A top-down strategy with temporal and spatial partition for fault detection and diagnosis of building HVAC systems

This paper presents a top-down strategy for detecting faulty HVAC units across different levels. Energy flow models of HVAC units are developed while the energy is the main feature extracted from the vast amount of real-time data of the HVAC system in a building. A temporal and spatial partition strategy for analyzing HVAC units is proposed. The partition takes into account of architectural, environmental, and human factors. The partition strategy leads to more appropriate thresholds for fault detection. Numerical examples are presented to demonstrate the detection of a systemwide fault and a fault at the VAV level.

► We present a top-down strategy for detecting faults in HVAC systems. ► Energy performance of HVAC units is monitored across different levels. ► PCA is implemented in a unique fashion for pattern recognition. ► A partition strategy is developed to tighten the fault detection threshold.