Abstract: In the operation of the power grid, energy storage systems, as a key means of flexible scheduling in the power system, play an important role in alleviating the pressure on the power grid and enhancing its stability. This paper proposes an intelligent scheduling optimization strategy based on Markov decision process(MDP) and reinforcement learning to address the limitations of traditional power grid energy storage scheduling strategies. Through continuous interaction between intelligent agents and the power grid environment, energy storage systems can learn the optimal charging and discharging strategies in real time, thereby reducing the operating costs of the power grid and effectively addressing issues such as load fluctuations and local blockages.