In order to solve the issues of standard scheduling techniques’ limited multi-objective optimization ability and lack of flexibility in dynamic contexts, this research suggests an intelligent scheduling model for energy storage systems based on reinforcement learning. Through reinforcement learning technology, the model learns the optimal charging and discharging strategies through the interaction between the agent and the environment, while incorporating energy cost, system efficiency and operational stability into the optimization objectives. The model introduces improvements such as priority experience replay and a dual network structure, which effectively improve training efficiency and decision-making performance. According to experimental findings, the suggested model offers a versatile and effective solution for contemporary energy management by lowering energy expenditures considerably, increasing system efficiency, and preserving operational stability in a changing environment. The research results provide theoretical support and a practical foundation for the future collaborative optimization of multiple energy sources and the construction of smart grids.