A hybrid mechanism-based and data-driven dynamic equivalence method for PMSG wind farms considering wake effect and fault severity

To improve the simulation efficiency of large-scale wind farms, it is necessary to construct the equivalent model of wind farms. Due to the shortcomings in the calculation of clustering indicators and adaptation to parameter changes of wind turbines, the traditional methods are difficult to accurately and efficiently simulate the transient response for wind farms after the fault. To this end, the dynamic equivalence method considering the wake effect and fault severity for permanent magnet synchronous generator (PMSG) wind farms is proposed in this paper. First, the response of PMSGs is classified according to the initial wind speed and fault severity. Second, the wind speeds of PMSGs are estimated by the wake model considering nonlinear interactions between multiple PMSGs, which are closer to the wind speed distribution in actual wind farms. The calculation method of terminal voltage for each PMSG during the fault based on the extended long short-term memory (xLSTM) network is proposed, which achieves efficient computation of clustering indicators. Subsequently, a parameter calibration method for the equivalent parameters of wind farms based on the Gaussian mixture model is designed, which can adapt to the parameter change of wind turbines. Finally, case studies verify that the clustering efficiency of wind farms can be enhanced and the equivalent error is reduced by more than 20% after using the proposed method. • The wake effect affects the clustering accuracy of the large-scale wind farms. • Calculating clustering indicators by neural networks improves the equivalence efficiency. • The calibrated equivalent parameters can adapt to parameter variation of wind farms.