基于SVR-OCSVM模型的多联机系统用能评估与诊断  

Energy Assessment and Diagnosis of Variable Refrigerant Flow System Based on SVR-OCSVM Model

作　　者：刘佳慧 刘江岩[1] 陈焕新[1] 黄荣庚 李正飞 Liu Jiahui;Liu Jiangyan;Chen Huanxin;Huang Ronggeng;Li Zhengfei(School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan, 430074, China)

机构地区：[1]华中科技大学能源与动力工程学院,武汉430074

出　　处：《制冷学报》2020年第4期75-81,共7页Journal of Refrigeration

基　　金：国家自然科学基金(51876070&51576074)资助项目。

摘　　要：直接根据多联机系统能耗数据的变化来判断导致能耗大幅波动的因素是很困难的。本文提出一种有效的可用于多联机系统的能耗评估与诊断方法:将支持向量回归(SVR)算法与单类支持向量机(OCSVM)算法相结合,首先通过提取系统能耗数据集特征,去除非稳态数据,根据提取的特征变量与系统能耗建立SVR模型,预测多联机系统能耗;然后将实际能耗值与预测能耗值之差和之比分别标准化,作为输入变量,建立单类支持向量机(OCSVM)模型进行样本判别,确定是否为导致系统能耗异常的原因,以此评估诊断多联机系统能耗情况。本文基于多联机能耗正常的数据集构建了能耗评估与诊断模型,并用多联机系统能耗异常数据集验证了模型的可靠性。结果表明:基于SVR-OCSVM模型的能耗评估与诊断模型具有较高的准确度,基本能达到70%以上。Normal factors such as complex control strategies,flexible meteorological parameters,and operation conditions or abnormal factors such as refrigerant charge faults in variable refrigerant flow(VRF)systems can lead to complex and varying fluctuations in the energy consumption.It is difficult to directly diagnose whether normal or abnormal factors cause such fluctuations based on the energy consumption.In this study,an effective energy assessment and diagnosis method is proposed,which combines the support vector regression(SVR)algorithm with the one-class support vector machine(OCSVM)algorithm to diagnose the energy performance of a VRF system.An energy assessment and diagnosis model is constructed based on the normal data set,and is verified by the abnormal energy data set.The results show that the energy assessment and diagnosis model based on SVR-OCSVM has a high accuracy of up to 70%.

关 键 词：多联机系统 用能评估与诊断 支持向量回归 单类支持向量机 

分 类 号：TB657.2[一般工业技术—制冷工程] TU831.3[建筑科学—供热、供燃气、通风及空调工程]