AI-Enhanced Inverter Fault and Anomaly Detection System for Distributed Energy Resources in Microgrids

TL;DR

This paper introduces AI-based techniques for detecting and localizing inverter faults in microgrids with high inverter integration, improving fault diagnosis and grid resilience.

Contribution

It presents the FaultNet-ML methodology, a novel AI approach for distinguishing inverter faults from anomalies in power electronics-driven grids.

Findings

FaultNet-ML accurately detects inverter faults in a 9-bus system.

The method effectively differentiates between system anomalies and true faults.

Enhanced grid resilience through improved fault localization.

Abstract

The integration of Distributed Energy Resources (DERs) into power distribution systems has made microgrids foundational to grid modernization. These DERs, connected through power electronic inverters, create power electronics dominated grid architecture, introducing unique challenges for fault detection. While external line faults are widely studied, inverter faults remain a critical yet underexplored issue. This paper proposes various data mining techniques for the effective detection and localization of inverter faults-essential for preventing catastrophic grid failures. Furthermore, the difficulty of differentiating between system anomalies and internal inverter faults within Power Electronics-Driven Grids (PEDGs) is addressed. To enhance grid resilience, this work applies advanced artificial intelligence methods to distinguish anomalies from true internal faults, identifying the specific malfunctioning switch. The proposed FaultNet-ML methodology is validated on a 9-bus system dominated by inverters, illustrating its robustness in a PEDG environment.