A sensor-fused BIM-based ıntelligent control system for energy-efficient ındoor environmental regulation using deep...

Area of Science:

• Building Science and Engineering
• Artificial Intelligence
• Control Systems

Background:

• The building sector faces increasing energy demands and the need for high indoor environmental quality (IEQ).
• Traditional control methods like rule-based control (RBC) and model predictive control (MPC) have limitations in adaptability and model dependency.
• Intelligent control strategies are crucial for optimizing building performance.

Purpose of the Study:

• To design and implement an intelligent indoor environmental control system integrating multi-sensor fusion and Building Information Modeling (BIM).
• To develop an energy-efficient dynamic environmental control strategy using a deep fusion actor-critic reinforcement learning (DACRL) algorithm.
• To evaluate the system's performance in reducing energy consumption and enhancing IEQ across diverse building types.

Main Methods:

• A four-layer cloud-edge-end collaborative architecture: perception, fusion, decision, and execution layers.
• Multi-modal sensor network for distributed data collection (temperature, humidity, CO2, light, occupancy).
• Spatio-temporal data fusion using sliding window dynamic weighting and extended Kalman filter (EKF), with anomaly detection via isolation forest and one-class SVM.
• Deep reinforcement learning control model incorporating the Lieb-Thirring (L-T) inequality as a spectral constraint within the Actor-Critic algorithm.
• Real-time closed-loop control of HVAC, lighting, and shading via BMS protocol.

Main Results:

• Annual energy consumption per unit area reduced by up to 63.2% compared to RBC and 23.5% compared to MPC.
• Significant improvements in IEQ metrics: CO2 concentrations below 800 ppm for 91.5% of the time, and thermal comfort (PMV) within ±0.3 for 94.2% of the time.
• DACRL algorithm demonstrated stable convergence in ~15,000 training steps with only 4.9% performance loss in cross-building transfer testing.

Conclusions:

• The proposed intelligent control system effectively balances energy efficiency and IEQ in buildings.
• The DACRL algorithm, enhanced with the L-T inequality spectral constraint, offers robust and generalizable control performance.
• The integration of multi-sensor fusion, BIM, and advanced AI presents a promising direction for smart building management.