With the continuous expansion of domestic metro operation scales, the issue of excessive power consumption by metro trains has become increasingly prominent. By optimizing the speed profiles of train control systems, mechanical energy losses incurred during train operation—due to overcoming resistance and the braking process—can be reduced, thereby improving traction energy efficiency. The rapid development of AI algorithms and computing equipment has brought new opportunities for the innovation of speed profile optimization algorithms. Based on this, a speed profile optimization method using the Deep Q-Network (DQN) algorithm is proposed: First, an optimization framework for speed profile optimization is constructed. Second, a reinforcement learning architecture is applied to iteratively optimize the intelligent agent. Finally, the DQN algorithm is used to solve for a speed profile with energy-saving characteristics. A simulation environment is established on the MATLAB platform, and a comparative simulation of the energy-saving effects between the traditional algorithm and the proposed algorithm is conducted. The results show that the proposed algorithm achieves approximately 5%–7% energy savings compared to the traditional algorithm.