| In recent years,oil prices have continued to rise,coupled with the serious environmental protection problems caused by pollutants emitted by automobiles.Energy conservation and emission reduction have become a new topic in the world.New energy vehicles have ushered in the best opportunity for development.In the future,urban transportation will be dominated by electric vehicles.Whether it is a hybrid car or a pure electric vehicle,it requires external power supply for charging and public charging facilities.However,with the large-scale development of electric vehicles,disorderly charging of electric vehicles will have a negligible impact on the power grid,charging facilities and users.Therefore,the electric vehicle charging problem has important research significance and practical value.Therefore,this paper studies the real-time charging problem of electric vehicles.Study how to prevent the driver from taking too long in the charging path and avoid "partially full and partly idle" charging stations in the real-time running scenario of the car.This article will design a real-time charging decision method for charging vehicles based on Dijkstra,SARSA,Q-learning,and DQN algorithms.Then propose a real-time charging scenario for electric vehicles based on the complex diversity of the urban environment,dividing the city into B*V grids where each grid represents an urban area of the city.Add Global Controller(GC)and RSU to the scene to make the vehicle run and make real-time charging selection more quickly and accurately.The main work of this paper is:(1)Realize real-time vehicle charging selection,optimize the vehicle and charging station from both micro and macro angles,and use V2V(Vehicle to Vehicle)communication to enable vehicles to efficiently and accurately obtain and transmit information.(2)Unlike the existing scheduling work,this paper considers the high-dimensional components of the urban environment in the scheduling process,including different space and time,and the optimal solution obtained is the optimal space-time path.(3)Gradually propose specific decision-making methods based on Dijkstra’s shortest path algorithm,Q-learning algorithm,SARSA algorithm,and DQN algorithm to solve the problems raised by the text and compare relevant factors,and implement scene-based algorithm operations to obtain the Dijksta algorithm.Vehicle travel path,get the Q-learning algorithm,SARSA algorithm training result table(Q-table value),DQN algorithm loss value,and calculate the next node of the vehicle according to the result table and finally determine the vehicle travel path,get the best excellent path.Combined with the simulation on the ONE simulation platform,the results of four algorithms in the same environment are obtained.Three evaluation indicators are designed for comparison,and finally it is concluded that in the real-time charging scenario of the vehicle,reinforcement learning will be greatly optimized in the shortest path algorithm.In the actual environment with extremely high complexity,DQN will also become the most advantageous algorithm. |
