摘要
局部放电诊断是对电缆状态进行评估的重要手段,振荡波电压法是有较好应用前景的一种测试方法。该文提出了一种考虑振荡波周期衰减特性的电缆缺陷类型识别算法,所设计的神经网络综合利用周期内的放电特性和周期间的时序信息。首先,为便于提取放电特征的同时保留周期间的时序信息,将原始放电数据转换为周期相位分布模式。随后,从每个周期内提取3种典型放电参量,再通过卷积神经网络对其进行特征提取,融合后得到组合序列特征。最后,将组合特征送入循环神经网络进行处理,利用全连接层对缺陷类型进行分类识别。为验证算法的有效性,对4种人工缺陷电缆进行振荡波试验和分析,结果表明:该文算法的识别准确率达92%,优于其他常用识别算法,振荡波周期间蕴含的时序信息对识别效果有显著提升;此外,考虑振荡波衰减特性的方法对振荡波电压下的电缆局放类型识别具有普适性,可作为一个通用方法对其他算法进行优化。
Partial discharge(PD)diagnosis is an important tool to assess the condition of the cables,and oscillating wave test is a promising method.This paper proposes a method for PD pattern recognition of cables considering attenuation of damped AC voltages.The designed neural network comprehensively considers PD characteristics within the cycle and the temporal information between cycles.Firstly,raw PD signals are transformed to periodic phase-resolved partial discharge(P-PRPD)mode to extract PD characteristics to retain relationship among various cycles.Then,three parameters are calculated in each cycle,features of which are extracted by convolutional neural network(CNN)and fused to get a combined feature.Finally,the combined feature is fed to recurrent neural network(RNN)for processing,after which a fully-connected neural network is utilized to make the final prediction of the defect type.In order to verify the availability of the model,four artificial defective cable models are tested and analyzed by an oscillating wave test system(OWTS).The results show that the recognition accuracy of the algorithm in this paper is 92%,which is better than other commonly used recognition algorithms.The periodic attenuation information can significantly improve the recognition accuracy of proposed model.Besides,considering the attenuation characteristics of the oscillating wave is a general method for pattern recognition of PD under damped AC voltage,which can be used as a universal method to optimize other algorithms.
作者
金森
张若兵
杜钢
JIN Sen;ZHANG Ruobing;DU Gang(Tsinghua Shenzhen International Graduate School,Shenzhen 518055,China;Guangzhou Power Supply Bureau Co.,Ltd.,Guangzhou 510410,China)
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2021年第7期2583-2590,共8页
High Voltage Engineering
基金
国家自然科学基金(51677105)。
