Identification of distributed optical cable external breakage events based on CNN-LSTM

To ensure the stable operation of the urban power optical cable network, highly reliable and accurate fault diagnosis technologies are essential. We propose a high-sensitivity monitoring method for external disruption events utilizing a combination of auto heterodyne beat frequency interference and CNN-LSTM. This method outlines the processing flow of multi-point input signals in the CNN-LSTM algorithm. Comparison of test error distributions reveals that in comparison to the LSTM algorithm, the CNN-LSTM algorithm generally exhibits smaller prediction errors, enhanced stability, and greater reliability. Based on event urgency, it can deliver effective early warnings and alarms for operation and maintenance personnel to ensure the safety of optical cable lines.