Research And Application Of Soft-sensor Technology In Cooling Capacity Test Of Room Air Conditioners In-service

At present,the test methods for cooling capacity of air conditioning system in laboratory have been relatively mature,mainly using indoor air enthalpy difference method and heat balance calorimeter method.The enthalpy difference method is mostly adopted by factories and testing institutions because of its low equipment investment cost and short overall measurement cycle,and has formed a standard which has been implemented by most countries in the world.However,there is no uniform and standard method for testing cooling capacity and long-term on-line monitoring of air-conditioner in-service installed in real life.For air-conditioning products installed in buildings,long-term measurement using indoor air enthalpy difference method will seriously affect the use of consumers due to the limitations of testing equipment and operating environment.It's not appropriate.Therefore,a soft sensor based on neural network is proposed in this paper.After measuring some easily measurable variables related to cooling capacity of air conditioning system,the cooling capacity can be deduced or estimated by computer technology and modelling variable relationship,so as to achieve the purpose of long-term on-line measurement of cooling capacity of household air conditioning system.This research is based on the performance of air-conditioning products in the actual use process.It focuses on the method of measuring and realizing long-term on-line monitoring of cooling capacity of household air-conditioning products in the actual use process.Firstly,starting from the detection methods described in national standards,this paper analyses and discusses the characteristics of different detection methods for measuring refrigeration capacity of air-conditioning systems.By analyzing the thermodynamic principle of the air conditioning system and the structure of the key components of the air conditioning system,the relationship between the interface parameters of the components of the air conditioning product system is studied,and the mathematical expression of the relationship between the comprehensive testing requirements and the physical characteristics of the system interface is established.Then,according to the actual situation of the in-service household air-conditioning system,a data acquisition platform is built.By means of data acquisition of system state parameters which are easy to measure on site and real-time transmission of cloud platform,the necessary training and learning data sets are provided for the establishment of soft-sensing model.Based on the collected data,the realizability of soft-measurement cooling capacity of in-service air-conditioning system using neural network is discussed.At the same time,the performance of BP neural network,radial basis function(RBF)neural network and support vector machine(SVM)in this practical application is compared.Finally,aiming at the strong dependence of BP neural network on initial weights,evolutionary algorithm is used to optimize the BP neural network,and MIV algorithm is used to screen the input variables of BP neural network,which achieves the purpose of improving system efficiency and reducing the cost of online measuring instruments.The results show that the soft sensing of cooling capacity of domestic air conditioner in service by using neural network can achieve better results,and it is feasible in practical engineering application.