Employing Federated Learning for Training Autonomous HVAC Systems

2024-05-01

Energy and Buildings

AI Summary

To address the challenges of data scarcity, poor generalization, and privacy constraints hindering efficient training of autonomous HVAC systems in real-world buildings, this paper pioneers the integration of federated learning into HVAC reinforcement control. We propose a lightweight, distributed training framework for multi-building collaboration, integrating Proximal Policy Optimization (PPO), LSTM-based state modeling, and an edge computing architecture. The framework features an adaptive model aggregation mechanism and a differential privacy protection strategy, ensuring data locality and privacy compliance while enhancing convergence speed and communication efficiency. Extensive experiments across six real-world building datasets demonstrate that our approach reduces energy consumption by 12.3% compared to baseline methods, decreases cross-building generalization error by 37%, and cuts communication overhead by 58%.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Search and Optimization: Learning to Search
• Multiagent Systems: Multiagent Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Improving data efficiency in HVAC reinforcement learning controllers
• Enhancing generalization across diverse building environments
• Accelerating learning speed for energy-efficient HVAC policies

Innovation

• Federated Learning enhances HVAC reinforcement learning
• Global policy aggregates local climate zone policies
• Improves learning speed and generalization stability