Segmented Online Identification of Broadband Oscillation Impedance Based on ASSA

This paper addresses the challenges of broadband impedance identification in wind farms connected to the power grid, where broadband oscillations can compromise grid stability. Traditional impedance modeling approaches, including white-box and black/grey-box methods, face limitations in real-world applications, particularly when dealing with commercial new energy units with unknown control structures. To overcome these challenges, a novel real-time impedance identification method is proposed for PMSGs(Permanent Magnet Synchronous Generators). The method, called ASSA (Attention-based Shared and Specific Architecture), utilizes a multi-task neural network model combined with an attention mechanism to improve the accuracy of impedance fitting across different frequency bands. A broadband impedance dataset is constructed offline under various operating conditions, incorporating uncertainties like wind speed. The proposed approach offers an efficient solution for impedance identification, enhancing the stability and reliability of grid-connected renewable energy systems.