Impedance Profile Prediction for Grid-Connected VSCs based on Feature Extraction

Impedance-based stability analysis has been widely adopted for voltage source converters (VSCs). Considering unknown controller parameters, impedance measurement based on frequency scan is always required for stability evaluation, which endures complicated implementation and can only be conducted under small amount of stable operating conditions. To solve this problem, a novel impedance profile prediction method for grid-connected VSCs has been proposed. A combined structure of stacked autoencoder (AE) and principal component analysis (PCA) is firstly proposed to extract VSC admittance feature under stable operating points, and a comprehensive VSC admittance set can be further predicted through searching on an enlarged feature space with unstable scenarios included. The stability can then be evaluated on the predicted VSC admittances with a stability boundary derived. Simulations and experiments prove the effectiveness of the proposed method.