Development and Validation of Home Comfort System for Total Performance Deficiency/Fault Detection and Optimal Comfort Control

Residential homes have the potential for substantial energy savings by using advanced home air conditioning systems for fault detection and optimal performance.

The University of Oklahoma developed a learning-based home thermal model that operates a model predictive control-based optimization agent to optimize residential HVAC operation. The learning-based model also contains an automated fault detection and diagnosis agent to detect and diagnose airflow reduction and refrigerant undercharge. Data was connected through smart thermostats, smart meters, and remote-control access of air conditioning units.

During testing of the learning-based model, it was concluded that the model can accurately predict 12-hours-ahead space air temperature and make optimal AC operation decisions that leverage information from the resident’s preferred temperatures, weather forecasts, home thermal condition, and utility time-of-use rate. A self-learning home thermal model that calculates thermal loads and learns the thermal properties of the home is important for optimization.

The fault-detection agent was successful in detecting and diagnosing airflow reduction and incorrect refrigerant charging level when the severity reached a 30% threshold. When both faults were detected at the same time, a second validation method was required to effectively detect airflow reduction.

Learn more about Building America's work.