Modeling of Unconventional Overcurrent Relay Curve in the Industrial Power Systems Using Radial Basis Function Neural Network

The standard mathematical model in inverse curve characteristics of overcurrent relay (OCR) refers to the International Electrotechnical Commission (IEC). Inaccurate relay coordination such as reversed curve intersections is possible when using IEC standards. Therefore, in this paper proposes an unconventional OCR curve modeling with a radial system function neural network (RBF Neural Network). RBF is implemented to examine and train data with various number of spreads of actual current as data input and OCR operation time as data output. Coordination of protection of Hess Indonesia Corporation is implemented to make OCR Curve. The R-VCB-09 curve that overlaps with the R-VCB-11 curve is edited to create data points. To obtain the optimal design of OCR curve, the training data is compared with the output of RBF output simulation; From the simulation result, the RBF method using 0.8 spread resulted in minimum error of 1.011%; Thus, the RBF method is very well applied in industrial power systems.