Fault Location and Fault Cause Identification Method for Transmission Lines Based on Pose Normalized Multioutput Convolutional Nets

Accurate identifying of fault causes and locating fault sites are fundamental requirements to ensure the safe operation of the power grid. Compared with extra-high-voltage transmission lines, digital fault recorders (DFRs) with high sampling rates and information synchronization are rarely applied in high-voltage transmission lines. One-ended DFRs and impedance-based methods are commonly used for fault identification and location. However, the impedance-based fault location and fault cause identification methods have coupling effect due to the fault resistance caused by different causes, which can affect the location accuracy. To address this, a fault cause identification and location method based on pose normalized multioutput convolutional nets (PNMCN) is proposed. First, the fault characteristics displayed by the volt-ampere curve are analyzed based on the fault mechanism. Second, a one-ended volt-ampere curve is used as an input, pose normalization is employed to improve feature extraction for fault resistance, and coupled prior knowledge of fault cause and fault location is used during training of PNMCN to mitigate their mutual effects. Finally, the results of the case study show that the median and mean relative errors of the fault location are 0.68% and 1.16%, respectively, and the fault cause identification accuracy is 99.5% when using one-ended low-sampling frequency data. The proposed method can be used for adaptive reclosing and is convenient for operation and maintenance personnel to detect faults.