Fault Diagnosis For Refrigerant Leakage Of Variable Refrigerant Flow System Using SEM-PCA-CART Algorithm

As a common HVAC system,VRF(variable refrigerant flow)system has been widely applied in various buildings due to its advantages,such as energy saving,high efficiency,reliable operation,flexible control,low noise and easy installation.However,in the process of actual operation,the operating conditions of the unit are often very complex,which inevitably leads to various typical faults,and the leakage of refrigerant is a typical fault of the VRF system.The refrigerant leakage of VRF system will reduce the unit’s performance of energy saving,user comfort,economy and even safety.Therefore,it is very important to effectively and online detect and diagnose the typical faults of VRF system,including refrigerant leakage.Therefore,the online fault diagnosis for refrigerant leakage of VRF system is mainly studied in this thesis.The experimental system and data sources of VRF system’s refrigerant leakage are introduced,and after the experimental data are preprocessed,13 characteristic variables are extracted.And VRC(virtual refrigerant charge)model is established,used to monitor the refrigerant state of the unit online during the process of VRF system’s refrigerant leakage,and the refrigerant state of the unit is classified,and the experimental data are also optimized.The decision tree model has been constructed,based on the basic principle of CART(classification and regression tree)algorithm,Gini index and system’s impurity gain.And the extracted characteristic variables have been substituted into it,so the fault detection and diagnosis for refrigerant leakage of VRF system has been carried out.The results show that the decision tree model fits the theoretical analysis and the actual situation of VRF system’s refrigerant leakage,and the diagnostic accuracy(94.31%)is relatively high.And after removing the frosting data,the diagnostic accuracy is improved again,reaching 97.39%.The SEM(structural equation model)is introduced into the fault diagnosis for refrigerant leakage of VRF system,and the fault diagnosis model and strategy of refrigerant leakage of VRF system based on SEM-PCA-CART algorithm are established.The experimental data,observed and latent variables,which has been selected,are used to construct the structural equation model,used to analyze the relationship between variables and the relationship between each variable and the refrigerant state of the unit.The results show that the relationship between the variables and the refrigerant state of the unit in the structural equation is consistent with the importance of variables in the decision tree,the theoretical analysis and actual situation of VRF system’s refrigerant leakage,therefore,the structural equation model can be well applied to the study of fault diagnosis for refrigerant leakage of VRF system.Then,the cluster of observed variables,which are closely related to the refrigerant state of the unit in the structural equation model,is substituted into PCA(principal component analysis)to construct the expert variable,and then substituting it into the decision tree model for online fault diagnosis,the diagnosis accuracy reaches 96.96%.After removing the frosting data of the unit,the diagnosis accuracy is again improved to 97.89%.In short,an online fault diagnosis model of VRF system’s refrigerant leakage based on SEM-PCA-CART algorithm,is suggested in this thesis,and it can effectively and online diagnose the refrigerant leakage of VRF system.In addition,the structural equation model is introduced into the research of fault diagnosis for refrigeration and air conditioning system for the first time in this paper.And the SEM can effectively analyze the relationship between variables and the relationship between each variable and the refrigerant state of the unit,and it can select the cluster of observed variables for the construction of expert variables,so as to improve the accuracy of fault diagnosis.