candidate site service areas and electric vehicle user demand.Finally,a distance priority principle is used to design an electric vehicle selection function to validate the effectiveness of the model;(3)A charging station location model based on reinforcement learning algorithm.A reward function was constructed based on comprehensive service coverage,charging efficiency,economic cost,and environmental friendliness factors.The action space was whether to build a station,and the state space was the number of charging stations,station spacing,coverage of interest points,and cost investment.A reinforcement learning model was constructed to optimize the location problem of electric vehicle charging stations.Through experiments,it was found that the performance of this algorithm is almost equal to that of the dual archive evolutionary algorithm when considering the number of POI coverage and the number of charging station selections.However,there is a significant improvement in the average selection distance of electric vehicles.This indirectly indicates that the two EVCS location optimization models proposed in this paper,which consider coverage range and user needs,each have different optimization capabilities and are reasonable.Users can choose optimization solutions based on their own needs. |
