| 引用本文: | 李俊卿,李斯璇,陈雅婷,王振兴,何玉灵.一种基于CGAN-CNN的同步电机转子绕组匝间短路故障诊断方法[J].电力自动化设备,2021,41(8): |
| LI Junqing,LI Sixuan,CHEN Yating,WANG Zhenxing,HE Yuling.Fault diagnosis method of inter-turn short circuit of rotor winding of synchronous motor based on CGAN-CNN[J].Electric Power Automation Equipment,2021,41(8): |
|
| 摘要: |
| 由于同步电机故障样本数量较少，为解决同步电机故障诊断中普遍存在的样本不平衡问题，提出了一种基于条件生成式对抗网络(CGAN)和卷积神经网络(CNN)的同步电机转子绕组匝间短路故障诊断方法。首先，对传感器收集到的数据进行预处理，对正常样本和故障样本分别添加标签后输入CGAN中生成大量新样本，将生成的新样本与原始样本混合并划分训练集和测试集；然后，利用CNN训练平衡后的数据集，充分、精准地提取有效故障特征；最后，在输出端利用Softmax分类器输出故障分类结果。通过实验证明，与非平衡数据集相比，利用平衡数据集后的故障识别准确率十分稳定且达到99.5 %以上，同时与平衡的原始样本数据相比，生成样本避免了噪声和其他干扰，故障诊断的准确率也更高。 |
| 关键词: 同步电机 条件生成式对抗网络 卷积神经网络 生成样本 转子绕组匝间短路故障 故障诊断 |
| DOI：10.16081/j.epae.202105020 |
| 分类号:TM341 |
| 基金项目:国家自然科学基金资助项目(51777074) |
|
| Fault diagnosis method of inter-turn short circuit of rotor winding of synchronous motor based on CGAN-CNN |
|
LI Junqing1, LI Sixuan1, CHEN Yating1, WANG Zhenxing1, HE Yuling2
|
|
1.Department of Electrical Engineering, North China Electric Power University, Baoding 071003, China;2.Department of Mechanical Engineering, North China Electric Power University, Baoding 071003, China
|
| Abstract: |
| Due to the small number of synchronous motor fault samples, in order to solve the problem of unbalanced sample problem in fault diagnosis of synchronous motor, fault diagnosis method of inter-turn short circuit of rotor winding of synchronous motor based on CGAN(Conditional Generative Adversarial Networks) and CNN(Convolution Neural Networks) is proposed. Firstly, the data collected by sensors is preprocessed, the normal and fault samples are respectively labeled and then put into CGAN to generate a large number of new samples. The generated new samples are mixed with the original samples, and they are divide into training set and test set. Then the CNN model is used to train the balanced data set to extract the effective fault features fully and accurately. Finally, the fault classification results are outputted by Softmax classifier at the output end. The experimental results show that compared with the unbalanced data set, the fault recognition accuracy of the balanced data set is very stable and reaches over 99.5 %. Moreover, compared with the balanced original sample data, the generated samples avoid noise and other interference, and the accuracy of fault diagnosis is also higher. |
| Key words: synchronous motor conditional generative adversarial network convolution neural network generated sample inter-turn short circuit of rotor winding fault diagnosis |
