Data-Driven Oscillation Mode Prediction Based Resonance Mitigation Scheme for Wind Turbine Generators Connected Power System

This paper proposes a data-driven oscillation mode prediction (DOMP) model, aimed at mitigating potential modal resonances in wind turbine generators (WTGs) connected power systems, without the need for establishing complex state-space analytical models. The proposed approach firstly involves the identification and clustering of system oscillation modes under historical operating scenarios, providing high-quality training data. It then employs a deep extreme learning machine algorithm to train the DOMP model, with scenario information as inputs and identified oscillation modes as outputs. Based on the DOMP predicted oscillation modes, WTG control parameters are optimized to prevent potential modal resonances and improve system small-signal stability under uncertain future scenarios. The effectiveness and scalability of the proposed DOMP scheme is validated through simulation results using the modified IEEE 39-bus and IEEE 118-bus model. The results demonstrate the scheme's ability to accurately predict system oscillation modes based on operating scenario information and effectively prevent potential modal resonances.