摘要
介绍深度学习的级联森林模型与原理。根据工程现场在线监测数据的特点,考虑到深度学习较强的样本特征转换和提取能力,选取H_2、CH_4、C_2H_6、C_2H_4、C_2H_2、CO和CO_2作为特征分析变量。实验结果证明:深度森林方法可以有效提高变压器故障诊断的正确率,且诊断效果优于传统方法。
Both cascading forest model and working principle of deep learning method was introduced. Basing on characteristics of on-line monitoring at the engineering site and considering obvious sample characteristics transform and extraction ability of the deep learning method, the H2, CH4, C2H6, C2H4 ,C2H2, CO and CO2 were taken as feature analysis variables. Experimental results show that, this deep forest method can improve the accuracy of fault diagnosis and it outperforms traditional method in diagnosis effect.
出处
《化工自动化及仪表》
CAS
2018年第1期69-72,共4页
Control and Instruments in Chemical Industry
关键词
故障诊断
油浸式变压器
溶解气体分析
深度森林模型
fault diagnosis, oil-immersed transformer, dissolved gas analysis, deep forest model
