Research On Operation Optimization Of Urban Carsharing System Based On Spatio-temporal Big Data

Urban car sharing is a new self-service car rental service mode based on the Internet and mobile device applications.It can meet the user’s demand for pick-ups/drop-offs.At the same time,it can also curb the problems of traffic congestion and tight parking spaces caused by the rapid increase in the number of urban motor vehicles.However,the urban car sharing system faces many urgent problems in operation and management.Its operational flexibility comes at the cost of system vehicle imbalance during operation.In particular,it is the matching of the distribution of vehicles with demand in phases,and the matching of the distribution of vehicles with the demand of the next phase after the end of an operating cycle.Operators lack the grasp and evaluation of system operation status.Dispatchers can only be dispatched in a certain period for manual deployment of unbalanced vehicles.This not only affects the operating efficiency of the system,but also increases operating costs and reduces service quality.To solve the above problems,this paper conducts research from four levels of user’ travel features,operation management,depot location and relocation of the one-way station-based car sharing system.The main work carried out in this dissertation includes:(1)First,the data visualization method is used to analyze and show the users’ features of car sharing.Then the POI data is introduced,and the urban functional area extraction framework is constructed using statistical theories.Based on the demands of individual users,a spatial structure of features of urban functional areas with time granularity is constructed.The spatial-temporal clustering method reveals the spatial-temporal dynamics of users’ mobility based on the urban functional structure.Finally,based on the multi-source data set,the Akaike information criterion was used to extract the spatial-temporal evolution of the demands of car sharing system among urban functional areas,which provided a direction for operators to optimize depots location.(2)To detect the spontaneous deviation problem derived from the user’s self-service demand involved in the system operation,so that the operator can monitor the operation status at any time according to the user’s demands,the urban car sharing system operation management evaluation method based on the LSTM deep learning structure is proposed.A discrete event model was established to capture the operating features of car sharing services from a time perspective,and at the same time quantitatively analyzed the user’s vehicle usage behavior.The discrete event model is used to evaluate the short-term prediction results of LSTM based on spatio-temporal big data.The results show that the evaluation method combined with LSTM prediction is feasible in describing the operating state of the system and can be used as a benchmark for further development of adaptive dynamic scheduling strategies.Aiming at the problem of urban car sharing system operation and management,in this dissertation,we propose a new approach based on deep learning techniques to assess the operation of a station-based car sharing system.Firstly,we analyse the pick-up and drop-off operations of the station-based car sharing system,capturing the operational features of car sharing service and the behaviours of vehicle use from a temporal perspective.Then,we introduced an analytical system to detect the system operation concerning the spontaneous deviations derived from user demands from service provisions.We employed Long ShortTerm Memory(LSTM)structure to forecast short-term future vehicle uses.An experimental case based on real-world data is reported to demonstrate the effectiveness of this approach.The results prove that the proposed structure generates high-quality predictions and can monitor the operation status at any time according to user demands.It can be used as a benchmark for further development of adaptive dynamic adjustment strategies.(3)To realize the adaptive scheduling strategy based on spatio-temporal big data,firstly,the operation locations are regionalized.Then,based on the initial depot distribution set by the operator,a MIQCP model for depot location optimization was constructed.Based on the optimal solution for depot location,a MILP model for single-day VRe P optimization is constructed.The current optimal theoretical adaptive scheduling strategy covering real demand and forecast demand is designed.The dissertation carried out a simulation experiment based on space-time big data.The results show that while ensuring the effectiveness and timeliness of the optimization method.This method can be used for operators to perform actual operations.In summary,this dissertation has completed a integrated set of urban car sharing system operation optimization methods that obtain operational problems from the market and solve operational problems with scientific methods in the scientific research field.It can provide scientific and effective theoretical support for urban car sharing operators in improving service quality,improving operating efficiency and reducing operating costs.