NN-induced Physical Information Dynamic Library for Transient Modeling of Large-Scale Wind Farm

The complexity of transient characteristics in large-scale wind farms (WF) hinders the application of machine learning algorithms. This paper proposes a neural network-based learning method to provide physical information cues for the learning framework of transient characteristics in large-scale WF induced by physical information. The complexity of the physical information repository is simplified through an iterative algorithm. The dynamic library obtained based on neural networks can induce the machine learning framework to rapidly learn the transient characteristics of large-scale WF. Moreover, there is no need for excessive mechanistic analysis and speculation regarding the transient behavior of WF. The effectiveness of the proposed method is verified in the simulation model of a WF.