To address the high operating costs of charging stations due to the uncertainty of charging times for new energy vehicles and the randomness of photovoltaic（PV）output, as well as to tackle the challenge of excessive action variables in large-scale electric vehicle （EV）charging processes, a two-layer sequential optimization scheduling model for a PV-hydrogen-storage charging stationis is proposed based on a deep reinforcement learning（DRL）algorithm. This model considers factors such as PV output, time-of-use pricing, load uncertainty, and the operational efficiency of each piece of equipment in the system, aiming to meet user demands while reducing the operating costs of the charging station. The twin delayed deep deterministic policy gradient（TD3）algorithm is employed to solve the two-layer sequential scheduling model. The simulation results show that the model can greatly reduce the operating cost of charging station under the premise of meeting the charging demand of users. In addition, when the number of charging piles increases, the real-time scheduling time of the model is not affected and solar curtailmentcan be effectively reduced.