A Tensor-Based Clustering Method for Dynamic Equivalent Modeling of Wind Farms

Abstract

Simulating large-scale wind farms (WFs) with detailed wind turbine (WT) models is computationally costly, which motivates exploration for simplified modeling strategies while maintaining accuracy. However, due to complicated wind speed conditions and network structure, it is still a challenge to achieve accurate transient equivalence of WFs. To address this problem, this paper proposed, for the first time, a tensor decomposition-based clustering method that can capture high-dimensional transient characteristics of WF to achieve precise reduced-order modeling through more rational grouping. More specifically, we first formulate a tensor-structure-based data set to preserve the intrinsic spatio-temporal properties of the dynamic WF responses. Then, a tensor decomposition strategy considering sparsity and smoothness is also proposed to extract the low-dimensional features that further assist us in the clustering design. Finally, we tailor an accurate WF network aggregation strategy to reduce power loss errors. The simulation results for different WF layouts, system faults, and wind scenarios reveal the excellent performance of the proposed method.