AI-driven explainable digital twin with adaptive decision support for multi-zone smart buildings

Smart buildings generate enormous amounts of sensor data, yet most digital twin systems still leave facility managers in the dark—offering predictions with no explanation of why. This paper introduces the X-DT (Explainable Digital Twin), a framework built for multi-zone smart buildings that not only forecasts zone-level energy consumption but also explains the reasoning behind every prediction in plain language. This work directly supports UN Sustainable Development Goal 7 (Affordable and Clean Energy) and SDG 11 (Sustainable Cities and Communities) by enabling data-driven, transparent energy management in smart buildings. The system connects live IoT data streams to an AI engine that identifies which factors—occupancy shifts, HVAC load, equipment usage—are driving energy behaviour at any given moment. Role-specific dashboards, what-if scenario simulation, and a full audit trail are also built in. Tested on two years of real building data from a five-zone commercial building in Bangkok, the X-DT’s XAI-guided decisions achieved 10.9% mean energy savings compared to 3.9% under a standard rule-based approach (p = 0.015, Cohen’s d = 0.46)—showing that when managers understand why the system recommends something, they make meaningfully better decisions.