Research on User-Side Data-driven Fault Diagnosis in Heating Systems for Engineering Application
With increasing digitization of heating systems,data-driven fault diagnosis is gaining traction on the network and supply side.User-side anomaly detection methods under limit input data are essential due to constrains from costs and privacy concerns.This study develops and validates a mechanistic model of the heating system based on actual data,which can generate training data that containing user-side anomalies.Utilizing feature selection,kernel function and hyperparameter optimization,a Support Vector Machine(SVM)model is established.The PSO-RBF-SVM method proves most suitable for engineering applications,with a classification error of 4.21%.Enhanced model only requires 3 types of measuring data,indoor,outdoor and return water temperature,thereby ensuring high engineering practicality.
district heatingsupport vector machinefault diagnosisheating user
NETL
NSTL
万方数据
