Vol: 1 Issue: 2
FAULT DETECTION AND CLASSIFICATION OF TRANSMISSION LINES USING ARTIFICIAL NEURAL NETWORKS WITH MULTIPLE DATASETS
Vincent Nsed Ogar, Archibong Etim, Akpama Eko James
ABSTRACT
This paper concentrates on classifying and detecting faults in transmission lines, which play a crucial role in transporting electricity from generation to distribution stations. However, these lines are frequently affected by faults caused by human interference, weather conditions, ageing conductors, and the long distances involved in transmission. A transmission line rated at 11/132 kV, 100 MVA, and 50Hz was modeled in MATLAB/SIMULINK to gather data on voltage and current under fault conditions. Training was conducted on 143 fault scenarios using the backpropagation algorithm of an Artificial Neural Network. The individual phases were analysed and subjected to fault detection and classification. 90% of the data was used for training, while validation and testing used 5% each, respectively. 77.6% of the data was ideally classified with a Root Mean Square Error (RMSE) of 0.12348, while 22.4% of the remaining data was in a confusing state. Also, RMSE 0.00415 for fault identification was recorded, and 95% of the data were correctly classified at the fault location zone. At the same time, 5% of the data was in a confused state. The proposed model can facilitate quick and accurate localization and detection of faults on transmission lines, considering their type and severity. This model delivers valid results, is user-friendly, and executes with precision and speed. However, it has limitations, as the output results indicate that fault classification accuracy is poor. Incorporating machine learning algorithms could enhance the accuracy of fault classification.
Keywords: Backpropagation, Artificial Neural Network, fault detection, Fault localisation, fault classification