Abstract
This study presents an integrated approach for detecting inter-turn faults in air-core dry-type shunt reactors. It combines finite-element modeling with a data-driven protection scheme that utilizes the eXtreme Gradient Boosting (XGBoost) algorithm. The method was applied to a 230 kV air-core reactor installed at the Penedo bus of the Nossa Senhora do Socorro-Penedo transmission line in Brazil. Finite-element analysis was used to determine the reactor's self and mutual inductances under various inter-turn fault conditions. These analyses served as the foundation for time-domain transient simulations. Fault signatures were generated using MATLAB, while external events, such as transmission line faults, busbar faults, reclosing operations, nearby equipment energization, line oscillations, and voltage unbalances, were simulated in ATP to enhance the model's security. The XGBoost-based protection scheme was trained and tested with both internal and external cases, including scenarios with noise, to evaluate its robustness. Results demonstrated high accuracy, dependability, and security, with most faults detected within two cycles. This proposed method effectively identifies inter-turn faults, even in situations where conventional protection schemes exhibit limited sensitivity.
Funding Information
  • Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, Brasil (001)
  • Fundação de Amparo à Pesquisa do Estado de São Paulo (2024/03306-8)

This publication has 17 references indexed in Scilit: