面向工程应用的供暖系统用户侧数据驱动异常诊断方法研究  

Research on User-Side Data-driven Fault Diagnosis in Heating Systems for Engineering Application

作　　者：丛铭阳 高连瑞 王源 徐海东 谢洪涛 崔建敏 Cong Mingyang;Gao Lianrui;Wang Yuan;Xu Haidong;Xie Hongtao;Cui Jianmin(School of Architecture and Design,Harbin Institute of Technology,Harbin,150000,Heilongjiang;Key Laboratory of Cold Region Urban and Rural Human Settlement Environment Science and Technology,Ministry of Industry and Information Technology,Harbin,150000,Heilongjiang;Daqing Heating Branch,Heilongjiang Longtang Power Investment Co.,Ltd.,Harbin,150028,Heilongjiang)

机构地区：[1]哈尔滨工业大学建筑与设计学院,哈尔滨150000 [2]寒地城乡人居环境科学与技术工业和信息化部重点实验室,哈尔滨150000 [3]黑龙江龙唐电力投资有限公司大庆供热分公司,哈尔滨150028

出　　处：《建设科技》2024年第15期19-22,共4页Construction Science and Technology

基　　金：国家自然科学基金:NFSC2022017。

摘　　要：随着供热系统数字化发展,数据驱动故障诊断在管网和热源侧的应用逐渐增多。但因成本与隐私问题,用户侧数据有限,故有限数据条件下用户侧异常识别方法对供暖系统高效运行意义重大。本文基于实测数据建立并验证了供暖系统机理模型,通过模型生成包含异常工况训练数据样本,通过特征工程,核函数和参数优化,建立支持向量机模型(SVM)。经分析,PSO-RBF-SVM方法的综合工程适用性最佳,分类误差为4.21%。研究改进了模型输入参数,仅需通过室内、外温度和用户回水温度三种测量值建立,具有较高的工程适用性。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.

关 键 词：集中供热 支持向量机 故障诊断 热用户 

分 类 号：TU832[建筑科学—供热、供燃气、通风及空调工程]