Thesis

Towards trustworthy AI systems for Smart Grid management : facilitating robustness, transparency and fairness in the energy transition

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2025
Thesis identifier
  • T17374
Person Identifier (Local)
  • 202178274
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • The integration of Artificial Intelligence (AI) into Smart Grid (SG) management presents significant opportunities for enhancing energy efficiency, reliability, and sustainability. However, deploying AI in critical energy infrastructure raises challenges related to robustness, transparency, and fairness, essential components of trustworthy AI. This thesis addresses these challenges by proposing novel methodologies and frameworks to enhance trustworthiness in AI-driven SG management. First, it introduces a quantitative framework for evaluating and visualizing explainability in deep learning based Non-intrusive Load Monitoring (NILM) systems. Next, it presents a new training enhancement approach that incorporates explainability principles directly into training of NILM models, achieving improvements in interpretability and predictive performance. Recognizing the constraints of deploying complex AI models on edge devices, the thesis proposes an explainability guided knowledge distillation framework that balances model efficiency with interpretability and reliability, facilitating robust edge deployment without compromising performance. Finally, it addresses equity concerns in Electric Vehicle Charging Station (EVCS) infrastructure placement by developing a geodemographic-aware placement strategy using Graph Neural Networks (GNNs), ensuring equitable access across diverse socioeconomic groups. Collectively, these contributions establish a comprehensive approach to embedding robustness, transparency, and fairness into AI applications within the SG context, aligning technical innovation with ethical and societal imperatives. This work supports broader adoption of trustworthy AI, contributing significantly to sustainable development and equitable energy transition objectives.
Advisor / supervisor
  • Stankovic, Vladimir
  • Stankovic, Lina
Resource Type
DOI
Funder

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T17374 2025-07-08 Public Download