双馈发电系统的神经网络转速辨识

Rotor Speed Identification of Doubly-Fed Generator System Based on Neural Network

根据双馈发电机的数学模型,建立了基于模型参考自适应系统(MRAS)的参考模型和可调模型。利用后项差分法推导了基于神经网络的MRAS可调模型,并用两层神经网络实现可调模型中的磁链运算,通过误差反传算法对两层神经网络进行训练,得出双馈发电机的辨识转速。仿真结果表明:基于神经网络的MRAS辨识转速能反映实际转速,且辨识精度得到了有效提高。

According to the mathematic model of doubly-fed induction generator, the adjustable model and reference model based on MRAS were found. The adjustable model of model reference adaptive system based on neural network was derived by backward differentiation method and the algorithm for magnetic flux linkage was determined using two-layer neural network. Besides, the speed identification of doubly-fed induction generator was obtained by training of two-layer neural network using error back propagation. Finally, the simulation results showed that compared with the speed identification of model reference adaptive system, the rotor speed can be reflected actually, and the speed estimation precision was effectively improved when neural network speed estimator was applied.