FaultXformer: A Transformer-Encoder Based Fault Classification and Location Identification model in PMU-Integrated Active Electrical Distribution System

Researchers developed FaultXformer, a Transformer-based AI model that achieves 98.76% accuracy in fault classification and 98.92% accuracy in fault location identification in electrical distribution systems using PMU data. The dual-stage architecture significantly outperforms traditional deep learning methods like CNN, RNN, and LSTM, particularly in systems with distributed energy resources integration.

• →FaultXformer uses a dual-stage Transformer encoder architecture to analyze real-time PMU current data for fault detection in electrical grids.
• →The model achieved 98.76% accuracy in fault classification and 98.92% in location identification, outperforming CNN, RNN, and LSTM baselines.
• →Testing was conducted on IEEE 13-node test feeder with 20 fault locations and multiple distributed energy resource integration scenarios.
• →The approach uses stratified 10-fold cross-validation and data from four strategically placed PMUs for robust performance evaluation.
• →Performance improvements were particularly notable in systems with high distributed energy resource penetration, addressing modern grid complexity.