Data-Driven Aggregation Model of Type-III Wind Farms for Power System Transient Characteristic Simulation

The transient characteristics of large-scale Type-III wind farms (WFs) exhibit high complexity and weak interpretability. To analyze their dynamic response behavior, this paper proposes a two-stage reduced-order modeling method that comprehensively incorporates factors, such as wind speed and collector network impedance. By aggregating both wind speed and line impedances, the proposed approach enhances the generalization and explanation of the aggregated WF model. Furthermore, an expert library integrating prior knowledge with neural network algorithms is constructed, endowing the model with both physical interpretability and predictive accuracy. As the proposed algorithm includes a linear regression approach to extract the critical elements in the designed expert library, the data-driven aggregation model only uses a few datasets for model generation, while balancing the model’s accuracy and simplicity. Validated in the MATLAB/Simulink environment through three-phase and two-phase short-circuit fault tests, the results indicate that the established aggregated model achieves high consistency with the detailed model in the full dynamic response of both power and voltage at the point of common coupling (PCC).