高级检索

  • ISSN 1006-3080
  • CN 31-1691/TQ

基于分布式深度神经网络的双馈风机低压故障穿越研究

张哲源 顾幸生

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). 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). doi: 10.14135/j.cnki.1006-3080.20220105005
ZHANG Zheyuan, GU Xingsheng. Low Voltage Ride Through Research on Distributed DNN-Based DFIG[J]. Journal of East China University of Science and Technology. doi: 10.14135/j.cnki.1006-3080.20220105005
Citation: ZHANG Zheyuan, GU Xingsheng. Low Voltage Ride Through Research on Distributed DNN-Based DFIG[J]. Journal of East China University of Science and Technology. 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

doi: 10.14135/j.cnki.1006-3080.20220105005
详细信息
    作者简介:

    张哲源(1997—),男,山西忻州人,硕士生,主要研究方向为电力系统仿真。E-mail:867878095@qq.com

    通讯作者:

    顾幸生,E-mail:xsgu@ecust.edu.cn

  • 中图分类号: TP183

Low Voltage Ride Through Research on Distributed DNN-Based DFIG

  • 摘要: 双馈风机的低电压穿越性能不仅依靠着控制策略,还取决于对控制参数选择。由于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,所以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(Doubly Fed Induction Generator,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。电网故障瞬间可以将故障数据直接通过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,并根据故障类型选择合适的控制策略。通过双馈风机模型的故障识别和参数优化方法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。

     

  • 图  1  转子侧变流器控制框图

    Figure  1.  Control block diagram of rotor side converter

    图  2  撬棒保护电路

    Figure  2.  Crowbar protection circuit

    图  3  电压跌落30%以内低压穿越特性对比

    Figure  3.  Comparison of low voltage ride through characteristics within 30% voltage drop

    图  4  电压跌落至32%时低压穿越特性对比

    Figure  4.  Comparison of low voltage ride through characteristics when the voltage drops to 32%

    图  5  电压跌落至45%时低压穿越特性对比

    Figure  5.  Comparison of low voltage ride through characteristics when voltage drops to 45%

    图  6  DNN结构图

    Figure  6.  DNN structure diagram

    图  7  分布式DNN训练流程图

    Figure  7.  Distributed DNN training flow chart

    表  1  控制器参数对比

    Table  1.   Controller parameter comparison

    Parameter item Initial value Optimization
    kpd 0.8 0.98
    kid 0.2 0.15
    kpq 972.222 1138
    kiq 8.0e-05 0.01
    hy 0.1 1.08
    下载: 导出CSV

    表  2  优化前后Crowbar电阻

    Table  2.   Crowbar resistance before and after optimization

    Crowbar resister/Ω
    Before optimizationAfter optimization/Ω
    3024.58
    下载: 导出CSV

    表  3  优化前后控制参数对比

    Table  3.   Comparison of control parameters before and after optimization

    Parameter
    value
    Before
    optimization
    Optimization
    (<1.25Ir)
    Optimization
    (>=1.25Ir)
    kpd 0.8 0.51
    kid 0.2 0.13
    kpq 972.222 1003.25
    kiq 8.0e-05 8.0e-05
    hy 0.1 0.15
    Rc 30 13.38
    下载: 导出CSV

    表  4  故障仿真训练集参数表

    Table  4.   Parameter table of fault simulation training set

    Fault divider resistor/Ω Fault fall degree Fault type
    5 94.22% Minor
    2 87.23% Minor
    1 77.35% Minor
    0.5 63.58% General
    0.35 52.44% General
    0.25 45.57% General
    0.15 32.69% Critical
    0.1 24.88% Critical
    下载: 导出CSV

    表  5  故障仿真测试集参数表

    Table  5.   Parameter table of fault simulation test set

    故障分压电阻/Ω Fault degree Fault type
    1.5 83.58% Minor
    0.3 49.88% General
    0.13 30.32% Critical
    下载: 导出CSV

    表  6  不同策略下DNN故障识别仿真结果

    Table  6.   Simulation results of DNN fault identification under different strategies

    StrategyNetwork structureMaximum number of iterationsTraining rateAverage errorTest accuracy/%Average training time/s
    130-20-1050000.010.0015698.102241
    230-1050000.010.0148895.791318
    330-20-1050000.010.0054196.931577
    430-20-1050000.010.0045997.641511
    下载: 导出CSV
  • [1] YAO Z H. Control strategy of doubly-fed induction wind turbine[J]. Journal of Physics:Conference Series, 2020, 1650(3): 159-221. doi: 10.1088/1742-6596/1650/3/032007
    [2] 徐殿国, 王伟, 陈宁. 基于撬棒保护的双馈电机风电场低电压穿越动态特性分析[J]. 中国电机工程学报, 2010, 30(22): 29-36.
    [3] AGUILAR M, COURY D V, REGINATTO R, et al. Multi-objective PSO applied to PI control of DFIG wind turbine under electrical fault conditions[J]. Electric Power Systems Research, 2020, 180(c): 810-819. doi: 10.1016/j.jpgr.2019.106081
    [4] 谷鑫. 基于有功功率平衡和蚁群算法优化PI参数的有源滤波器[D]. 天津: 天津大学, 2006.
    [5] GEBRU F M, KHAN B, ALHELOU H H, et al. Analyzing low voltage ride through capability of doubly fed induction generator based wind turbine[J]. Computers & Electrical Engineering, 2020, 86(3): 1214-1218.
    [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.
    [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.
    [8] TAMAARAT A . Active and reactive power control for DFIG using PI, fuzzy logic and self-tuning pi fuzzy controllers. 2019, 74(2/4) : 95-102.
    [9] MASA A, KKM B, SB A, et al. Modified rotor-side converter control design for improving the LVRT capability of a DFIG-based WECS - ScienceDirect[J]. Electric Power Systems Research, 2020, 186(C): 440-444.
    [10] 贺益康, 周鹏. 变速恒频双馈异步风力发电系统低压穿越技术[J]. 电工技术学报. 2009, 24(9): 140-146.
    [11] 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, 27(11): 83-87. doi: 10.3969/j.issn.1006-6047.2007.11.021
    [12] CHONG C H, RIGIT A R H, ALI I. Wind turbine modelling and simulation using Matlab / SIMULINK[J]. IOP Conference Series:Materials Science and Engineering, 2021, 1101(1): 307-316.
    [13] AKANTO J M, HAZARI M R, MANNAN M A. LVRT and stability enhancement of grid-tied wind farm using DFIG-based wind turbine[J]. Applied System Innovation, 2021, 4(2): 33-40. doi: 10.3390/asi4020033
    [14] 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, 8(03): 122-130.
    [15] PENA, R C. Doubly fed induction generator using back-to-back PWM converters and its application to variable-speed wind-energy generation[J]. Electric Power Applications Iee Proceedings, 1996, 143(3): 231-238. doi: 10.1049/ip-epa:19960288
    [16] STALLON S R D, RAJKUMAR M N. Improving the performance of gridヽonnected doubly fed induction generator by fault identification and diagnosis: A kernel PCA〦SMO technique[J]. International Transactions on Electrical Energy Systems, 2021, 31(4): 48-55.
    [17] LI X F, QASIM S, YU C W. Influence of GA-BP artificial neural network based on PCA dimension reduction in yarn tenacity prediction[J]. Advanced Materials Research, 2014, 3569(1048): 358-366.
    [18] ZHENG L Y, HE H M. Share price prediction of aerospace relevant companies with recurrent neural networks based on PCA[J]. Expert Systems With Applications, 2021, 183(18): 518-525.
    [19] 郭林, 唐晶, 唐黎哲, 等. 一种基于改进BP神经网络的变压器故障诊断方法[J]. 控制与信息技术, 2021(05): 71-77.
    [20] YAO Q , FAN Y , HU W , et al. On the training aspects of Deep Neural Network (DNN) for parametric TTS synthesis[C]// IEEE International Conference on Acoustics. Florence, Italy: IEEE, 2014.
    [21] WU Z , SWIETOJANSKI P , VEAUX C , et al. A study of speaker adaptation for DNN-based speech synthesis[C]//Sixteenth Annual Conference of the International Speech Communication Association. [l. n. ]: Dresden, Germany: ISCA, 2015.
    [22] 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.
  • 加载中
图(7) / 表(6)
计量
  • 文章访问数:  55
  • HTML全文浏览量:  21
  • PDF下载量:  1
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-01-05
  • 网络出版日期:  2022-05-14

目录

    /

    返回文章
    返回