Research On Fault Diagnosis Of Key Equipment In Central Air-Conditioning System

Central air-conditioning is widely used in various buildings.The application of fault diagnosis technology in central air-conditioning can ensure the stable operation of the whole system and save energy at the same time.However,it is extremely difficult to build a complete fault diagnosis system for central air-conditioning system due to the non-linearity and the parameter coupling.Therefore,fault diagnosis of chillers and centrifugal pumps,as the key equipment of central air-conditioning system,is studied in this paper.The main research contents and innovations of this paper are shown as follows:(1)Fault analysis of chillers and centrifugal pumps is studied to explain their fault types and causes,and to determine the characteristic variables of fault diagnosis.For key equipment of the central air-conditioning,a fault diagnosis method based on neural network is proposed in the context of big data.(2)Back Propagation(BP)neural network,Radial Basis Function(RBF)neural network and Elman neural network are used to build fault diagnosis models of key equipment.For a particular parameter of each neural network,traversal optimization in a certain range is used to find the optimal parameters points of the network fault diagnosis.Then the fault diagnosis results become more accurate and effective.(3)The feasibility of these three kinds of neural networks is studied by fault diagnosis simulation of chillers.And the accuracy and validity are verified by centrifugal pump fault simulation experiments.(4)The actual working process of the centrifugal pump is operated under non-frequency conversion conditions,but the load of system end is variable.The method of changing the valve opening of the centrifugal pump outlet is used to simulate the different load conditions.And the different load conditions are taken as the research variables in the failure simulation experiment.(5)The effect of BP network,RBF network and Elman network on fault pattern recognition is analyzed by comparing the data simulation results of chillers and centrifugal pumps.The fault diagnosis effects of the three networks are comprehensively analyzed from four aspects: network memory,network structure complexity,network training time and network testing accuracy.It is concluded that RBF network structure is simpler than BP network and Elman network.RBF network memory is better,and network training time is short.Of course,test output error is small.Then the conclusion that RBF network is more suitable for fault diagnosis of key equipment in central air-conditioning system is confirmed.