The Simulation Research Of Subway Station Chiller Fanlt Detection And Diagnosis

In order to build a safe and comfortable bus and work environment for subway station, the ventilation and air conditioning system is introduced in the station, and for which the chiller is to provide cooling source. In the daily running of the chiller, there will be various degrees of faults, if not solved timely, will cause the setting parameters of the ventilation and air conditioning system deviate from the preset value. This will bring discomfort to passengers, and increase the subway station of energy consumption, shorten the service life of the equipment also. So, how to ensure the efficient operation of the chiller has become a research hotspot in the field.The chiller of subway station is the research object of this thesis. According to the process data has non-Gaussian and non-Linear characteristics, a fault detection and diagnosis(FDD) method based on Indpendent Component Analysis(ICA) and Least Squares Support Vector Machine(LSSVM) was proposed. The main researchs of this paper were done as follow:(1) The research of chiller's fault detection method based on ICAAiming at the characteristic that the process variables of the chiller can not be strictly obeyed Gauss distribution, and serious correlation between the large number of variables,a fault detection method based on ICA is described. The method is used to extract the independent components of the high dimensional data, and then to construct the monitoring statistics, to detect the fault. Finally, the method is verified by using the experimental data of ASHRAE 1043-RP. Compared with the PCA, the results showed that the detection model based on ICA has better fault detection performance and has good sensitivity for early fault, can reduce the false alarm rate effectively.(2) Proposed a fault detection method of chiller based on the improved Fast ICAA new method of fault detection based on improved Fast ICA algorithm is proposed, in order to solve the problem of the sensitivity of the initial point selection of traditional Fast ICA. In the improved method, the relaxation factor is introduced, to change the original iterative way, and reduce the dependency of the algorithm to the initial value; with improved Fast ICA algorithm is used to extract independent components and constructs statistics for fault detection. It used to extract the independent components and construct the monitoring statistics, to construct the statistic model for fault detection. The experimental data of ASHRAE 1043-RP is used to verify the detection performance of the improved method. The results show that the improved Fast ICA algorithm has reduced the dependency of the algorithm to the initial value, and improve the accuracy of fault detection.(3) Research on a fault diagnosis method based on ICA-LSSVMThe correlation between the variables of the chiller is serious, and the symptoms and causes of the fault are different, which lead to the difficulty of the fault diagnosis of the chiller. To solve this problem, a fault diagnosis method based on ICA-LSSVM is studied. ICA method is used to extract the independent information of the process variables, which is used as the input value of the LSSVM classifier to identify the fault. Finally, the performance of the fault diagnosis model is verified by the experiment with ASHRAE 1043-RP data. Compared with the traditional fault diagnosis methods, this new method not only improves the efficiency of fault diagnosis, but also accelerates the calculation speed.(4) Fault simulation experiment of chillerThe fault simulation experiment of chiller was done in subway training platform of some university in Beijing. In the experiment, a water-cooled piston type chiller rated cooling capacity 173 kw is selected as the research object, has simulated some faults, like refrigerant leakage?reduced cooling water flow, etc. Through the analysis of experimental data, and verifying the FDD model based on ICA-LSSVM, results show that it has good performance. On the one hand, the detection model based on ICA has a high fault detection rate, and sensitivitive to early fault. On the other hand, The fault diagnosis method based on ICA-LSSVM can effectively extract the higher order statistics of the data, reduce the redundancy of the data, and improve the efficiency of fault diagnosis.