Neural Network Modelling and Parameter Prediction of Drive Train in a DFIG Wind Turbine

The drive train is an important part of the doubly-fed induction generator (DFIG) wind turbine (WT), its model and parameters have vital influence on power system synchronous stability and frequency stability analysis. Therefore, an accurate drive train model is the prerequisite for studying the dynamic characteristics of new energy power systems. In order to solve the difficulty of identifying model parameters due to insufficient measurement information for large disturbances, a neural network model is proposed based on the rich historical response data under random small disturbances excitation during normal operation of the unit, and the corresponding relationship between the response data and model parameters is used to predict the driving system model parameters based on the current response data. Firstly, the BP neural network modelling principle is introduced. Secondly, the power spectrum characteristic data of response signal is extracted based on a simulation system with a DFIG wind farm integrated into an infinite system. Thirdly, the key parameters are selected based on the power spectrum sensitivity. Finally, the BP neural network model is built to reflect the nonlinear mapping between the response signal power spectrum and model parameters, then the model parameters are predicted based on trained neural network. The model error is also analyzed to validate the feasibility of data-driven modelling method for WTs.