Clustering and Equivalent Modeling Method for Direct-Drive Wind Farms Based on Particle Swarm Neural Network

In the realm of large-scale integration of wind farms or clusters into power grids, the fault transient characteristics of these grids have experienced significant transformations. Unfortunately, the prevalent single-machine equivalent models are inadequate in accurately depicting the fault transient behavior of wind farms or clusters. Moreover, their simulation computations are exceedingly intensive, resulting in suboptimal simulation efficiency. To address these challenges, this paper introduces a clustering and equivalent modeling approach tailored for direct-drive wind farms, leveraging the capabilities of a particle swarm neural network. Initially, a transient model for direct-drive wind turbines is formulated, and fault characteristic influence factors are established through theoretical analysis. These factors encompass the distance between the wind turbine and the point of common coupling (PCC), the implementation of DC-side current limiting measures, wind speed, and reactive power at the outlet. Following this, the Euclidean distance of the influence factor set is computed, and the initial cluster centers of wind turbines are derived using an enhanced max-min distance method. Ultimately, an equivalent clustering model is formulated by utilizing the influence factors and initial cluster centers as the training set. Simulation outcomes reveal that, in comparison to the single-machine equivalent model, this method enhances simulation efficiency by a range of ${6 2 \%}$ to ${6 7 \%}$ and reduces the steady-state error in fault current simulation by ${0 . 8 \%}$ to ${1 . 9 \%}$.