Ensemble learning based transmission line fault classification using phasor measurement unit (PMU) data with...

Area of Science:

• Electrical Engineering
• Power Systems Analysis
• Machine Learning Applications

Background:

• Phasor Measurement Unit (PMU) data offers new insights into power network states for transmission line fault studies.
• Traditional voltage and current magnitudes are insufficient for precise fault analysis, necessitating additional parameters.

Purpose of the Study:

• To develop and evaluate machine learning models for accurate transmission line fault classification using comprehensive PMU data.
• To investigate the effectiveness of ensemble learning and explainable AI (XAI) in improving fault detection and interpretability.

Main Methods:

• Generated synthetic transmission line data including voltage, current, frequency, and phase angles using ePMU DSA tools and Matlab Simulink.
• Trained and compared individual machine learning models (Decision Tree, Random Forest, K-NN) and an ensemble model using soft voting.
• Applied explainable AI (XAI) to interpret model predictions and validated performance on the IEEE 14 bus system.

Main Results:

• Individual models achieved high cross-validation accuracies (DT: 99.84%, RF: 99.83%, K-NN: 99.76%).
• The ensemble model demonstrated superior performance with a 99.88% cross-validation accuracy.
• Explainable AI provided insights into the influence of different parameters on fault classification.

Conclusions:

• The developed ensemble model effectively classifies transmission line faults with high accuracy.
• Comprehensive PMU data and advanced machine learning techniques significantly improve fault detection capabilities.
• Explainable AI enhances the understanding and trustworthiness of the fault classification models.