Power transformers are major power system equipment. Transformer failure can cause disconnected to power system. This project proposed an application of artificial neural network (ANN) in transformer failure detection system. The fault detection is based on dissolved gas in oil analysis (DGA). DGA diagnose five different dissolved gases (H2, CH4, C2H2, C2H4, and C2H6) to get result. The result maybe normal, arching, discharge or overheat. Several neural networks will be developed and analyzed to find the most efficient network that will be use to model the transformer failure detection system in this project. The networks are Multilayer Peceptron network (MLP), Radial Basis Function network (RBF) and Hybrid Multilayer Perceptron network (HMLP). All this networks will be train using different training algorithm. This study proves that the HMLP network trained with MRPE algorithm has the best result. HMLP accuracy is 95.34%. MLP network that trained with BP algorithm obtain accuracy of 54.05%. Accuracy for second MLP network that trained with LM algorithm is 92.74% and accuracy for last MLP network that trained with BR algorithm is 92.38%. RBF network that trained with k-Means Clustering and GLS algorithm has accuracy of 89.41%. The final product of this project is a software system to detecting transformer failure. This system can be used as an inductive, exploratory, accurate and analytical tool for transformer failure detection
SISTEM PENGESANAN KEROSAKAN TRANSFORMER BERASASKAN RANGKAIAN NEURAL BUATAN_WAN MOHD FAHMI WAN MAMAT_E3_2007_875002554_NI