基于深度子领域自适应的直驱风机次/超同步振荡源定位

Location of Sub-/Super-synchronous Oscillation Source for Direct-driven Wind Turbines Based on Deep Subdomain Adaptation

直驱风机接入弱交流系统会引发形态较复杂的次/超同步振荡,为了更加精准有效地抑制振荡,亟须及时定位引发振荡的风电机组。考虑到电力系统实际振荡数据匮乏的现状,文中提出了一种基于深度子领域自适应的振荡源定位方法。该方法将仿真系统中的强迫振荡泛化到实际次/超同步振荡领域,通过输入样本与迁移模型的构建,实现次/超同步振荡源在线定位,为振荡数据匮乏导致机器学习困难提供了一种新的解决方案。此外,设计了含直驱风机并网的仿真系统测试算例,并将所提方法与不同振荡源定位方法进行了对比。结果表明,所提方法能够在较短的时间内给出更为准确的定位信息,为进一步实现振荡源在线识别奠定了基础。

Direct-driven wind turbines connected to weak AC systems can cause sub-/super-synchronous oscillation with complex patterns.In order to suppress the oscillation more accurately and effectively,it is urgent to locate the wind turbines causing the oscillation.Considering the lack of actual oscillation data in the power system,this paper proposes an oscillation source location method based on the deep subdomain adaptation.This method generalizes the forced oscillation in the simulation system to the domain of actual sub-/super-synchronous oscillation,and realizes the online location of the sub-/super-synchronous oscillation source through the construction of the input sample and the transfer model,which provides a new solution for the difficulty of machine learning caused by the lack of oscillation data.In addition,a test case of a simulation system with multiple direct-driven wind turbines connected to the power grid is designed,and the proposed method is also compared with different oscillation source location methods.The results show that the proposed method can provide more accurate location information in a shorter time,which lays the foundation for further online oscillation source identification.