AI Agent-based Intelligent Digital Twins for Building Operation

Ontology-enabled AI agent-driven intelligent digital twins for building operations and maintenance

This study proposes a method for implementing an Intelligent Digital Twin (IDT) for autonomous building operations and maintenance (O&M) by leveraging AI agents and an ontology-based framework. The proposed IDT model incorporates interactions between the existing five-dimensional DT model and AI agents, defining the roles of DT administrators—who construct, operate, and manage the DT model through AI agents—and DT users—who utilize digital twin services to achieve efficient building operations and maintenance. In particular, the role of the DT administrator is divided into (1) establishing the basic DT environment and (2) executing a Knowledge Engineering process, which enables AI agents to systematically learn and apply domain knowledge, emphasizing the importance of providing ontologies that represent the target system. In a case study applied to an actual building HVAC system, the AI agent demonstrated the ability to analyze ontologies and operational data in response to user requests, autonomously generate highly accurate virtual models, detect sensor faults using a fault detection and diagnosis (FDD) algorithm, visualize and summarize the results into a Word document, and update ontologies and system information in real time. The proposed methodology shows that advanced digital twin operations and continuous expansion and synchronization can be achieved with minimal human intervention, even in complex building systems.