Physics-Informed Learning Based Wind Farm Two-Machine Aggregation Model for Large-Scale Power System Stability Studies

The aggregation model (AM) of wind farms (WFs) decreases the computation burden but may reduce accuracy. Further, it may be difficult to appropriately determine the behaviors of wind turbines under power system transients. This paper proposes a novel aggregation modeling method of WFs for power system transient analysis. It involves mainly two design stages. First, a dendrogram algorithm is utilized to generate a simple and generic model (GM). Second, the GM is refined using a WFs-tailored partial differential equations functional identification of nonlinear dynamic (PDE-FIND) algorithm so that the GM accuracy is improved. In this stage, the dynamic library of PDE-FIND is reformulated to contain the probably existing variables which can express the power error equations. To extract dominant variables and produce a precisionadjustable AM, a requirements-oriented regression algorithm is then proposed, whilst also balancing the accuracy and simplicity of the model. To verify the effectiveness of the propo-sed method, the dynamic responses of the GM and refined model are compared against the traditional model.