This paper presents a vehicle platoon control method based on reinforcement learning(RL) to solve the multi-objective optimization problem. The actor network is designed to receive the state information of each vehicle in the platoon and the inter-vehicle state error, and outputs the desired acceleration based on the longitudinal dynamics of the vehicle. The proposed approach ensures both the individual vehicle stability and the string stability of the platoon under V2X communication. To model the platoon driving scenario with the spacing policy and communication topology, the Markov decision process(MDP) model of the platoon is established. In addition, considering the multi-input and multi-output high-level sample characteristics of the platoon, the deep deterministic policy gradient(DDPG) algorithm is adopted with the priority experience replay strategy to improve the convergence efficiency. To better approximate the actual platoon vehicle fuel consumption, the simulation is based on PreScan to build a high-degree fuel vehicle dynamics model. A co-simulation environment is created using Matlab/Simulink to train the actor network and critic network in the platoon controller by adding noise. The simulation results demonstrate that the reinforcement learning-based vehicle platoon control approach reduces fuel consumption and achieves faster and smoother vehicle control.