Research On Intelligent Fault Diagnosis Algorithms For Sewage Source Heat Pump System In A Residential Area Of Xi'an

Sewage source heat pump technology is a kind of "zero pollution" green energy which promoted by the government.It has broad application prospects.In this paper,the sewage source heat pump system in Xian Vanke Golden Yuecheng District is taken as a case study.This project is a key environmental protection and new energy construction project in Xi'an.In the process of participating in the project investigation,it is found that the heat pump system may fail in the operation process.After the failure,because of the complexity of the system,workers need to test different parts of the heat pump system according to the characteristics of the fault,and then locate the fault point,which is a costly and time-consuming business.More importantly,during the maintenance period,the heat pump system can not be working properly,causing a lot of economic losses.At the same time,it also needs to face the complaints of end users.The generation of heat pump fault is a gradual process.In order to solve the practical problems encountered in the above projects,the key lies in how to early warn and seek the location of the fault before the outage of the system equipment occurs,so as to maintain it in time.This paper explores the use of intelligent algorithm to establish fault diagnosis model of sewage source heat pump and diagnose it.Firstly,based on the data of sewage source heat pump system collected during the heating period of 2016-2019(November 15 to March 15 each year),three common types of faults are selected(compressor solenoid valve faults,evaporator scaling and board replacement scaling),and 100 sets of training data of each type of faults areextracted as samples to establish fault state set.According to the comparison 100 sets of "normal operation state" are taken at the same time.Because the operation of heat pump system is a dynamic process,there are varying degrees of correlation between the changes of equipment parameters in the system.This paper establishes the mapping relationship between operation data and operation status by means of the embedded sensor feedback data.Secondly,based on the set of fault states established earlier,the corresponding BP neural network structure model is built.BP algorithm is used to diagnose heat pump.The experimental results show that the accuracy is 40%.In order to improve the experimental results,adaptive weighted particle swarm optimization(AWPSO)is used to optimize the weights and thresholds of BP model and the adaptive learning training algorithm with momentum term is used to train the model.The experimental results show that the optimized algorithm achieves better experimental results,and the accuracy of fault diagnosis reaches 85%.However,considering the huge amount of data in the long run of the project,it is necessary to further improve the experiment.After analysis,the limited number of samples has a great impact on the neural network algorithm.In view of the good performance of the support vector machine(SVM)algorithm under small samples,based on the platform of MATLAB,this paper uses the LIBSVM toolbox to construct the SVM model for fault diagnosis of heat pump system,and the diagnostic accuracy reaches 95%.Finally,through in-depth analysis of the above experimental results,it is found that although the diagnostic accuracy of the above algorithms has been greatly improved at the qualitative level,there is a greater potential risk of false positives in the quantitative analysis level.The result of fault diagnosis in this paper is based on the membership value of different operation states.When the membership value of a certain type of state is the largest,it belongs to this type of fault.However,when calculating the membership value of the above algorithm,part of the output results show that there is no obvious difference in the membership value under different state types.How to improve the level of confidence in diagnosis has become the focus of research.In Chapter 4,a Multi-source information fusion method based on D-S evidence theory is introduced.The results of BP,AWPSO-BP and SVM aretransformed into evidence bodies for fusion analysis.The results show that the model established by the information fusion method achieves ideal results,not only improves the level of confidence in diagnosis,but also corrects some of the misdiagnosis cases in the above single algorithm.Based on the above discussion,using intelligent algorithm to diagnose the fault of sewage source heat pump system can achieve the purpose of early warning before system outage failure,ensure the normal and efficient operation of heat pump system,and play a better guiding role in the actual operation and maintenance of heat pump.