Optimization of PSS parameters using surrogate-assisted evolutionary algorithms based on KAN

This research presents an approach for tuning Power System Stabilizer (PSS) parameters using surrogate-assisted evolutionary algorithms based on the recently proposed Kolmogorov-Arnold Network (KAN). Developed from the Kolmogorov-Arnold representation theorem, KAN replaces traditional weight parameters with spline-adjustable univariate functions, enhancing both accuracy and interpretability over conventional methods. The proposed approach has been tested on a ten-machine system, with performance comparisons made against various surrogate models based methods. Our preliminary numerical simulation results indicate that PSS parameters obtained by the proposed method are able to improve low-frequency oscillation damping. In addition, using surrogate-assisted evolutionary algorithm accelerates the convergence towards optimal solutions compared to conventional parameter tuning methods.