Static Voltage Stability Margin Evaluation Considering Stochastic Power Injection Based on Deep Reinforcement Learning Theory

With the high proportion of renewable energy grid-connected and the increase of power electronic equipment, applicability of the traditional voltage stability margin assessment approach in power system decreases, which brings new challenges to the static voltage stability evaluation of power system. A static voltage stability margin evaluation method based on deep Q-learning network (DQN) algorithm is proposed in this paper. To begin with, the boundary conditions of static voltage stability margin are identified by the singularity of Jacobian matrix. The voltage stability margin is measured by the nearest distance between the system operation point and the boundary point, which establishes the evaluation index of static voltage stability margin. In additional, DQN algorithm is introduced to explore the nearest stability boundary point swiftly in the high dimensional load action space. Then according to the decision trajectory information of the load action, the stability margin is evaluated quickly and effectively. Finally, the voltage stability margin under different load states is evaluated in IEEE 39-node system, and the effectiveness of the proposed method is verified by comparison with existing methods.