Anomaly Detection System Of Heating Secondary Pipe Network Based On Big Data Platform

In recent years,the balance control and energy saving of the heating secondary pipe network have gradually become the focus of heating enterprises.With the large-scale installation of heat-measuring appliances and the collection of massive data,heating companies have accumulated a large amount of data on heating terminals.However,these data have the problems of large data volume,many missing values and a certain degree of distortion,which makes heating companies lack suitable tools to play the role of these data.Based on the above problems,this paper cleans and screens the heat meter data,screens outliers through a distributed isolated forest algorithm,and clusters outliers through a Gaussian mixture model algorithm to correctly identify heating anomalies.In this paper,the heating data under actual working conditions is used as the original data.The advantage of Spark parallelization is used to integrate and filter the information of multiple large tables in the database.The Lagrange interpolation algorithm is used to fill in the missing values and apply statistical methods.Simple analysis of the characteristics of heating data with the method of drawing,and the feature selection of the data set through the combination of mutual information filtering method and random forest embedding method.In this paper,k-means algorithm(k-means),gaussian mixture model algorithm(GMM)and IForest algorithm(IForest)were designed and tested respectively to detect the data of heating anomalies.The experiment proved that the isolated forest algorithm was superior to the commonly used k-means algorithm and gaussian mixture model algorithm in indicators such as false alarm rate(FPR)and recall rate(TPR).Developed based on large data,this paper analyses the heating quality monitoring system,the system integrates the Spark big data reading module,a data cleaning module,data visualization module,the module,the generated reports to encapsulate a variety of anomaly detection algorithm,make the heating enterprise can more convenient to learn the quality of heating,in order to quickly and timely to take corresponding measures to solve the problem of heating anomaly.