Enhancing Resilient Operation of Distributed Energy Resources Using an Improved Machine Learning-Based IoT Framework Against Cyberattack Perturbation

Abstract

The most realistic way to achieve large-scale integration of distributed energy resources (DERs) into the current grid system is to implement microgrids. These systems function as localized power grids that can operate independently or seamlessly integrate with utility grids. They include DERs, energy storage options, and a variety of loads. Microgrid architecture is changing toward greater distribution, intelligence, and close network integration as communication network technology develops quickly. Microgrids are useful in many fields, including Industry 4.0, smart cities, and the Internet of Things (IoT). However, the connection of the microgrid components and transfer of data via the internet network expose the data to cyber threats, including false data injection, side-channel noise intrusion (SNI), and adversarial attacks, which deteriorate the operation data analysis models such as machine learning (ML). This paper outlines the enhancement of resilient operation using reliable IoT based on reliable ML as a case study for online monitoring of smart inverters. Furthermore, a new ML algorithm is developed using an image processing procedure against cyberattack perturbation. Different levels of attack perturbation are executed using the fast gradient sign method (FGSM) and SNI to validate the robustness of the developed model.