LSTM-based Sub-Synchronous Oscillation Detection Scheme for Type 4 Wind Farm Interfaced with Weak AC Grid

Sub-synchronous oscillations (SSO) pose a significant threat to the stability of a power grid integrated with wind energy resources. Specifically in Type 4 wind farm's operation under weak grid conditions, SSO may be triggered because of the interaction of the fast dynamics of converter's control with the AC grid. The adverse effects of this type of SSO include sustained oscillations in the active power, point of common coupling (POC) voltage, and output current which can pose severe threat to the stability of the system and may lead to cascaded tripping across the network. Given these risks, it is crucial to ensure the effective and timely detection of SSO to mitigate potential instability and prevent potential cascading failures. Conventionally, SSO is identified by performing frequency scanning, impedance scanning approaches, developing and analyzing mathematical models (small signal and eigenvalue analysis). This paper presents a lightweight intelligent machine learning-based approach for swift detecting SSO in wind farms equipped with Type 4 WTG units. The proposed SSO detection scheme leverages a double layer LSTM (DL-LSTM) based neural network to devise swift, effective, and accurate SSO detection, addressing the shortcomings of previous approaches. Insightful case studies are presented in this paper to validate the effectiveness of the proposed ML based SSO detection scheme.