Improvement on optimal coordination of directional overcurrent relays in mesh distribution network system using artificial intelligence technique

The electrical meshed distribution network (MDN) protection coordination scheme,poses great challenges to protection coordination scheme setup, due to the network topology structure. This always resulted to unexpected miscoordination among selected primary and backup relay pairs due to multi-directional fault current infeed’s contributions by all interconnected electrical power sources to the short circuit fault current magnitude level. Moreover, other challenges to be addressed is in the ineffective prediction of the nonlinear time-current characteristic function curve from empirical data as earlier proposed in previous research, for the future determination of the relay operation time response to short circuit fault in other locations. This research work propose the artificial intelligent (AI) solution on the conventional objective function (COF) and the modified objective function (MOF) formulation, with the application of genetic algorithm (GA) optimization solver, to determine each relay best optimal operation parameters selection for the time dial settings (TDS), plug setting (PS) and response time to fault accordingly. Also, the elimination of pending miscoordination amongst relay pair for effective coordination scheme. Furthermore, a novel hybrid GAANN technique is proposed for the supervised training, to predict the nonlinear timecurrent characteristic function fitting of each relay operation time function. A directional overcurrent relay (DOCR) coordination in IEEE 9 bus test system is proposed for this research work with three integrated multi distribution generation electrical power sources (DG) in DigSiLent power factory and Matlab Simulink software. The obtained result from the GA solution of the MOF produced a 91.67% improvement in the obtained optimal parameter values against the 8.33% reduced value from COF. This also translated into the same percentage values in operation time response to fault within each relay protection coverage zones. Furthermore, the pending miscoordination amongst selected relay pairs of 16.67% earlier experienced in COF solution is been eliminated by the GA solution of the MOF with 100% elimination between the selected primary and backup relay pairs. This is substantiated by the lower fitness mean value of 1.3358 from MOF against the 4.7679 from the COF for the same minimization problem. However,the Levenberg–Marquardt nonlinear function fitting algorithm application on solving the novel hybrid GA-ANN technique predicted the nonlinear time –current function fitting of each relay effectively with minimum mean square error (mse) between the target output and the actual output for effective generalization during supervised training of the network. This research work has achieved all proposed objective function by improving and eliminating all pending problem encountered in multi sources MDN