双馈风电机组传动系统神经网络建模及参数预测  

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

作　　者：丁新虎 潘学萍[1] 孙晓荣 和大壮 陈海东 DING Xinhu;PAN Xueping;SUN Xiaorong;HE Dazhuang;CHEN Haidong(College of Energy and Electrical Engineering,Hohai University,Nanjing 210098,Jiangsu Province,China)

机构地区：[1]河海大学能源与电气学院,江苏省南京市210098

出　　处：《现代电力》2024年第2期201-208,共8页Modern Electric Power

基　　金：国家自然科学基金资助项目(52077061)。

摘　　要：传动系统是双馈风电机组的重要组成部分,其模型对电力系统同步稳定及频率稳定分析具有重要影响,准确的传动系统模型参数是分析新能源电力系统动态特性的前提。为解决因大扰动量测信息不充裕导致模型参数难以辨识的困难,提出利用机组正常运行状态时随机小扰动激励下丰富的历史响应数据,根据响应数据与模型参数的对应关系构建神经网络模型,并根据当前响应数据进行驱动系统模型参数预测。首先讨论了基于BP神经网络进行数据建模的基本流程;针对含双馈风电机组的无穷大系统仿真算例,提取随机风速扰动下响应信号受扰轨迹的功率谱特征;定义功率谱灵敏度指标,提出选取功率谱灵敏度较大的参数作为重点参数;最后基于BP神经网络构建响应信号功率谱与模型参数之间的非线性映射,基于训练得到的BP网络辨识新响应下的模型参数。通过分析BP神经网络动态模型的误差,验证数据驱动建模方法的可行性。The drive train is an important part of the doublyfed 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.

关 键 词：双馈风电机组 参数预测 功率谱特征 可辨识性 BP神经网络 

分 类 号：TM73[电气工程—电力系统及自动化]