Fault Identification Method for Measured Travelling Wave of Transmission Line Based on CSCRFAM-Transformer

The ability to accurately classify fault type within traveling wave data is crucial for real-time online fault location and protection using traveling wave technology. However, the current common practice in the power industry relies on manual data screening followed by offline processing, leading to several limitations such as poor timeliness, low accuracy, and high skill requirements for operators. These drawbacks restrict the application of traveling wave acquisition devices. To address these issues, this paper proposes a fault identification method for measuring the traveling wave of transmission lines based on the CSCRFAM-Transformer. Firstly, CSCRFAM is used to encode the temporal and spatial information of the measured traveling wave data. Next, pixel-level features are further aggregated through dimensional interaction. Then, an adaptive encoding hierarchy Transformer adjustment mechanism is employed to extract multi-level differentiated traveling wave high-frequency information from the aggregated features to complete fault identification. This method combines the dimensional interaction of the EMA mechanism and the self-attention mechanism of the Transformer's sensitivity to the spatiotemporal characteristics of traveling waves. The proposed method is trained and tested using a massive dataset of 396 672 measured samples from 110 kV to 220 kV transmission lines in Yunnan Power Grid. The method is used to identify, classify, test, and compare four distinct types of traveling wave data. The obtained results show that the method reduces the number of model parameters and improves the identification accuracy. The mAUC, Accuracy, Precision, and F1 values of the algorithm reach 0.969, 0.969, 0.965, and 0.957, respectively, indicating better detection accuracy and identification efficiency.