CNN with Attention Mechanism for Accurate Faulty Line Localization in High Impedance Fault Scenarios with Distributed Generators

The global power shortage has led to the widespread deployment of grid-integrated distributed energy resources or generators (DERs). A dependable and continuous power supply requires the protection of power systems, which have grown increasingly complex due to the integration of more dispersed generation. The most crucial aspect of complicated power sys-tem protection is precise and accurate faulty line or section localization. Modern power networks have bidirectional power flow and low magnitude of fault current due to high impedance fault (HIF) with integrated DERs, making conventional fault analysis approaches inappropriate. This study presented a hybrid Convolutional Neural Network with Attention Mechanism (CNN-AM) model that uses end-to-end deep learning for faulty line localization tasks. This automated methodology yields excellent accuracy and enhanced reliability, unlike machine learning, which necessitates preprocessing and human participation. Tests are conducted on various fault-simulated data sets on a grid with three distributed generators built into it. A Doubly Fed Induction Generator (DFIG) wind turbine, Solar Photovoltaic (PV), and Diesel Generator together serve as the distributed generators. The proposed CNN with AM model provides better accuracy in the presence of adversarial vulnerability for faulty line localization. Using our proposed model, 10-fold cross-validation yielded 99.03% accuracy for faulty line Localization. The proposed model compares with various existing deep learning models and performs outstandingly. Real-time data from the OPAL-RT simulator has been used to validate the proposed model, which shows its robustness.