Remote Fault Detection of Building HVAC System Using a Global Optimization Program

Abstract

The energy efficiency of HVAC systems in buildings often degrades with the passage of time. Significant engineering time is normally required to determine the cause(s) of the degradation. This paper reports a preliminary investigation of a remote fault detection method using the global optimization program Solver® (Frontline Systems, 2000) coupled to a simplified simulation program, which is a coding of the ASHRAE 'Simplified Energy Analysis Procedure' (Knebel, 1983). This approach uses the simulation program in conjunction with synthetic measured data to identify faults in the building operation. This fault detection approach has successfully identified all of the faulty parameters with noise levels of 1%, 3% and 6%. It successfully detected 117 of 120 faulty parameters in the presence of 10% noise. These results are based on acceptable errors for each parameter defined as ±5°F for Tcl, ±2°F for Tr, ±0.4cfm/sf for Vs, and ±5% for OA(%). If the acceptable errors for each parameter are reduced to ±2.5°F for Tcl, ±1°F for Tr, ±0.2cfm/sf for Vs, and ±2.5% for OA(%), the number of faults successfully detected decreases to 118 out of 120 faulty parameters with 1% noise, 112 out of 120 with 3% noise, 102 out of 120 with 6% noise, and 94 out of 120 with 10% noise that were introduced into synthetic “measured data” that was generated with the simulation program.