Hybrid Double Deep Q Network for Active Distribution Network Equivalent Modeling

As more and more distributed renewable energy generation is connected to the active distribution network (ADN), the power interaction between the ADN and the backbone network becomes more complex. In addition, due to the lack of numerous branches of power measurement units, it is difficult to accurately simulate the power flow in ADN. Therefore, we proposes an equivalent model of ADN, consisting of a motor and a synthetic ZIP load. Equivalent modeling utilizes a Hybrid double deep Q-network (HDDQN) to track the dynamics of the original ADN, using a small amount of power measurements in the boundary to achieve high accuracy. In HDDQN, select the types of ZIP loads and motors from the load pool to obtain the dynamic equivalent modeling of the first layer ADN. Through the powerful feature extraction ability of deep learning and the optimization decision-making ability of reinforcement learning, determine the weight of the model, and obtain the $PQ$ components of the motor and ZIP loads in the second layer. The results indicate that HDDQN can accurately equivalent ADN compare with DQN.