Abstract
<jats:p>Due to variability of short circuit current magnitude and their direction in presence of distributed generators into distribution network, conventional protection scheme based directional overcurrent relays (DOCR) may exhibit mal operation in term of selectivity, coordination time interval (CTI) and operating time (OT). This paper aims to determine a suitable adaptive protection scheme (APS) based DOCR settings using two processes: 1- detection of fault, 2- predetermined setting selection. A machine learning based on support vector machine ensures the fault detection, while a Particle Swarm Optimization (PSO) is employed to determine the optimal settings i.e., time multipliers setting (TMS) and pickup current (Ip) for all operating conditions (OC). In each scenario, a load flow calculation, 4 types-based fault analysis, generation of feasible solution, and optimal settings are carried out to bult a database. The pretrained SVMs with 82.3 % and 98% of accuracies are implemented into the relays as well as the offline setting during the configuration such that to perform real time setting selection without the need of continuous online communication with the control center. A comparison analysis for different feeders shows that the proposed APS is able to correct the maloperation faced by the Convention Protection Scheme (CPS)</jats:p>