Combined fault diagnosis method of centrifugal water chilling unit and primary air system

Abstract

The present invention discloses a combined fault diagnosis method of a centrifugal water chilling unit and a primary air system. The method comprises: historical data denoising processing and featurescreening, clustering analysis, determining a support vector machine training parameter, establishing a fault diagnosis model and the like. The denoising processing and feature screening is performedto remove inevitable noise and abnormal value data from original data, convert the inevitable noise and abnormal value data into a data structure recognizable by the support vector machine, and selecta variable significantly correlated with a fault. A storage space is saved on the data structure, and a model training time is shortened. The clustering analysis is performed to determine closeness of fault features, and improve diagnosis precision. The support vector machine training parameter can be determined to improve training precision of the fault diagnosis model and a compression model generation time. According to the method, dual system combined fault diagnosis can be widely performed with high precision, few time consumption and wide applicability, and a problem of a dual system can be discovered in time, so that objectives of improving operating performance, saving energy and reducing maintenance costs are achieved.

Description

technical field [0001] The invention belongs to the technical field of fault diagnosis, and in particular relates to a joint fault diagnosis method of a centrifugal chiller and a fresh air system. It is applicable to the diagnosis of a total of 9 kinds of faults in centrifugal chillers and fresh air supply systems in the field of HVAC. Background technique [0002] The energy consumption in the whole process of construction accounts for a huge proportion of my country's total energy consumption. Taking the total energy consumption of building operation as an example, it accounted for about 20% of the country's total energy consumption in 2015. In the energy consumption of building operation, the energy consumption of the HVAC system accounts for about 60% of the total energy consumption of the building, and due to the widespread problems of unreasonable design and improper maintenance, various HVAC subsystems often operate at low The energy efficiency ratio range has exace...

AI Technical Summary

Problems solved by technology

In the above methods, although the fault tree and expert system can diagnose multiple faults at the same time, they require a large amount of prior knowledge and tedious rule settings, rely too much on individual buildings, and have obvious personality characteristics, so they are not too high. Universality; artificial neural network, regression model, fuzzy algorithm, etc. all need a large amount of input operating data under normal and fault conditions to construct the model, which is often difficult to achieve in actual engineering, and it is difficult to achieve multiple faults The joint diagnosis may not guarantee the accuracy of the joint diagnosis, and it is also prone to overfitting problems resulting in poor generalization ability; the independent meta-analysis, designated meta-analysis and principal component analysis in the statistical method need the operating data of the unit under normal conditions Carry out model construction and judge whether a fault occurs by comparing some characteristic indicators, but the types of faults it is suitable for diagnosing are relatively single, and if two faults have the same fault characteristics, it is difficult to give an accurate diagnosis result

Images