DefinePK

Abstract

The desired accuracy level of the nonlinear time-current characteristic curve prediction of each overcurrent protective relay can best be obtained from the practical coordination scheme data, applied for the optimal prediction of the relay operation time, rather than empirical data application as mostly seen in most computational applied method for the neural network training and relay operation time prediction. This paper presents a global optimal determination of relay operational parameters settings for time dial setting (TDS) and plug setting (PS) as apply for the time-current characteristic curve prediction for each relay in a propose IEEE 9 bus test system for optimal determination of individual relay response time to fault within its protection zones. A propose hybrid genetic algorithm with artificial neural network (GA-ANN) technique is propose for the prediction of the time-current characteristic curve obtained from global accurate operational parameter settings of each relay to short circuit fault. The GA is applied for global optimal operational parameter determination for each relay by solving a modified objective function (MOF) equation for accurate training data extraction. These valid obtained operation parameters are supplied as training inputs data for the training of ANN to predict accurately the time-current characteristic for each relay. The level of obtained accuracy of the nonlinear time-current characteristic curves will predict accurately the operation time of each relay to different fault current level with minimum mean square errors (MSE) obtained from the applied Levenber-Maequardt algorithm as compared with the obtained outputs from other two applied ANN nonlinear function fitting training algorithm.