Self Classification of Multifunction Relay Based on Neural Network for Industrial Scale

As time goes by, the changing of the electrical system must be balanced with the development of a protection system. It is ranging from simple with limited capabilities and features to complex systems that have better capabilities in terms of selectivity and operation. A good protection system must be able to secure and minimize disturbance quickly and precisely so undesirable things such as equipment damage, blackout, and danger to humans are not happens. Moreover, the incident of mal-tripping from the protection system must be avoided. For example, an incident that occurs mal-tripping of overcurrent relays in the industrial system is the starting of an induction motor or energizing transformer. The transient current that appears from the starting motor itself will trigger a relay to work. The surge in the motor start is only temporary, so the relay should not be able to work during the starting period. Therefore it can be said that at this time, relay works only based on the settings of the user and has not been able to classify disturbance or not. This research proposes a method for detecting and classifying disturbance using Neural Network methods based on time series data and conducted in a radial system, which is usually used in industrial scale. Where the time series data as a neural network input consists of pre and on fault data. Then, the results of the neural network output will be applied to the modeling of multifunction relays in Simulink.