HVAC Fault Detection Benchmark Data from Seven Building Systems

Seven datasets provide operational time-series data for evaluating Fault Detection and Diagnostics algorithms on HVAC equipment. The collection includes simulation, lab, and field measurement data for rooftop units, air handlers, VAV boxes, fan coils, chillers, and boilers. It was created by a consortium of national laboratories and universities including LBNL, PNNL, NREL, ORNL, and Drexel.

Use Cases

• Benchmark FDD algorithm accuracy using time-series data for faults at different severity levels across seven HVAC system types.
• Compare fault-free operational data against fault-injected cases for equipment like chiller plants and boiler plants.
• Visualize building operational data according to the BRICK schema using the provided TTL files.

Strengths

• Includes data from seven distinct HVAC systems/equipment types, enabling broad algorithm testing.
• Combines data from three sources: simulation, laboratory experiments, and field measurements from real buildings.
• Each dataset is documented with a PDF file detailing key information for understanding content and scope.

Limitations

• The specific number of data rows, columns, and file sizes are not provided, limiting precise assessment of scale.
• Data is provided in multiple CSV files per dataset, requiring integration and parsing before analysis.

Provenance

Source

LBNL, PNNL, NREL, ORNL, and Drexel University.

Collection Method

Operational data gathered from simulation, laboratory experiments, and field measurements from real buildings.

Freshness

Last updated in February 2024.