Automated commissioning of HVAC systems using first principle models
Commissioning of HVAC systems has potential for significant improvements in occupant satisfaction, comfort and energy consumption, but is very labour-intensive and expensive as practiced at this time. Previous investigators have capitalized on digital control systems' capability of logging and storing data and of interfacing with external computers for open loop control by developing methods of automated fault detection and diagnosis during normal operation. Some investigators have also considered the application of this technique in commissioning. This thesis investigates the possibility of utilizing first principles and empirical models of air-handling unit components to represent correct operation of the unit during commissioning. The models have parameters whose values can be determined from engineering design intent information contained in the construction drawings and other data available at commissioning time. Quasi-dynamic models are developed and tested. The models are tested against design intent information and also against data from a real system operating without known faults. The results show the models agree well with the measured data except for some false positive indications, particularly in the damper and fan models, during transients. A procedure for estimating uncertainty in the instrumentation and the models is developed. The models are also tested against artificial faults and are able to detect all of the faults. Methods of diagnosing the faults are discussed.