Research On Operation Data Diagnosis Of Central Air Conditioning System Based On Graph Network Model

In the current HVAC system operation and maintenance industry in China,the operation and management of central air conditioning system lacks scientific guidance.In addition,because engineers can not directly and effectively diagnose and analyze the overall operation state of the air conditioning system,there is a huge potential for energy conservation.The application of building automation system usually collects and stores a large number of actual operation data,which is the most direct and original message of the actual operation state of central air conditioning system.If through the above data,the information and knowledge inside the data can be deeply explored,and the operation and maintenance analysis and diagnosis of the central air conditioning system can be assisted to a higher efficiency of energy-saving.However,the traditional single data mining method often ignores the integrity and relevance of the central air-conditioning system,pays less attention to the comprehensive analysis of the whole system and ignores the information hidden in the association structure between different data dimensions.Therefore,based on the graph network data mining technology,this paper establishes the central air conditioning operation data Graph Network model,and proposes a complete set of central air conditioning operation data analysis and diagnosis method based on the model.Firstly,the data is preprocessed by linear interpolation to solve the problem of inconsistent data sampling frequency;secondly,the data is preliminarily analyzed by the combination of descriptive analysis and statistical analysis,and the data exploration work is carried out by using the indexes of mean,maximum,median,standard deviation,skewness,kurtosis and Pearson correlation coefficient.Then,we combine symbol aggregation processing with monotonic trend processing to acquire the association attributes from time series data dimension reduction processing;then we propose the detailed modeling steps of Graph Network model from "data node","association edge" and "network structure",which can effectively model the original building data set into graph network data storage and expression;finally,we propose a set of analysis and diagnosis methods based on frequent subgraph mining,using gspan algorithm to analyze the frequent pattern of the above graph network data.In addition,we propose an automatic screening method based on difference coefficient analysis to realize frequent pattern diagnosis and daily running diagnosis of historical data.This method is analyzed from the overall system perspective,which is convenient for engineers to quickly find out the problem and specific influencing factors.The method is applied to 2 months operation data of an actual central air-conditioning system for case analysis,and four major energy-saving potential problems and some irregular operation behaviors are effectively diagnosed.The results show that the method is reasonable and effective.The research results of this paper are applicable to more diagnosis and analysis of central air conditioning system operation in the future,which is convenient to improve the work efficiency of engineering and researchers,and has certain practical significance.