Abstract
The increasing penetration of inverter-based resources (IBR) in medium voltage (MV) networks presents significant challenges for traditional overcurrent (OC) protection systems, particularly in ensuring selectivity, reliability, and fault isolation. This paper presents an adaptive protection system (APS) that dynamically adjusts protection settings based on real-time network conditions, addressing the challenges posed by distributed energy resources (DER). The methodology builds on ongoing research and development efforts, combining an offline phase, where operational scenarios are simulated using historical data, clustered with fuzzy c-means (FCM), and optimized with evolutionary particle swarm optimization (EPSO), and an online phase. To overcome the static nature of conventional schemes, a machine learning (ML)-based classifier is integrated into the APS, enabling real-time adaptation of protection settings. In the online phase, a centralized substation protection controller (CPC) leverages real-time measurements, communicated via IEC 61850 standard protocols, to classify network conditions using a support vector machine (SVM) classifier and activate the appropriate protection settings. The proposed APS has been validated on a Hardware-in-the-Loop (HIL) platform, demonstrating significant improvements in fault detection times, selectivity, and reliability compared to traditional OC protection systems. As part of a continued effort to refine and expand the system's capabilities, this work highlights the potential of integrating artificial intelligence (AI) and real-time/online decision-making to enhance the adaptability and robustness of MV network protection in scenarios with high DER penetration. © 2025 Elsevier B.V., All rights reserved.