【作者】 赵鹏程；

【导师】 李冬辉；

【作者基本信息】 天津大学 ， 控制科学与工程， 2017， 硕士

【摘要】 大型公共建筑能耗巨大,空调系统的优化控制和及时的故障诊断对节省建筑能耗意义重大。空调系统传感器的准确读数是优化控制和部件故障的精确诊断的前提,为保证空调传感器读数准确,不少国内外学者都在研究空调系统传感器故障诊断方法。空调传感器故障中的难点是渐变故障和既有故障诊断,传统故障诊断方法不能有效解决这些问题,本课题在传统故障诊断基础上主要做了以下工作和改进:首先,本文考虑大型公共建筑中包含大量相似空调单元,基于相似单元的模型参数服从相同统计特性的原理设计基于相似单元参数统计特性的故障诊断方法,这种方法克服了传统故障诊断方法只能和本单元数据对比的缺点,能够更加有效的检测系统的既有故障。其次,本课题采用卡方拟合检验对单元参数的整体分布进行拟合检测,对样本的整体分布规律的检测较传统的对单个统计参数进行检验的方法能够更加有效的检测出系统的微小渐变故障。最后,本文通过基于主元分析的方法来描述传感器读数之间的波动相似度,根据波动相似度的变化和故障辨识规则来区分传感器故障和部件故障,能够节省系统维护成本,仿真结果表明此种方法较其他故障区分方法准确性更高。本课题利用实验用空调系统实时采集大量传感器数据,利用MATLAB进行仿真分析。对数据滤波后获得单元模型参数,利用基于高斯混合模型的EM聚类算法进行聚类,在故障检测部分,对比卡方拟合检验与传统故障检测方法的ROC曲线,验证了卡方拟合检验对渐变故障的检测效果更佳。在检测出单元故障后,利用监测波动相似度方法识别故障单元的故障来源,能够有效的识别传感器故障和部件故障。更多还原

【Abstract】 Large public building energy consumption is huge,the air conditioning system optimization control and timely fault diagnosis to save building energy consumption is significant.Accurate readings of air-conditioning system sensors is the prerequisite for optimal control and accurate diagnosis of component failure.In order to ensure accurate air-conditioning sensor readings,many domestic and foreign scholars are studying air conditioning system sensor fault diagnosis method.The difficulty of air conditioning sensor fault is the gradual failure and the existing fault diagnosis.The traditional fault diagnosis method can not effectively solve these problems.In this paper,the following work and improvement are mainly done on the basis of traditional fault diagnosis:Firstly,this paper considers the large number of similar air-conditioning units in large public buildings,and the model parameters based on the principle that similar units obey the the same statistical characteristics.This method overcomes the limitation of traditional fault diagnosis methods.The shortcomings of this unit data comparison,can more effectively detect the system’s existing failures.Secondly,the chi-square fitting test is used to fit the whole distribution of the unit parameters,and the detection of the whole distribution law of the sample can detect the small gradient of the system more effectively than the traditional method of checking single statistic parameter.malfunction.Finally,the paper describes the fluctuation similarity between sensor readings based on principal component analysis(PCA),which can save system maintenance costs by distinguishing between sensor failure and component failure according to the variation of wave similarity and fault recognition rules.The simulation results show that the method is more accurate than other fault classification methods.In this paper,a large number of real-time sensor data are collected in the experimental air-conditioning system,and simulated by MATLAB.After filtering the data,the parameters of the units are obtained,and the EM algorithm is used to cluster these units.By the ROC curve of the chi-square fitting test and the traditional fault detection method,the chi-square fitting test detect gradient fault better.After detecting the fault of the unit,we can effectively identify the sensor failure and the component failure by monitoring the fluctuation similarity.更多还原