基于深度学习attention机制的不带并联电抗器输电线路自适应重合闸策略  

Deep learning attention mechanism based adaptive reclosing strategy for transmission lines without shunt reactors

作　　者：周红 王海云 王维庆 王利超 ZHOU Hong;WANG Haiyun;WANG Weiqing;WANG Lichao(Engineering Research Center of Ministry of Education for Renewable Energy Generation and Grid Connection Technology,Xinjiang University,Urumqi 830047,China;Electric Power Research Institute,State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830011,China)

机构地区：[1]新疆大学可再生能源发电与并网技术教育部工程研究中心,新疆乌鲁木齐830047 [2]国网新疆电力有限公司电力科学研究院,新疆乌鲁木齐830011

出　　处：《现代电子技术》2022年第23期171-176,共6页Modern Electronics Technique

摘　　要：常规的线路重合闸装置在线路跳闸后采用的固定延时后直接重合的方式存在一定的弊端,为避免断路器重合于永久性故障给电网造成的冲击,提出一种基于注意力机制的深度学习算法,用于改进对于不带并联电抗器的输电线路在进行重合闸前完成对故障状态的自适应判断,从而进一步提高断路器重合的准确性和可靠性。在线路两侧的断路器跳开后通过在线路的一端增设常规电源作为辅助的外加激励,另一端采用物理短接的方式;根据线路电源端电流所呈现的波形特征,利用attention机制的深度学习算法进行特征提取,从而判定故障是否为永久性故障,再确定是否闭锁重合闸出口。最后在PSCAD/EMTDC中搭建仿真模型,利用Python API接口实现联合仿真,仿真结果表明,该方法能够在不同的过渡阻抗情况下对不同的故障距离和短路类型具有较强的辨识能力。Direct reclosing mode after a constant delay is adopted for the conventional line reclosing device after line tripping,which has certain disadvantages.In order to avoid the impact on the power grid due to the fact that the circuit breaker is coincident with the permanent fault,a deep learning algorithm based on attention mechanism is proposed to improve the self⁃adaptive judgment of the fault state before the reclosing of the transmission line without the shunt reactor,so as to further improve the accuracy and reliability of the circuit breaker reclosing.After the circuit breakers on both sides of the line are tripped,a conventional power supply is added at one end of the line as an auxiliary external excitation,and the physically short⁃circuited mode is adopted on the other end.According to the waveform features of the current at the power supply end of the line,the deep learning algorithm based on attention mechanism is used for feature extraction,so as to determine whether the fault is permanent or not,and then determine whether to block the reclosing exit.A simulation model is built in PSCAD/EMTDC,and the Python API(application programming interface)is used to realize co⁃simulation.The simulation results show that the method can identify different fault distances and short⁃circuit types under different conditions of transition impedance.

关 键 词：深度学习 attention机制 输电线路 重合闸 故障性质 特征提取 PSCAD/EMTDC 

分 类 号：TN710-34[电子电信—电路与系统]