为实现有载分接开关(on-load tap changer)在复合故障下的精准故障诊断,提出一种基于多尺度特征提取与改进天鹰算法(improved aquila optimizer,IAO)和长短时记忆神经网络(long short-term memory networks,LSTM)的变压器OLTC故障诊断方法.首先提取OLTC振动信号时域尺度、频域尺度和能量熵尺度特征组成特征向量;通过混合初始化策略和精英解保留策略对天鹰优化算法(aquila optimizer,AO)进行改进,以提高收敛性;利用改进天鹰算法对LSTM的隐含层节点数和学习率进行优化,得到最优LSTM模型;以单一故障和复合故障融合特征向量为输入,以故障状态作为输出,在最优网络模型中训练,完成后进行故障诊断.结果表明,文中所述方法平均准确率达97.2%,适用于OLTC的故障诊断.
Fault Diagnosis Method for On-load Tap Changer Based on Multiscale Feature Extraction and IAO-LSTM
To realize the accurate fault diagnosis of on-load tap changer(OLTC)under compound faults,a fault diagnosis method for transformer OLTC based on multi-scale feature ex-traction and IAO-LSTM was proposed.Firstly,features of the time domain scale,frequency domain scale and energy entropy scale were extracted from OLTC vibration signals to form fea-ture vectors.By incorporating the mixing initialization strategy and elite solution retention strategy,the aquila optimizer(AO)was improved to enhance the convergence.The improved aquila optimizer(IAO)was used to optimize the number of hid-den layer nodes and learning rate of LSTM,and thus an optim-al LSTM model was obtained.Taking the fusion eigenvector of the single fault and compound fault as the input and the fault state as the output,the optimal model was trained.After that,the fault diagnosis was carried out.The results indicate that the method yields an average accuracy of 97.2% and is appropriate for OLTC fault diagnosis.
国家科技期刊平台
NSTL
万方数据
