Research On Charging Demand Prediction And Recommendation Method For Electric Taxis Driven By Trajectory Data

With the progress of human society and the development of urban modernization,many cities begin to replace traditional fuel taxis with electric taxis.Although electric taxis have the advantages of energy conservation and environmental protection,they also bring difficulties to taxi drivers and government departments.On the one hand,electric taxis themselves have the characteristics of long charging time,high charging frequency,queuing at peak hours,and so on.On the other hand,due to the high cost and scarce land resources,it is difficult to build charging stations.Even if sufficient charging stations are deployed,uneven spatial distribution will still lead to low resource utilization.Therefore,in practical application and real life,the charging recommendation scheme is more operable and challenging than the direct deployment of charging stations.This paper studies and implements a set of charging recommendation methods for large-scale electric taxis based on charging demand prediction.Firstly,based on the historical trajectory data,we implemented three prediction models using cyclic neural network and attention mechanism to predict the charging demand of electric taxis:Where is the driver’s next drop-off point?Will the driver choose to charge at this drop-off point?If you choose to charge,which charging station will you choose?Then,considering the charging demands of current and potential users,we model the charging recommendation method as a Pareto optimization problem,which minimizes the idle time of users(traveling time+queuing time)under the practical constraints.Finally,a prototype system based on the proposed method is implemented and a simulation experiment is carried out using a two-week data set of the actual electric taxi in Shenzhen.The experimental results show that the recommendation method proposed in this paper can effectively reduce the queuing time and idle time of electric taxis.