Digital Overcurrent Relay Implementation With Non-Standard Inverse Curve Modelling Using Adaptive Neuro Fuzzy Inference System

The inverse curve characteristic of overcurrent relay in an industrial power system refers more to the IEC standard. Complex load demands and variation types of relay curve are possible to occur curves intersection. Therefore, in this paper suggest to model an unconventional curve using adaptive neuro fuzzy inference system or ANFIS. Test data made consist of two intersection inverse curve to represent the relay coordination system. New data points of relay are designed by user to obtained the preliminary data. Fault current and tripping time of the relay are used as input and output for ANFIS training. To get the optimum result between data target and design, g-bell with different membership function are implemented in the proposed method. The number of 10-gbell mf obtains more accurate results. The modelling results with a test current of 3.57 amperes obtained a trip time of 5.53 seconds while for prototype testing the trip time is 5.46 seconds, so that between the target modelling results to the prototype relay current testing has better accuracy with the smallest error average value of 3,4,E-03.