Linearized MILP Model With Improved Soft Actor-Critic Algorithm for Dynamic and Efficient Active Distribution Network Planning

With the rapid development of distributed generations (DGs), the operation scale of distributed networks is getting larger and larger, and the operation modes are becoming more diversified. The traditional AC-based nonlinear planning model has many shortcomings in computational analysis. To this end, an active distributed network planning model based on linear constraints is proposed. Firstly, the basic principle and analysis method of the distributed network model linearization are proposed from two aspects, namely, power flow linearization and distributed network connectivity linearization. Then, by analyzing the distributed network planning objective function and constraints, a mixed integer linear planning (MILP) model for distributed networks is proposed. And an improved Soft Actor-Critic (ISAC) algorithm is proposed to solve the model. Finally, the feasibility and efficiency of the proposed method are verified by comparing the proposed MILP model and the traditional nonlinear planning model for distributed networks through multi-scenario case analysis. The experimental results demonstrate that the proposed MILP model significantly improves computational efficiency while maintaining accuracy. Compared to the traditional AC-based nonlinear model, the MILP model achieves a maximum relative error of only 3.3%, with computation time reduced from over 460 seconds to less than 0.2 seconds. Furthermore, ISAC exhibits outstanding performance in dynamic scenarios, reducing total costs by 20.63% and 15.25% compared to Deep Q-Network (DQN) and Deep Deterministic Policy Gradient (DDPG), respectively.