Reactive power optimization in active distribution systems with soft open points based on deep reinforcement learning

In the context of the increasing penetration level of photovoltaic energy, its intermittence and randomness bring challenges such as voltage over-limit and increased network loss to the distribution network. The paper proposes a real-time two-time-scale voltage regulation method with soft open point (SOP) to solve the reactive power optimization problem. To obtain hourly scheduling strategies for on-load tap changers and switchable capacitor banks, the centralized optimization problem is formulated as a mixed integer second-order cone programming in the first stage. In the second stage, the active distribution system is divided into multiple subsystems according to the proposed partitioning method, and the SOP is regulated to adjust the voltage in real-time through an effective control strategy and local measurement information of the subsystems. The proposed method takes full account of the randomness of photovoltaic, and can conduct distributed real-time voltage regulation in the active distribution system based on the rapid and continuous response of SOP. Compared with traditional methods, this method has lower communication costs, better real-time performance, and does not dependent on an accurate power flow model. Finally, the effectiveness of the proposed method is verified by the IEEE 33-bus and IEEE 123-bus simulation examples.