基于深度学习的特高压三端混合直流输电线路波形特征故障区域判别方法OACSTPCD

When the existing DC line fault zone identification methods are applied to three-terminal hybrid UHVDC transmission lines,there exist problems such as difficulty in distinguishing faults on both sides of the T-zone,weak ca-pability to endure the transition resistance,and difficulty in threshold setting.Aimed at these problems,a method is proposed to identify the fault zones of three-terminal hybrid UHVDC transmission lines by using deep learning and wave-form characteristics.First,the fault characterization of different fault zones of three-terminal hybrid DC lines is carried out.Second,a multi-scale wavelet decomposition of line mode voltage and line mode current is carried out to extract the low-and medium-frequency components of line mode current and high-frequency components of line mode voltage.These components form the input to the deep learning model by combining the waveform characteristics of positive and negative voltage,while the fault zone is taken as the output,the deep learning faulty region recognition model is con-structed.Third,the acquired fault characteristics are processed by the trained deep learning model to achieve the fault zone identification.Through a lot of simulations,it is verified that the proposed fault zone identification method has a high accuracy and is basically unaffected by the transition resistance.