CNN-Based Fault Phase Identification Method of Double Circuit Transmission Lines

In order to address the problems that model-based distance protection cannot correctly identify the fault phases in the DCTL (Double Circuit Transmission Lines on the same tower), a CNN-based (Convolutional Neural Network) fault phase identification method is proposed. The proposed method only uses the current and voltage waveforms of single-ended and single-circuit of DCTL to identify the fault phases. The CNN includes two convolution layers and two pooling layers, which are used to automatically extract the features of the waveforms. Furthermore, the CNN realizes the phase identification function by the extracted features. A training set is constructed by the simulation result of PSCAD, which is used for training the CNN to fit the mapping relationship between the waveforms and the fault types. Then, a validation set is used to optimize the hyper-parameters of CNN. Finally, several test sets are used to test the performance of CNN for phase identification. Compared with the traditional model-driven methods, the CNN-based method avoids manually analyzing the fault characteristics under different fault types and constructing identification criteria. The test results show that the proposed method can greatly improve the fault phase identification accuracy of DCTL.