摘要
随着分布式电源的大量接入,微网系统的谐波阻抗测量更具有动态性和不确定性,对系统谐波阻抗的主动测量是谐波分析、滤波器设计等的重要基础。当分析微网系统的稳定性时,其dq轴谐波阻抗参数的获得显得至关重要。文章中对微网系统诺顿等效电路的dq轴谐波阻抗进行了理论的推导,建立了RL系统谐波阻抗模型。为了提高预测精度,文章引入了BP神经网络预测模型,介绍了其核心算法梯度下降原理,并建立了基于系统dq轴谐波阻抗的预测模型。然后为了得到不相关的注入谐波电流矢量以求解系统谐波阻抗矩阵,文章中介绍了旋转电流相量法,主动获得系统dq轴谐波阻抗。对微网系统dq轴谐波阻抗的测量和预测进行了仿真验证,其仿真结果证明了文章中所提策略的有效性。
With the large access of distributed power supplies,the harmonic impedance measurement of the micro-grid system is more dynamic and uncertain.The active measurement of the harmonic impedance of the system is an important basis for harmonic analysis and filter design.When analyzing the stability of the micro-grid system,the acquisition of the dq-axis harmonic impedance parameters is crucial.In this paper,the dq-axis harmonic impedance of the Norton equivalent circuit of the micro-grid system is theoretically deduced,and the harmonic impedance model of the RL system is established.In order to improve the prediction accuracy,this paper introduces the prediction model of BP neural network,and introduces its core algorithm-gradient descent principle,and establishes a prediction model based on the dq-axis system harmonic impedance.Then,in order to obtain the irrelevant injecting harmonic current vector to solve the system harmonic impedance matrix,this paperintroduces the rotating current phasor method and actively obtains the system dq-axis harmonic impedance.Finally,the simulation and prediction of the dq-axis harmonic impedance of the microgrid system are carried out.The simulation results prove the effectiveness of the proposed strategy.
作者
王胜
冯兴明
周宇
曾江
李志华
Wang Sheng;Feng Xingming;Zhou Yu;Zeng Jiang;Li Zhihua(Yancheng Power Supply Company,State Grid Jiangsu Electric Power Company,Yancheng 224001,Jiangsu,China;School of Electric Power Engineering,South China University of Technology,Guangzhou 510640,China)
出处
《电测与仪表》
北大核心
2021年第3期118-125,共8页
Electrical Measurement & Instrumentation
基金
国家自然科学基金资助项目(51777078)
国网江苏省电力有限公司科技项目(J2018107)。
关键词
微电网
dq轴谐波阻抗
BP神经网络预测
旋转电流相量
梯度下降
micro-grid system
dq-axis harmonic impedance
BP neural network prediction
rotating current phasor
gradient descent
