Validation of Machine Learning-Aided and Power Line Communication-Based Cable Monitoring Using Measurement Data.

Area of Science:

• Electrical Engineering
• Data Science
• Materials Science

Background:

• Smart electricity grids leverage power line communications (PLC) for real-time cable monitoring.
• Machine learning (ML) models show promise for ML-aided cable diagnostics.
• Existing research often relies on synthetic data, limiting real-world applicability.

Purpose of the Study:

• To validate ML-aided cable diagnostics using measured PLC channel data.
• To develop an anomaly detection system for power cables.
• To assess the effectiveness of clustering and principal component analysis (PCA) for cable health monitoring.

Main Methods:

• Integrated measured PLC channel data with ML models.
• Employed clustering for data pre-processing.
• Utilized principal component analysis (PCA) for dimension reduction and anomaly detection.
• Trained models on healthy network data and tested on degraded cable conditions.

Main Results:

• The proposed anomaly detection scheme achieved high accuracy in identifying cable degradation.
• The system demonstrated a low false alarm rate.
• Validation was performed using an experimental setup with introduced power cable degradations.

Conclusions:

• ML-aided diagnostics using measured PLC data are effective for cable anomaly detection.
• The combination of clustering and PCA provides a robust method for real-time cable monitoring.
• The developed system offers a reliable solution for enhancing power grid safety and maintenance.