Research On EV Charging Guidance And Calibration Based On Improved Deep Reinforcement Learrning

In the context of the development of new energy era,electric vehicles(EVs)have made significant contributions in energy conservation and emission reduction.However,in the face of this emerging industry,whether electric vehicles,charging stations,and distribution networks can be "interconnected" is the key to improving user convenience and improving social charging networks.Therefore,it is necessary to establish a charging guidance strategy that takes into account the interests of users,charging operators,and distribution networks.In response to the shortcomings of poor economic profits in existing electric vehicle charging guidance strategies,this article integrates the coordinated pricing problem of charging stations into the charging guidance strategy of electric vehicles,and utilizes improved deep reinforcement learning for solving training,thereby guiding the reasonable distribution of charging loads in time and space.The specific research content is as follows:Firstly,in response to the complexity of the vehicle station network interaction network,the travel and charging characteristics of electric vehicles,as well as the queuing behavior of charging stations,were modeled,and a basic disordered charging process was established to compare with subsequent orderly charging.At the same time,a theoretical basis was provided for real-time pricing of charging stations and orderly charging guidance of electric vehicles.Secondly,a coordinated pricing model for charging stations based on finite Markov decision is proposed to address issues such as the inability of existing pricing strategies for charging stations to guide users through electricity prices,resulting in imbalanced utilization of charging station equipment.At the same time,in response to the problem of the above model being unable to cope with complex random disturbances in actual working conditions,it is transformed into an FMDP process and a solution strategy based on improved DDPG is proposed to train and solve the model.Finally,the effectiveness of the proposed charging station pricing model and solution strategy in reducing user charging costs and improving charging station revenue was verified through numerical examples,providing a pricing basis for subsequent charging guidance.Then,the obtained electricity price,interactive information during driving and charging are used as model inputs.In response to the problem of traditional guidance methods being difficult to provide real-time charging decision-making solutions for car owners,a decision model based on double-layer FMDP is proposed to transform the charging guidance model of electric vehicles.Based on this decision model,the problem of insufficient generalization performance,computational performance,and convergence performance in the basic solution method is addressed,An improved Rainbow algorithm was proposed as the solution for the two-layer FMDP decision model.Finally,an actual traffic network example was used to verify that the proposed method can provide realtime charging and driving decision-making solutions for car owners,and further reduce the travel costs of electric vehicle owners.Finally,to further demonstrate the practical application effectiveness of the proposed pricing strategy and charging guidance strategy,the application results were analyzed through a real-time coupling scenario of vehicle station network under the same example configuration.The proposed strategy has been verified to not only achieve peak shaving and valley filling for the distribution network,but also improve the utilization efficiency of the charging network,and reduce the charging cost for EV users while meeting real-time traffic path planning requirements.