Chiller Performance Detection And Diagnosis Method Research

With the continuous improvement of people’s economic level and the improvingrequirements of environmental quality, the HVAC(Heating, Ventilating and Air-Conditioning) Engineering system is more widely used in different types ofconstruction and production sectors of life. In the energy consumption of central airconditioning systems, chiller energy consumption is the most important part.Thechiller operating conditions directly affect the energy consumption of air conditioningsystems. Chiller failure will result in reduced performance, increasing of energyconsumption, reduced safety, and affect the thermal comfort of the indoorenvironment. Therefore, it is necessary to chiller performance monitoring, faultdetection and diagnosis, to keep the chiller in a normal and efficient operation of stateand the prevention of the generation of chiller failure.In this paper,a semi-physical model has been used and the water temperature ofthe chilled water, return water temperature of cooling water, unit load rate are selectedas the three regression variables. After data preprocessing, the steady-state andtrouble-free datas are selected to create a centrifugal chiller regression model. Inletand outlet water temperature difference of chilled water,inlet and outlet watertemperature difference of cooling water, lubricating oil temperature, condensingtemperature, condensing approximation temperature are secleted as five characteristicparameters.The reduced evaporator water flow,reduced condenser water flow,excessoil,refrigerant overcharge, refrigerant leakage and condenser fouling are choosed assix fault for fault strategies in this study. After compared the actual measured valuewith the predictive regression model value of each characteristic parameters,it will getresiduals,adaptive thresholds relative sensitivity of each fault index. Relativesensitivity of each fault index to a fault is analyzed so that the most sensitive one isselected and used in the rule-based fault diagnosis.It means the most sensitive onewill have the biggest change when the corresponding fault occur. On the basis faultdetection and diagnosis classifier is formulated for the six fault online fault detectionand diagnosis. Fault detection and diagnosis process include data pretreatment, theregression model, the characteristics of the parameter sensitivity produced and faultdiagnosis classifier four parts.The FDD(Fault Detection and Diagnosis) strategy is validated using laboratory data provided by an ASHRAE research project RP-1043and applied to a real buildingHVAC system in Hong Kong. Finally, the software of chiller performance diagnosis isdeveloped used the FDD strategy on the Visual Basic6.0in this paper.