基于级联极限学习机的基站空调在线监测系统  

Air Conditioning Online Monitoring System for Base Station Based on Cascaded Extreme Learning Machines

作　　者：罗方芳[1] 陶求华 LUO Fangfang;TAO Qiuhua(Computer Engineering College,Jimei University,Xiamen 361021,China;School of Mechanical and Energy Engineering,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学计算机工程学院,福建厦门361021 [2]集美大学机械与能源工程学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2018年第6期475-480,共6页Journal of Jimei University：Natural Science

基　　金：国家自然科学基金资助项目(51508225);福建省自然科学基金资助项目(2016J01244);福建省教育厅资助项目(B16162;JB13137)

摘　　要：提出一种基于级联极限学习机的基站空调在线监测系统。首先,基于某基站空调公司提供的监测数据集构建多个原子极限学习机分类器,每一个原子极限学习机对应一种故障类别;再将各原子分类器以级联方式组合用于未知样本的故障诊断;最后将级联极限学习机与单独的多类极限学习机算法、SVM算法、BP神经网络算法、C4. 5决策树算法进行比较测试。结果表明,级联极限学习机算法提高了小类样本的故障识别率,具有更高的故障诊断精度和较短的训练时间,且诊断时间达到在线实时的要求。An efficient and real-time fault detection and diagnosis system for on-line base station air conditioner can ensure the stable operation of various equipment in the base station. Due to the fault categories of the base station air conditioner are nonlinear and unbalanced,an on-line monitoring system of cascaded extreme learning machines is proposed for the fault diagnosis of air conditioner in field base station. Firstly,based on the data set provided by a base station s air conditioning company, a collection of basic binary extreme learning machine classifiers are constructed, in which each classifier corresponds to a fault category. Then,these basic binary classifiers are combined in a serial cascade to be applied to the fault diagnosis of new samples. Finally, the cascaded extreme learning machine algorithm is compared with single multi-class extreme learning machine,SVM algorithm,BP neural network algorithm and C4. 5 decision tree algorithm when they are used in the on-line monitoring system of the field base station air conditioner. Experimental results show that the cascaded extreme learning machine algorithm can improve the recognition rate for minority class,and has the advantages of higher fault diagnosis accuracy and shorter training time than traditional algorithm. Furthermore,the test diagnosis time can meet the on-line and real-time requirement.

关 键 词：基站空调 故障诊断 级联极限学习机 在线监测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]