The Study Of Fault Detection Method For HVAC System Based On Hardware-in-loop Simulation Platform

In order to save energy, improve indoor air quality and ensure ahealthy running status of heating, ventilation and air conditioning (HVAC)systems, various of fault detection and diagnosis (FDD) strategies havebeen embedded into the Building Automation System. At the same time,more and more FDD strategies are proposed by researchers. These FDDstrategies can be divided into several categories: the model-based FDDmethod, the knowledge-based method and the date-driven based method.The validations of these FDD strategies are mainly conducted throughdigital simulation platform and real HVAC system. It’s convenient tovalidate FDD strategies through digital simulation platform, but the FDDstrategies validated through digital simulation platform may be inefficientin practical application. The FDD strategies validated through real HVACsystem will be effective, but the process of validation is time consuming.In this study, a hardware-in-loop simulation platform is proposed andapplied to the validation of FDD strategies. The hardware-in-loopsimulation platform combines the digital simulation platform and the realcontroller together. It’s also convenient to validate FDD strategies throughhardware-in-loop simulation platform, and the validation result is practical.A comparison between PCA based FDD method and BPNN basedFDD method is conducted in this study. The result shows that for both of these two FDD methods, the fault detection efficiency withhardware-in-loop simulation platform is lower than the value with digitalsimulation platform. It exemplified that the FDD methods, which havebeen validated with digital simulation platform, still need to be improvedin the real application. The hardware-in-loop simulation platform, which ismore close to the real HVAC system, contributes to the development ofreliable FDD strategies.On the basis of aforementioned comparison study, improvementapproaches are made for the PCA based FDD method and BPNN basedFDD method. The fault detection method based on wavelet transform andPCA principles and the fault detection method based on reconstructedBPNN are proposed. These two methods are validated withhardware-in-loop simulation platform. The validation results show that thewavelet transform can eliminate the effects of noise signal in HVACsystem, while the reconstructed BPNN can improve the predictionaccuracy of neural network.