My research is dedicated to advancing the large-scale application of smart building information systems, particularly at the intersection of artificial intelligence and building HVAC systems. By systematically integrating innovations at the data, platform, and application layers, I explore how to leverage time-series data mining, machine learning, and visualization techniques to address core issues such as building energy efficiency, thermal dynamic modeling.

In the data layer, I developed a portable data collection system to reduce the complexity of deploying AI models, designed a long-term data stream generation model to provide high-fidelity data support across seasons, and proposed a real-time data augmentation model to dynamically optimize thermal dynamic models.

In the platform layer, I built an intelligent HVAC platform that integrates model development, testing, and user interaction. By introducing ML models including large model technology, I enhanced the stability and generalization of prediction models and developed a contrastive curriculum training method tailored for the HVAC domain. Additionally, I established the benchmarking framework to systematically evaluate different models' performance and created LLM-assisted visualization tools to help non-technical personnel quickly generate test code.

In the application layer, my research has been applied in real-world scenarios: automatically modeling highly reliable thermal dynamic models for individual rooms to reduce expertise and experimental studies involved, significantly improving modeling efficiency. I also utilized data generation techniques to assess the potential of buildings in demand response, laying the foundation for large-scale intelligent control.

These efforts form a complete loop from theory to application, providing key technological support for smart building initiatives.