Smart buildings: Federated learning-driven secure, transparent and smart energy management system using XAI

Abstract

In modern smart grids and decentralized systems, smart buildings face several key energy management challenges, including data privacy concerns, the need for accurate real-time decisions, the complexity of managing Distributed Energy Resources (DERs), and the lack of transparency in Artificial Intelligence (AI) systems, which erodes user trust. Traditional energy management systems rely on centralized data gathering and processing, where energy data from various sources is accumulated and processed in one location. While centralization aids in decision-making regarding energy distribution, it also raises concerns about data privacy, cybersecurity, and the opaque nature of AI decisions, all of which undermine user confidence. To address these issues, Federated Learning (FL) and Explainable Artificial Intelligence (XAI) offer promising solutions. FL decentralizes model training, enhancing data privacy and security, while XAI provides clear explanations of AI decisions, fostering user trust. When combined, FL and XAI create a secure, transparent, and interpretable framework for managing energy in smart buildings. This paper proposes an FL-driven XAI model that aims to improve data privacy, accelerate real-time decision-making, enhance efficiency, and increase transparency, thereby building user trust. The proposed model demonstrates superior performance in simulations compared to previously published approaches.