Early Warning of System Oscillation Events based on Precursory Signs

(SSO) events since 2021, presenting a significant threat to system security. SSOs can compromise power quality, hinder renewable integration, and disrupt critical loads. Presently, the underlying mechanisms of SSO events, especially those influenced by inverter based resources (IBRs) are still not well understood. Furthermore, modelling complex IBR dynamics and their interaction with the wider power system remains challenging due to proprietary and variable IBR control design. As an alternative, real-time measurement-based methods offer a promising solution by avoiding complex modelling. This paper presents such an approach, which uses precursory signs identified from analysing real-world SSO events to provide early warning of SSO risks. In the analysis, the statistical change-point detection (CPD) technique was employed to automatically process large volumes of historic data and detect SSO signatures. An early warning system that employs a machine learning based method to automatically detect SSO events and their precursory signatures in synchronised measurements is proposed and demonstrated. The early warning system is validated using actual SSO event measurements from phasor measurement units (PMUs) to detect oscillatory instability.

Discussions of the potential limitations of using fundamental phasors in characterising IBRdriven SSOs are provided, where it is compared with analysis from waveform recordings. It is recommended that synchronised waveforms can better detect and accurately measure multiple oscillation modes associated IBR-driven oscillations.