Processing math: 100%
高级检索

基于深度神经网络的DFIG低电压穿越技术研究

余欣梅, 陈豪君, 王星华

余欣梅,陈豪君,王星华.基于深度神经网络的DFIG低电压穿越技术研究[J].南方能源建设,2021,08(03):122-130.. DOI: 10.16516/j.gedi.issn2095-8676.2021.03.018
引用本文: 余欣梅,陈豪君,王星华.基于深度神经网络的DFIG低电压穿越技术研究[J].南方能源建设,2021,08(03):122-130.. DOI: 10.16516/j.gedi.issn2095-8676.2021.03.018
YU Xinmei,CHEN Haojun,WANG Xinghua.Research on Low Voltage Ride Through Technology of DFIG Based on Deep Neural Networks[J].Southern Energy Construction,2021,08(03):122-130.. DOI: 10.16516/j.gedi.issn2095-8676.2021.03.018
Citation: YU Xinmei,CHEN Haojun,WANG Xinghua.Research on Low Voltage Ride Through Technology of DFIG Based on Deep Neural Networks[J].Southern Energy Construction,2021,08(03):122-130.. DOI: 10.16516/j.gedi.issn2095-8676.2021.03.018
余欣梅,陈豪君,王星华.基于深度神经网络的DFIG低电压穿越技术研究[J].南方能源建设,2021,08(03):122-130.. CSTR: 32391.14.j.gedi.issn2095-8676.2021.03.018
引用本文: 余欣梅,陈豪君,王星华.基于深度神经网络的DFIG低电压穿越技术研究[J].南方能源建设,2021,08(03):122-130.. CSTR: 32391.14.j.gedi.issn2095-8676.2021.03.018
YU Xinmei,CHEN Haojun,WANG Xinghua.Research on Low Voltage Ride Through Technology of DFIG Based on Deep Neural Networks[J].Southern Energy Construction,2021,08(03):122-130.. CSTR: 32391.14.j.gedi.issn2095-8676.2021.03.018
Citation: YU Xinmei,CHEN Haojun,WANG Xinghua.Research on Low Voltage Ride Through Technology of DFIG Based on Deep Neural Networks[J].Southern Energy Construction,2021,08(03):122-130.. CSTR: 32391.14.j.gedi.issn2095-8676.2021.03.018

基于深度神经网络的DFIG低电压穿越技术研究

基金项目: 

国家自然科学基金项目“基于张量技术的多视图特征选择方法研究” 61903091

详细信息
    作者简介:

    余欣梅(通信作者)1978-,女,浙江衢州人,高级工程师、注册咨询师,工学博士,主要从事能源与电力规划咨询研究的工作(e-mail)yuxinmei@gedi.com.cn

    陈豪君1995-,男,广东佛山人,工学硕士,从事新能源接入及运行控制等研究(e-mail)2621344218@qq.com

    王星华1972-,男,江苏扬州人,副教授,工学硕士,从事电力系统高级应用软件开发,大数据分析、含分布式发电的配电网络规划优化研究(e-mail)1805211@qq.com

  • 中图分类号: TM7

Research on Low Voltage Ride Through Technology of DFIG Based on Deep Neural NetworksEn

  • 摘要:
      目的  双馈风机(DFIG)的低电压穿越(LVRT)性能在一定程度上依赖于控制参数的优化,而目前对控制参数的优化基本都是离线模式,原因在于优化算法难以满足实时控制对计算速度的要求。
      方法  基于深度神经网络(DNN)原理,提出基于“离线训练、在线计算”思路的低电压穿越实时优化控制方法。首先针对含DFIG电网在不同运行方式下发生的大量不同故障进行控制参数的离线优化并形成相应的LVRT优化控制策略,将不同的故障状态进行归类并形成成对的故障模式和参数表,从而构成深度神经网络的训练样本。
      结果  电网故障瞬间可以将输入参数直接通过训练好的DNN网络迅速实现控制方案和最优参数的优选。
      结论  基于PSCAD和Matlab的联合仿真结果论证了所提思想在优化效果和优化速度方面的优势,并说明了其实用性。
    Abstract:
      Introduction  During the power grid fault, the low voltage ride through (LVRT) performance of the doubly-fed induction generator (DFIG) depends on the control parameters. At present, the optimization of control parameters is basically in the offline mode, which lies in the fact that it's hard for algorithm optimization to meet real-time control's requirement of the calculation speed.
      Method  Therefore, the real-time optimization control method of LVRT following“offline training, online computation” was presented based on the principles of deep neural networks (DNN). Firstly, the appropriate LVRT strategy for optimization control was proposed for different fault severity levels. The parameters were clustered and optimized according to the respective objectives of each strategy, then the parameter list was formed.
      Result  At the moment of power grid fault, the input parameters can be directly input into the trained DNN networks to quickly realize the optimization of control scheme and optimal parameters.
      Conclusion  The joint simulation results based on PSCAD and MATLAB demonstrate the advantages of the proposed idea in optimization effect and optimization speed and the practicability is also illustrated.
  • 图  1   联合模块示意图

    Figure  1.   Illustration of combined model

    图  2   转子侧变流器控制框图

    Figure  2.   Schematic diagram

    图  4   撬棒电路的拓扑结构

    Figure  4.   Topology of crowbar circuit

    图  7   DNN网络结构图

    Figure  7.   Structure of DNN

    图  8   仿真系统图

    Figure  8.   Diagram of simulation system

    表  1   PI控制器参数对比

    Table  1   Comparison of PI parameters

    参数项kpkI
    初始控制参数值0.90.3
    优化后控制参数值0.7420.335
    下载: 导出CSV

    表  2   优化前后撬棒阻值对比

    Table  2   Comparison of Crowbar crowbar resistance

    参数项参数值
    固定Crowbar电阻整定值/Ω2
    优化后Crowbar电阻值/Ω1.625
    对应D1占空比0.923
    下载: 导出CSV

    表  3   最优控制参数

    Table  3   Optimal control parameters

    参数kpkIRCB
    优化值0.8610.2371.344
    下载: 导出CSV

    表  4   策略参数表

    Table  4   Parameter table

    序号故障程度LVRT策略优化参数
    0轻度故障改进控制kP=0.742; kI=0.335
    1中度故障联合控制kP=0.861; kI=0.237; Rcb=1.344(Ω)
    2严重故障撬棒保护Rcb=1.625(Ω)
    下载: 导出CSV

    表  5   训练样本仿真参数设置

    Table  5   Simulation parameters for training set

    类型参数取值数量
    故障位置/km0,10,15,20,25,30,35,40,45,5010
    过渡电阻0,5,10,15,20,25,30,35,409
    故障程度轻度(0),中度(1),严重(2)3
    下载: 导出CSV

    表  6   测试样本仿真参数设置表

    Table  6   Simulation parameters for testing set

    类型参数取值数量
    故障位置/km5,12,18,24,32,38,497
    过渡电阻0,5,10,15,20,25,30,35,409
    故障程度轻度(0),中度(1),严重(2)3
    下载: 导出CSV

    表  7   不同网络参数下DNN故障识别仿真结果

    Table  7   Results of simulation

    序号网络结构(层数、神经元数量)最大迭代次数训练率平均训练误差测试准确率/%训练时间/s
    1500-400-300-1005 0000.000 10.005 695898
    2500-400-3005 0000.000 10.009 092.5839
    3500-400-300-1005 0000.0010.003 995974
    4200-100-100-505 0000.000 10.007 592.5501
    5200-100-100-502 5000.000 10.036 190253
    下载: 导出CSV
  • [1] 张学广,徐殿国,潘伟明,等. 基于电网电压定向的双馈风力发电机灭磁控制策略 [J]. 电力系统自动化,2010,34(7):95-99.

    ZHANGX G,XUD G,PANW M,et al. A flux damping control strategy of doubly-fed induction generator based on the grid voltage vector oriented [J]. Automation of Electric Power Systems,2010,34(7):95-99.

    [2] 杨晨星,杨旭,童朝南. 双馈异步风力发电机低电压穿越的软撬棒控制 [J]. 中国电机工程学报,2018,38(8):2487-2495+2558.

    YANGC X,YANGX,TONGC N. An LVRT control strategy based on soft crowbar control for doubly fed induction wind power generations [J]. Proceedings of the CSEE,2018,38(8):2487-2495+2558.

    [3] 朱晓东,石磊,陈宁,等. 考虑Crowbar阻值和退出时间的双馈风电机组低电压穿越 [J]. 电力系统自动化,2010,34(18):84-89.

    ZHUX D,SHIL,CHENN,et al. An analysis on low voltage ride through of wind turbine driven doubly fed induction generator with different resistances and quitting time of crowbar [J]. Automation of Electric Power Systems,2010,34(18):84-89.

    [4] 郑涛,魏旭辉,李娟,等. PI控制参数对双馈风电机组短路电流特性的影响分析 [J]. 电力自动化设备,2016,36(7):15-21.

    ZHENGT,WEIX H,LIJ,et al. Influence of PI control parameters on short circuit current characteristics of DFIG [J]. Electric Power Automation Equipment,2016,36(7):15-21.

    [5] 谷昱君,聂暘,王东,等. 基于改进免疫遗传算法的双馈风机控制系统PI参数优化 [J]. 陕西电力,2016,44(11):25-30.

    GUY J,NIEY,WANGD,et al. PI parameters optimization of DFIG control system based on improved immune genetic algorithm [J]. Shanxi Electric Power,2016,44(11):25-30.

    [6] 黄琳妮. 基于群飞蛾扑火算法的风力发电系统PI控制参数优化整定 [D]. 广州:华南理工大学,2018.

    HUANGL N. Optimization of PI control parameters of wind energy system based on swarm moths flame algorithm [D]. Guangzhou:South China University of Technology,2018.

    [7] 李达,薛卿,孔德健,等. 基于PSO-BP算法的高压输电线路故障分类 [J]. 电气自动化,2018,40(6):46-48.

    LID,XUEQ,KONGD J,et al. Fault classification of high voltage transmission line based on PSO-BP algorithm [J]. Electrical Automation,2018,40(6):46-48.

    [8] 马浩淼,高勇,杨媛,等. 双馈风力发电低电压穿越撬棒阻值模糊优化 [J]. 中国电机工程学报,2012,32(34):17-23+4.

    MAH M,GAOY,YANGY,et al. Fuzzy optimization of crowbar resistances for low-voltage ride through of doubly-fed induction generators [J]. Proceedings of the CSEE,2012,32(34):17-23+4.

    [9] 王国权. 双馈风力发电系统低电压穿越优化研究 [D]. 郑州:华北水利水电大学,2017.

    WANGG Q. Study on optimization of low voltage ride through for doubly fed wind power generation system [D]. Zhengzhou:North China University of Water Resources and Electric Power,2017.

    [10] 贺益康,周鹏. 变速恒频双馈异步风力发电系统低电压穿越技术综述 [J]. 电工技术学报,2009,24(9):140-146.

    HEY K,ZHOUP. Overview of the low voltage ride-through technology for variable speed constant frequency doubly fed wind power generation systems [J]. Transactions of China Electrotechnical Society,2009,24(9):140-146.

    [11] 陈豪君. 基于人工智能的双馈风机低电压穿越控制策略研究 [D]. 广州:广东工业大学,2020.

    CHENH J. Research on control strategy for low-voltage ride through of double-fed fan based on artificial intelligence [D]. Guangzhou:Guangdong University of Technology,2020.

    [12]

    MORRENJ,HAANS W H D. Short-circuit current of wind turbines with doubly fed induction generator [J]. IEEE Transactions on Energy Conversion,2007,22(1):174-180.

    [13] 孔祥平,张哲,尹项根,等. 计及励磁调节特性影响的双馈风力发电机组故障电流特性 [J]. 电工技术学报,2014,29(4):256-265.

    KONGX P,ZHANGZ,YINX G,et al. Fault current characteristics of DFIG considering excitation and regulation characteristics [J]. Transactions of China Electrotechnical Society, 2014,29(4):256-265.

    [14] 杨健维,麦瑞坤,何正友. PSCAD/EMTDC与Matlab接口研究 [J]. 电力自动化设备,2007(11):83-87.

    YANGJ W,MAI R K,HEZ Y. Interface between PSCAD/EMTDC and Matlab [J]. Electric Power Automation Equipment,2007(11):83-87.

    [15] 卫志农,缪新民,王华伟,等. 基于PSCAD-MATLAB联合调用的高压直流控制系统参数优化 [J]. 高电压技术,2014,40(8):2449-2455.

    WEIZ N,MIAOX M,WANGH W,et al. Parameter optimization for HVDC control system based on PSCAD-MATLAB combined invocation [J]. High Voltage Engineering, 2014,40(8):2449-2455.

    [16]

    QIANY,FANY,HUW,et al. On the training aspects of deep neural network(DNN)for parametric TTS synthesis [C]//IEEE.ICASSP 2014-2014 IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP), Florence,Italy,May 4-9, 2014. Florence:IEEE,2014:3829-3833.

    [17]

    WUZ Z,SWIETOJANSKIP,VEAUXC,et al. A study of speaker adaptation for DNN-based speech synthesis [C]//International Speech Communication Association. Proceedings of Interspeech 2015,Dresden,Germany,Sep.6-9,2015. Dresden:International Speech Communication Association,2015: 879-883.

    [18] 余达. 基于深度学习的风力发电系统故障在线诊断研究 [D].广州:华南理工大学,2018.

    YUD. Online fault diagnosis of wind Power systems using deep learning Algorithms [D]. Guangzhou:South China University of Technology,2018.

  • 期刊类型引用(4)

    1. 华文,董炜,陆翌,郑晨一. 考虑暂态功角稳定的风机短路电流与无功支撑关键参数优化方法. 可再生能源. 2025(02): 275-284 . 百度学术
    2. 刘洪涛,马骞,黄河,黄兆棽,朱益华,李成翔. 考虑功角稳定的新能源电压穿越控制参数优化方法. 能源与环保. 2025(03): 219-226 . 百度学术
    3. 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. 华东理工大学学报(自然科学版). 2023(03): 401-409 . 百度学术
    4. 赵康,张志轩,周宁,王亮,贺政华,陈俊超. 基于多层神经网络的直驱风机低穿控制参数辨识. 山东电力技术. 2022(04): 1-6 . 百度学术

    其他类型引用(1)

图(23)  /  表(7)
计量
  • 文章访问数:  783
  • HTML全文浏览量:  209
  • PDF下载量:  45
  • 被引次数: 5
出版历程
  • 收稿日期:  2021-07-25
  • 修回日期:  2021-08-23
  • 刊出日期:  2021-09-24

目录

    Xinghua WANG

    1. On this Site
    2. On Google Scholar
    3. On PubMed

    /

    返回文章
    返回