Cross-Bonded Cable Circuits Identification Based on Deep Embedded Clustering of Sheath Current Sensing

Area of Science:

• Electrical Engineering
• Power Systems Analysis
• Signal Processing

Background:

• Online identification of high-voltage (HV) cable circuits is crucial for effective inspection and maintenance.
• Existing offline methods, such as passive electromagnetic wave injection, are insufficient for real-time circuit identification.
• A gap exists in methodologies for online identification of cross-bonded HV cable circuits.

Purpose of the Study:

• To propose and validate an online circuit identification methodology for HV cross-bonded cable circuits.
• To leverage sheath current temporal characteristics and deep embedded clustering for accurate circuit determination.
• To address the limitations of offline identification methods in HV power systems.

Main Methods:

• Development of an equivalent circuit model for multi-circuit cross-bonded cable sheaths to analyze sheath current temporal similarity.
• Verification of temporal similarity robustness using Dynamic Time Warping (DTW) distance simulations.
• Implementation of a combined Temporal Convolutional Network Autoencoder (TCN-AE) and K-medoids model for temporal clustering of sheath currents.

Main Results:

• The proposed method achieved a high identification accuracy of 95.37% on a full-scale 110 kV cable test platform.
• Demonstrated robustness against domain gap variations, ensuring reliable performance under different operating conditions.
• Successfully transformed circuit identification into a temporal clustering problem using synchronized monitoring of multi-circuit sheath currents.

Conclusions:

• The developed online circuit identification methodology effectively determines the circuits of cross-bonded HV cables.
• The approach offers significant engineering application value due to its high accuracy and robustness.
• This method provides a viable solution for real-time monitoring and maintenance of HV power cable systems.