冷水机组系统中,温度传感器出现故障会严重影响机组工作效率及使用寿命。针对冷水机组温度传感器偏差故障,本文提出一种基于单类支持向量机(one-class support vector machine,OC-SVM)的故障检测方法,采用冷水机组正常数据建立OC-SVM模...冷水机组系统中,温度传感器出现故障会严重影响机组工作效率及使用寿命。针对冷水机组温度传感器偏差故障,本文提出一种基于单类支持向量机(one-class support vector machine,OC-SVM)的故障检测方法,采用冷水机组正常数据建立OC-SVM模型,通过十折交叉验证法获得模型优化参数。分别采用工程实测数据和实验数据(共4组)对该方法进行了验证,结果表明:基于OC-SVM的方法能有效检测出4组冷水机组的温度传感器偏差故障。其中对于螺杆式冷水机组(数据集Ⅰ)的故障检测效果明显,当冷冻水侧温度传感器偏差故障幅值绝对值大于1℃时,检测效率达到100%。展开更多
Power quality assessment is an important performance measurement in smart grids.Utility companies are interested in power quality monitoring even in the low level distribution side such as smart meters.Addressing this...Power quality assessment is an important performance measurement in smart grids.Utility companies are interested in power quality monitoring even in the low level distribution side such as smart meters.Addressing this issue,in this study,we propose segregation of the power disturbance from regular values using one-class support vector machine(OCSVM).To precisely detect the power disturbances of a voltage wave,some practical wavelet filters are applied.Considering the unlimited types of waveform abnormalities,OCSVM is picked as a semisupervised machine learning algorithm which needs to be trained solely on a relatively large sample of normal data.This model is able to automatically detect the existence of any types of disturbances in real time,even unknown types which are not available in the training time.In the case of existence,the disturbances are further classified into different types such as sag,swell,transients and unbalanced.Being light weighted and fast,the proposed technique can be integrated into smart grid devices such as smart meter in order to perform a real-time disturbance monitoring.The continuous monitoring of power quality in smart meters will give helpful insight for quality power transmission and management.展开更多
文摘冷水机组系统中,温度传感器出现故障会严重影响机组工作效率及使用寿命。针对冷水机组温度传感器偏差故障,本文提出一种基于单类支持向量机(one-class support vector machine,OC-SVM)的故障检测方法,采用冷水机组正常数据建立OC-SVM模型,通过十折交叉验证法获得模型优化参数。分别采用工程实测数据和实验数据(共4组)对该方法进行了验证,结果表明:基于OC-SVM的方法能有效检测出4组冷水机组的温度传感器偏差故障。其中对于螺杆式冷水机组(数据集Ⅰ)的故障检测效果明显,当冷冻水侧温度传感器偏差故障幅值绝对值大于1℃时,检测效率达到100%。
基金supported in part through U.S.National Science Foundation(No.1553494).
文摘Power quality assessment is an important performance measurement in smart grids.Utility companies are interested in power quality monitoring even in the low level distribution side such as smart meters.Addressing this issue,in this study,we propose segregation of the power disturbance from regular values using one-class support vector machine(OCSVM).To precisely detect the power disturbances of a voltage wave,some practical wavelet filters are applied.Considering the unlimited types of waveform abnormalities,OCSVM is picked as a semisupervised machine learning algorithm which needs to be trained solely on a relatively large sample of normal data.This model is able to automatically detect the existence of any types of disturbances in real time,even unknown types which are not available in the training time.In the case of existence,the disturbances are further classified into different types such as sag,swell,transients and unbalanced.Being light weighted and fast,the proposed technique can be integrated into smart grid devices such as smart meter in order to perform a real-time disturbance monitoring.The continuous monitoring of power quality in smart meters will give helpful insight for quality power transmission and management.
文摘工作在复杂环境下的多元退化设备面临失效数据少、多源信息融合准确度低和监督学习数据不平衡等问题,对此本文提出一种基于时间序列生成对抗网络(Time-series Generative Adversarial Networks,TimeGAN)与单分类支持向量机(One-Class Support Vector Machine,OCSVM)组合模型的小子样数据增广方法.方法引入了TimeGAN模型拟合真实数据时间序列相关性,从而生成新的多元退化设备数据.本文提出了一种基于最大均值差异改进方法的可信度判据,避免强相关特征对生成数据质量评价的影响,通过使用T-分布随机邻近嵌入(T-distributed Stochastic Neighbor Embedding,T-SNE)和全局最大均值差异(Global Maximum Mean Discrepancy,GMMD)的组合方法,定性定量地评价生成数据的质量水平.基于训练后的OCSVM模型,对生成数据进行异常检测与剔除,进一步提高生成数据的质量.以航空发动机数据集C-MAPSS为例进行方法验证分析,通过与其他数据增强模型对比验证了所提方法的可行性和有效性.
