“Deep reinforcement learning is widely used in the field of user side energy storage regulation, but there are problems such as difficulty in executing constraints and threats to system security. Experts propose an optimization operation strategy based on MILP-TD3, establish a real-time operation optimization model, convert the DRL agent action value function into MILP formula, and enable the agent to strictly execute constraints. The case analysis shows that this strategy can determine the optimal operating strategy, reduce operating costs, meet real-time scheduling requirements, and ensure the safe operation of energy storage systems.”