Repair and maintenance in power distribution is an important factor that
affects the continuous productivity services and power efficiency in electrical
supply systems. Thermographic inspection has been often used as a
maintenance tool, as it allows detection of early-stage failure from the system
in electrical distribution. Failure in the system can lead to catastrophic
failure like a high-voltage arc fault. The presence of fault is caused by the
higher temperature of the instrument that leads to the formation of hotspots.
The use of infrared inspection is useful in detecting the hotspot that is hardly
noticeable. It helps to overcome the problems that arise during operation
and maintenance in the distribution systems. In this research, a fault
detection system is proposed with the application of Artificial Neural
Network (ANN) in identifying faults on electrical equipment. This method
was trained by using the temperature parameter on the IR images taken from
TNB Distribution. As a result, it will lead to faults detection. Thus, the
purpose of this project is to ensure the correct recommendation of corrective
actions in the maintenance procedure of the electrical system. The actions to
the detection of faults taken are based on the results of the temperature
measured. The neural network training performance for the temperature of
hotspot detection was developed with a minimum error of 0.00084165 MSE
at epoch 39. The study shows the best-fitting allows detection of early-stage
failure. It can be concluded that the current method in conducting the
prediction process by using Thermographic inspection is suitable for
electrical equipment based on the training result.