Reinforcement Learning-Based BEMS Architecture for Energy Usage Optimization.

Area of Science:

• Building Science and Engineering
• Artificial Intelligence in Energy Management
• Sustainable Building Technologies

Background:

• Existing intelligent building energy management systems (BEMS) often rely on simple equipment upgrades, limiting innovation and advanced energy savings.
• Stand-alone BEMS in existing buildings do not fully leverage modern Information and Communications Technologies (ICT) for comprehensive energy optimization.
• The need for advanced energy management solutions is critical for achieving sustainable societies worldwide.

Purpose of the Study:

• To propose and evaluate a dynamic heating, ventilation, and air conditioning (HVAC) scheduling method using artificial intelligence (AI).
• To implement an intelligent energy optimization strategy for buildings based on reinforcement learning (RL).
• To demonstrate the application of AI and ICT for enhanced energy saving policies in existing building infrastructures.

Main Methods:

• Collection, analysis, and inference of Internet of Things (IoT) data from a hotel testbed.
• Application of reinforcement learning (RL) algorithms for dynamic HVAC scheduling.
• Development of a purpose-oriented energy-saving methodology driven by AI-based data analysis.

Main Results:

• The proposed method successfully collects and analyzes building energy usage data.
• AI-driven inference of energy usage patterns enables intelligent energy saving policies.
• Reinforcement learning facilitates dynamic HVAC scheduling for optimized energy consumption.

Conclusions:

• The integration of AI, specifically reinforcement learning, significantly enhances BEMS capabilities.
• The developed methodology provides an effective approach for intelligent energy savings in buildings.
• This research contributes to realizing futuristic HVAC systems and achieving energy saving goals through advanced ICT.