Coordinated Preventive-Corrective Control for Power System Transient Stability Enhancement Based on Machine Learning-assisted Optimization

Abstract

A machine learning-assisted optimization method is proposed in this paper to coordinate the preventive generation rescheduling and the corrective generatio\load shedding for power system transient stability enhancement. The coordinated control problem is firstly formulated as an extended security constrained optimal power flow (SCOPF) model that includes the corrective generation\load shedding as controlling measures. Considering that the conventional methods are computationally intensive due to the repeated time-domain simulation (TDS)-based transient stability assessment (TSA), the neural process (NP) is proposed to develop the machine learning-based predictor for fast and simulation-free TSA. With the neural process-based predictor to be the surrogate model for fast TSA, the surrogate-assisted evolution programming approach is proposed for online coordinated preventive and corrective control against power system transient instability. The effectiveness of the proposed machine learning-assisted optimization method is demonstrated by the case study on the IEEE 39-bus system and numerical results show that the proposed method can generate the effective coordinated control actions for transient instability mitigation.