Machine-Learning-Based Adaptive Settings of Directional Overcurrent Relays With Double-Inverse Characteristics for Stable Operation of Microgrids

Abstract

Microgrids (MGs) with distributed energy resources (DERs) provide significant benefits in terms of energy efficiency and sustainability. However, they bring challenges to protection schemes, particularly relay settings and coordination. This article investigates the deployment of directional overcurrent relays (DOCRs) in MGs. Given the limited inertia of DERs and the potential instability resulting from extended DOCR operating times postfault, a novel DOCR setting is proposed. This setting uses shifted user-defined characteristics that integrate two inverse curves to ensure relay coordination and MG stability. Meanwhile, recognizing that MGs can operate in various topologies, a single DOCR setting may prove ineffective for many scenarios. Therefore, this article configures DOCRs with adaptive settings to manage diverse operating conditions. Due to the limited number of settings supported by commercial DOCRs, a self-organizing map is used to categorize MG potential scenarios into coherent groups aligned with available DOCR settings. The stability-constrained settings of each DOCR are optimized using the genetic algorithm and then stored within the relay for seamless activation when needed. The efficacy of the proposed approach is evaluated on a modified IEEE 33-bus system with synchronous and inverter-based DERs using DigSILENT and MATLAB.