Abstract：To realize the real-time charging guidance of electric vehicles and improve the charging efficiency of

charging stations，a real-time charging guidance strategy for electric vehicles based on hierarchical deep

reinforcement learning was proposed. Considering the mutual coupling characteristics of vehicle-station-road

multiple agents，a double-layer electric vehicle charging navigation model was constructed based on the

characteristic information of electric vehicles，charging stations，distribution networks and transportation networks.

The above-mentioned model was decoupled into a two-layer finite Markov decision process network architecture，

the upper network evaluated and recommended charging stations，and the optimal selection result were passed to the lower network. The lower network planed the driving path for the user. The deep Q-network algorithm based on

rainbow framework was used to solve the above-mentioned two-layer decision-making process. Finally，the

simulation results in a specific urban area show that compared with the disorderly guidance method，the proposed

method can reduce the user time cost and save the user cost，and ensure the safe operation of the distribution network.

      Key words：electric vehicle（EV）；real-time charging guidance；recommending charging station；planning

driving path；two-layer deep reinforcement learning（DRL）；deep Q-network algorithm