Voltage Stability Monitoring based on Adaptive Dynamic Mode Decomposition

This paper develops a new voltage stability monitoring method using dynamic mode decomposition (DMD) and its adaptive variance. First, state estimation (SE) is used to estimate the voltage in the system. Then, the measured voltages from the phasor measurement units (PMU) and estimations from SE are used as the inputs for DMD to predict the long-term voltage dynamic. Furthermore, to improve the prediction performance, the normal DMD is improved by adaptively changing the size of input samples based on the error in the training phase, named adaptive DMD (ADMD). The effectiveness of the proposed method is validated on the Nordic32 test system, which is recommended as the test system for voltage stability studies. Different contingency scenarios are used, and the results show that the proposed method is able to monitor the voltage stability after a disturbance (i.e., $4.3\times 10^{-4}$ MAPE for a stable case and 0.0041 MAPE for an unstable case). Furthermore, the results from a scenario in which a real-world oscillation event is used show high accuracy in voltage stability monitoring of the proposed ADMD method.