Research On Relocating Optimization Of Carsharing Vehicles Based On User Travel Demand Prediction

With the improvement of people’s material foundation,the number of private cars is increasing.As a result,traffic jam,environmental pollution,difficult to find parking space and other problems are becoming more and more prominent.As a new means of transportation that can not only improve this kind of problems but also meet users’ various travel demands,especially one-way carsharing are gradually emerging in major cities.However,due to the uncertainty of users’ travel demands and the inconsistency between the picking-up and returning stations of one-way carsharing,a series of problems emerge,such as poor user experience,imbalance between the supply and demand of vehicles at the station,and loss of profits of carsharing operators.To solve these problems,this paper uses a long and short-term memory model(LSTM)to predict the travel demand of potential carsharing users based on operational data characteristics;Based on a similar Logit model,users’ demand for carsharing is determined.The vehicle relocating optimization model of carsharing is constructed by combining manual relocating with dynamic pricing.And a nested genetic algorithm combined with optimization simulation was designed to solve the problem.The research can provide reference for the operation of carsharing enterprises.The specific work carried out is as follows:Firstly,the characteristics of users’ travel are analyzed.Using a carsharing operation in Gansu Province,China,as a research subject.According to the order data provided by the operation,the travel rules of users are analyzed from many aspects.The purpose is to provide characteristic quantity for user travel demand prediction.According to the characteristics of time travel,5-19 o’clock in a day is divided into three periods,namely5-9 o’clock,9-16 o’clock and 16-19 o’clock,and the subsequent studies were based on this time period division.Secondly,the LSTM was designed to predict the travel demand of potential carsharing users at different stations at different times,using the historical picking-up data and characteristics volume of each station as model inputs.Compared with the ARIMA,the results show that the accuracy of LSTM demand prediction is better and the error is smaller.In this paper,the mean absolute percentage error(MAPE)and mean square error(MSE)are used as evaluation indicators for the prediction results.Its mean absolute percentage error is 6.2%,mean square error is 0.0087,which meets the needs of the company.The prediction results of 30 stations with good accuracy are selected as the input data of vehicle relocating optimization scheme.Then,the vehicle relocating optimization scheme under the dynamic user demand is put forward.In order to improve the relocating efficiency and operators’ income,this paper adopts the relocating mode of combining manual vehicle relocating with userinitiated vehicle relocating.Based on the price-sensitive relationship of carsharing demand,dynamic pricing induces users to change the original pick-up and return stations to realize user relocating of vehicles.And a similar Logit model is constructed to determine the demand through the ratio of carsharing utility to private car utility.Second,manually relocate vehicles to complete the remaining relocate tasks.The model aims at maximizing the profit of enterprises.A nested genetic algorithm is designed to solve it.Finally,the effectiveness and scientific validity of the relocating optimization model is verified using an example of a carsharing operation in Gansu Province,China.The objective function value,that is,operator’s income,is solved by the algorithm.Meanwhile,the picking-up price in different regions,the actual travel demand corresponding to the stations,the manual relocating cost in each period and the number of the manual relocating in each period under the scheme can be solved.The analysis shows that the operator’s income has increased by 39.24% compared with that before the implementation of the program.The manual relocating cost of each time period is reduced by 20.40%,5.92% and 17.08% respectively.The total cost of manual relocating is reduced by 15.44%,The change in total station demand for each time period was-1.92%,+2.89% and-1.12%.The results show that the vehicle relocating optimization scheme based on dynamic demand proposed in this paper is reasonable and effective.