Neural Network-Assisted Identification of Power Converter Control Parameters: A Robust and Noise-Filtered Approach

With the widespread penetration of power converters into the electrical grid, Transmission System Operators are compelled to assess the dynamics of these converters to gain a comprehensive insight into their operations and facilitate new planning studies. However, due to manufacturer confidentiality, access to control parameters for developing a model is impossible. In this article, a method for identifying the control parameters of converters connected to the grid as a black box is presented. This method relies on a feedforward neural network and the admittance frequency model measured at the Point of Common Coupling. The robustness of the method to noise is assessed, and an autoencoder-based filtering approach is proposed to enhance identification performance, even in the presence of noise.