高级检索

    张哲源, 顾幸生. 基于分布式深度神经网络的双馈风机低压故障穿越研究[J]. 华东理工大学学报(自然科学版), 2023, 49(3): 401-409. DOI: 10.14135/j.cnki.1006-3080.20220105005
    引用本文: 张哲源, 顾幸生. 基于分布式深度神经网络的双馈风机低压故障穿越研究[J]. 华东理工大学学报(自然科学版), 2023, 49(3): 401-409. DOI: 10.14135/j.cnki.1006-3080.20220105005
    ZHANG Zheyuan, GU Xingsheng. Low Voltage Ride through Research on Distributed Deep Neural Network-Based Doubly Fed Induction Generator[J]. Journal of East China University of Science and Technology, 2023, 49(3): 401-409. DOI: 10.14135/j.cnki.1006-3080.20220105005
    Citation: ZHANG Zheyuan, GU Xingsheng. Low Voltage Ride through Research on Distributed Deep Neural Network-Based Doubly Fed Induction Generator[J]. Journal of East China University of Science and Technology, 2023, 49(3): 401-409. DOI: 10.14135/j.cnki.1006-3080.20220105005

    基于分布式深度神经网络的双馈风机低压故障穿越研究

    Low Voltage Ride through Research on Distributed Deep Neural Network-Based Doubly Fed Induction Generator

    • 摘要: 双馈风机的低电压穿越性能不仅依靠控制策略,还取决于对控制参数的选择。然而,控制参数优化算法耗时太久,在实时控制中达不到相应的效果。本文提出了一种基于离线参数优化和模型训练、在线故障识别的方法。首先通过已经建立的双馈风机(DFIG)并网模型获取大量不同类型的故障数据,并根据故障类型进行控制参数的离线优化,形成相应的低压穿越方式;然后将不同的故障数据进行分类,构成神经网络的训练样本;最后在电网故障瞬间利用训练好的分布式深度神经网络(Deep Neural Networks, DNN)迅速判断故障种类,并选择合适的控制策略。通过半实物仿真的方式验证了该方法的可行性以及在控制效果和速度方面的优势。

       

      Abstract: The low voltage ride through performance of doubly fed induction generator (DFIG) not only depends on the control strategy, but also depends on the selection of control parameters. Due to the time-consuming optimization algorithm of control parameters, it cannot achieve the corresponding effect in real-time control. Therefore, this paper proposes a method based on offline parameter optimization and model training, as well as online fault identification. Firstly, a large number of different types of fault data are obtained through the established DFIG grid connection model, and offline optimization of control parameters is carried out based on the fault type to form the corresponding low voltage ride through mode. Then, different fault data are classified to form training samples for neural network. At the moment of a power grid fault, the fault data can be directly used to quickly judge the fault type by the trained distributed deep neural network (DNN), and appropriate control strategies can be selected based on the fault type. The feasibility of this method and its advantages in control effectiveness and speed are verified through fault identification and parameter optimization method of doubly fed fan model.

       

    /

    返回文章
    返回