摘要
针对高压直流(high voltage direct current,HVDC)输电线路故障暂态行波具有时序性和强非线性的特点,导致高过渡电阻情况下故障识别率低的问题,提出基于卷积神经网络(convolutional neural networks,CNN)和双向循环门单元(bidirectional gate recurrent unit,BiGRU)的HVDC输电线路故障识别方法。首先,采用故障后整流侧的双极暂态电流行波作为特征向量,利用CNN提取全局特征,并从中剔除噪声和不稳定成分,完成对数据的降维处理。然后,采用BiGRU来捕获CNN提取到特征的前后时间信息,进一步提取数据中的时序特征,以实现HVDC输电线路故障识别。仿真结果表明:该方法可在不同故障地点以及不同过渡电阻下对单极接地、双极短路、雷击故障、雷击干扰共四种故障实现准确识别,可靠性高,具有较强的耐受过渡电阻能力,同时具备一定的抗噪性能。
Aiming at the problem that the fault transient traveling wave of high voltage direct current(HVDC)transmission line has the characteristics of timing and strong nonlinearity,resulting in low fault identification rate under high transition resistance,a fault identification method for HVDC transmission line based on convolutional neural network(CNN)and bidirectional gate recurrent unit(BiGRU)is proposed.Firstly,the bipolar transient current traveling wave on the rectification side after the fault is used as the feature vector,and CNN is used to extract global features,removing noise and unstable components,and completing the dimensionality reduction processing of the data.Then,BiGRU is used to capture the temporal information of the features extracted by CNN,and further extract the temporal features in the data to achieve HVDC transmission line fault recognition.The simulation results show that this method can accurately identify four types of faults:single pole grounding,bipolar short circuit,lightning fault,and lightning interference at different fault locations and under different transition resistances.It has high reliability,strong resistance to transition resistances,and certain anti noise performance.
作者
赵妍
王泽通
邢士标
朱建华
陈阔
张思博
ZHAO Yan;WANG Zetong;XING Shibiao;ZHU Jianhua;CHEN Kuo;ZHANG Sibo(Northeast Electric Power University,Jilin 132012,China;State Grid Siping Power Supply Company,Siping 136000,China;Rundian Energy Science and Technology Company Limited,Zhengzhou 450000,China)
出处
《吉林电力》
2024年第1期29-34,39,共7页
Jilin Electric Power
关键词
深度学习
高压直流
卷积神经网络
双向循环门单元
故障识别
deep learning
high voltage direct current
convolutional neural network
bidirectional gate recurrent unit
fault identification
