Data-driven Power System Collapse Predicting Using Critical Slowing Down Indicators

In this paper, a data-driven voltage collapse predicting method is proposed based on the critical slowing down phenomenon of dynamic systems. First, the dynamic model of power systems with fluctuations is established using the stochastic differential algebraic equations. The system model is used to simulate operating data of power systems, and the proposed voltage collapse predicting method does not rely on a detailed model. Second, the critical slowing down phenomenon of dynamic systems is introduced, and the statistical indicators such as the variance and autocorrelation of state variables are designed. Third, a machine learning method is proposed to predict voltage collapse based on the statistical indicators. Finally, the proposed method is tested using the IEEE 14-bus system with renewable energy generation. The noise of PMU measurement is taken into account and Monte Carlo simulation (MCS) is used to simulate PMU data. The predicting method is used to show the warning signals of voltage collapse in such system.