Artificial Neural Network For Power Fault Detection And Location | ||||
| The Egyptian International Journal of Engineering Sciences and Technology | ||||
| Articles in Press, Accepted Manuscript, Available Online from 30 June 2025 | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/eijest.2025.372472.1327 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Godswill Omatsone Jemiriayigbe  ; Benjamin Akinloye; Ngozichukwuka Gomenti; Daniel Adadiaha
| ||||
| Electrical and Electronics Engineering Department, Faculty of Engineering, Federal University of Petroleum Resources, Warri, Delta State | ||||
| Abstract | ||||
| This paper presents an artificial neural network approach for fault detection and location in a 200-kilometre, 33 kilovolts power distribution system. The research addresses the growing complexity of modern power grids and the increasing need for adaptive, reliable fault detection methods that can swiftly identify and isolate faults. The methodology involves developing a system capable of accurately classifying fault types and determining their precise locations along the distribution line. The system architecture integrates a measurement subsystem with a neural network trained to recognize fault patterns. Various fault scenarios were simulated to evaluate the system's performance. The fault classification model achieved a cross-entropy loss of 0.9849, indicating a need to further refine this part of the system, while the fault location model attained a mean squared error of 2.826 and a regression value of 0.9996, indicating excellent performance in fault location prediction. Test results demonstrated the system's ability to distinguish between different fault types by analyzing current and voltage profiles, though some instances of misclassification were observed, particularly between similar fault types. The mean margin of error for fault location was approximately 5.72%. The research concludes that while artificial neural network approach-based fault detection systems offer significant advantages over traditional methods in handling non-linear fault scenarios, their effectiveness depends largely on the quality of training data, suggesting opportunities for further optimization of the architecture of the artificial neural network. | ||||
| Keywords | ||||
| Data Preprocessing; Fault Classification; Machine Learning; Neural Network Training; Power Distribution Lines | ||||
|
Statistics Article View: 49 |
||||
