Early detection, classification and proactive mitigation of cyberattacks in microgrids embedded with renewables and electric vehicles charging infrastructure

Microgrids are localized energy systems that can operate independently or in conjunction with the main utility grid. Microgrids incorporate renewable energy resources such as solar panels, wind turbines, and increasingly support the deployment of Electric Vehicle Charging Stations (EVCSs). This integration promotes cleaner energy generation and enhances operational efficiency and reliability. However, the deployment of such advanced technologies necessitates robust communication networks to facilitate data exchange among microgrid components, which exposes them to cybersecurity vulnerabilities. Cyberattacks targeting microgrid infrastructure can severely disrupt operations and cause significant technical and economic damage, emphasizing the urgent need for early detection and proactive mitigation strategies against those attacks. This thesis proposes a novel approach for the early detection, accurate classification, and proactive mitigation against cyberattacks. The methodology begins by analyzing electrical signals by employing an advanced signal processing technique known as Continuous Wavelet Transform (CWT), which converts input signals into time–frequency representations called scalograms. These scalograms accurately visualize and highlight sharp transition events, making them an effective tool for capturing anomalies indicative of cyberattacks. These scalograms are further enhanced using an image processing tool called a morphological operation and the Bartlett observation window to suppress irrelevant variations and emphasize discriminative features. Subsequently, the enhanced scalograms are fed into a Convolutional Neural Network (CNN), which is a class of deep learning, to learn the relevant features from these scalograms and classify different types of cyberattacks. The proposed approach achieved a high detection and classification accuracy of 99.53% and 98.80%, respectively. Moreover, the computational time is accelerated by leveraging the parallel processing capabilities of an advanced Graphics Processing Units (GPUs), achieving a 95.85% reduction in training time. Furthermore, the Long Short-Term Memory (LSTM) network is employed to provide valuable insights of attack patterns, thus, enhancing the transparency, and reliability of CNN's decisions. Finally, a novel technique is developed to proactively mitigate the impacts of such attacks on the operation of the microgrid system.

Name thesis:degree_name

Doctor of Philosophy (PhD)

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor

University of Ontario Institute of Technology

Year dc:date.issued

2025

Author dc:creator

• Abu Nassar, Ahmad

Advisor dc:contributor.advisor

• Ibrahim , Walid Morsi

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/10155/1980

OAI identifier oai:identifier

oai:ontariotechu.scholaris.ca:10155/1980