Meshfree and Interpretable Data-Driven Dynamic Equivalence Modeling of Doubly-Fed Induction Generators Based on Kolmogorov-Arnold Networks
Yushan Yin, Xin He, Yuhong Wang, Shilin Gao, Xu Zhou · 2025
With the increasing integration of renewable energy sources into power systems, the accurate modeling of doubly-fed induction generators (DFIGs) has become crucial for ensuring grid stability and efficiency. However, current technologies struggle to balance the modeling accuracy and simulation efficiency required for large-scale wind turbines. To address the issue, this paper proposes a modeling approach that comprehensively considers both precision and timeliness. Kolmogorov-Arnold Networks (KANs) provide unique advantages such as interpretability and targeted modeling capabilities, enabling accurate representation of DFIG dynamics under complex conditions. This paper constructs a meshfree and interpretable data-driven dynamic equivalent model of DFIGs based on KANs. The study primarily accomplishes two tasks: Firstly, construct KANs for capturing dynamic characteristics of DFIGs, solving the trouble of model accuracy and simulation efficiency. Secondly, design a co-simulation interface integrated into the CloudPSS cloud simulation platform based on the first task, achieving integrated simulation under closed-loop conditions. Through multi-scenario simulation validation, the proposed method demonstrates higher accuracy, higher computational efficiency and exhibits certain practical utility in engineering applications.