A new optimization and scheduling strategy based on reinforcement learning for smart community energy system (SCES) with electric vehicles (EVs) as energy storage devices is proposed to improve the economic efficiency. Firstly, a new SCES model is established based on the operation mechanism of EVs and other energy devices. The impact of EVs on load and supply-demand balance is taken into account when EVs are connected to the grid. Secondly, EVs are classified into energy consuming groups and energy storing groups, which participate in the system operation as electricity loads and energy storage devices respectively. The energy coupling relations of various energy devices in the new SCES are analyzed, and the nonlinear constraints of EVs and other energy devices are established accordingly. The state space, action space, and reward function of the algorithm are redefined to accommodate the EVs and other energy devices. Thirdly, the proposed novel SCES optimization strategy based on deep double Q-learning network (DDQN) is utilized to solve the energy system optimization problem. Finally, the application of this strategy verifies that the proposed strategy can effectively improve the economic efficiency of SCES.