The randomness of user travel behavior and electricity prices poses a severe challenge to the charging and discharging optimization of electric vehicles under the guidance of real-time electricity prices. To meet this challenge

this paper proposes a real-time charging and discharging strategy for the electric vehicle based on an improved soft actor-critic. First

considering the characteristics of the efficiency changing with power during the charging and discharging process of the electric vehicle and the power boundary constraints

a nonlinear charging and discharging efficiency model and a power boundary model are established respectively to more accurately model the SOC update process. Secondly

a user anxiety model is constructed to characterize their range and time requirements to alleviate the irrational charging behavior caused by anxiety and the decision-making bias caused by it. The user travel uncertainty is characterized by the Markov chain

and their charging location and time behavior are simulated. Then

the electric vehicle charging and discharging optimization problem is modeled as a Markov decision process with unknown transition probability

and then an improved soft actor-critic integrating a triple critic network and a comprehensive experience playback mechanism is proposed to formulate the strategy for electric vehicle charging and discharging in real time. The simulation results of the example show that the proposed optimization method based on the improved soft actor-critic can meet user needs and reduce charging costs under the premise of ensuring reliable charging and discharging power

and has good adaptability to uncertainty.