Cognitive Digital Twins: Integrating Deep Learning Models for Proactive Facilities Management in Smart Buildings

Digital twins (DTs) are revolutionizing the AECO industry by offering real-time digital replicas of physical assets. However, the integration of continuous, automated predictive forecasts within DTs remains limited. A framework that integrates AI and ML technologies with operational DTs was developed and implemented, enabling predictive capabilities. Using a Long Short-Term Memory neural network, daily energy consumption was predicted based on historical energy usage and real-time weather forecasts. The model achieved an MAE of 178.79 kWh, a root RMSE of 236.02 kWh, and an MAPE of 2.98%, demonstrating its ability to generalize while capturing seasonal and operational trends. By functioning as predictive systems, DTs transform into comprehensive common data environments (CDEs) that aggregate and contextualize data, empowering stakeholders with actionable insights. This framework not only demonstrates practical applicability but also lays the foundation for cognitive digital twins (CDTs), enabling them to evolve into intelligent, adaptive systems for smarter, more sustainable building operations.