【作者】 周守军； 孙文虎； 刘祥瑞； 张弛； 李法昌； 姜国斌； 孙钱行；

【Author】 Zhou Shoujun;Sun Wenhu;Liu Xiangrui;Zhang Chi;Li Fachang;Jiang Guobin;Sun Qianxing;Shandong Jianzhu University;Zibo Boshan District Heating Company;

【机构】 山东建筑大学； 淄博市博山区热力公司；

【摘要】 为了保证供热管网的安全稳定运行,降低管网堵塞发生率及其造成的能源浪费,本文以实验室的供热管网实验系统为研究对象,提出了一种基于BP神经网络并结合数据标准化和主成分分析(PCA)的管网堵塞诊断方法。针对供热管网实验系统的四种不同拓扑结构:单热源枝状(单枝)、单热源单环(单环)、双热源枝状(双枝)、双热源双环(双环),以管网实验数据、仿真数据以及二者的交叉数据为数据样本,采用主成分分析(PCA)和数据标准化方法,构建并训练管网堵塞位置及堵塞程度诊断的BP神经网络诊断模型。模型预测结果表明:以实验数据集构建的诊断模型,四种管网的预测准确率分别为99.07%,99.28%,99.41%,99.12%;以仿真数据集构建的诊断模型,四种管网的预测准确率分别为99.96%,99.94%,99.91%,99.96%;以交叉数据集构建的诊断模型,随着交叉数据比(仿真数据:实验数据)的增大,四种管网的预测准确率分布区间为:99.51%-98.86%,93.96%-89.29%,99.41%-90.26%,99.44%-89.63%,且准确率变化趋势稳定。更多还原

【Abstract】 To ensure the safe and stable operation of heating networks and reduce the incidence of heating network blockage and the energy waste caused by it,this paper proposes a heating network blockage diagnosis method based on BP(Back Propagation) neural network combined with data standardization and principal component analysis(PCA),which takes the heating network experimental system in the laboratory as the research object.In view of the four different topological structures of the heating network experimental system:branch with single heat-source(single branch),single loop with single heat-source(single loop),branch with double heat-source(double branch),double loop with double heat-source(double loop),the BP neural network diagnosis model of blockage location and degree is constructed and trained by principal component analysis(PCA) and data standardization method,using experiment data,model data and cross data as data samples.The model prediction results show that:For the four heating-networks,the prediction accuracy of the model build with experimental datasets is 99.07%,99.28%,99.41%,and 99.12% respectively;the prediction accuracy of the model build with model datasets is 99.96%,99.94%,99.91%,and 99.96% respectively;the prediction accuracy of the model build with cross datasets is in the range of 99.51%-98.86%,93.96%-89.29%,99.41%-90.26%,and 99.44%-89.63%,and the accuracy change trend is stable.更多还原