高压断路器作为现代电力系统的核心设备之一，在电力系统运行中发挥着不可替代 的双重作用：既承担着正常工况下负荷投切的操作功能，又肩负着快速切除故障线路的 保护使命。高压断路器一旦发生故障，不仅会影响其他电气设备的正常运行，更可能危 及整个电力系统的安全稳定。因此，开展高压断路器故障诊断方法的研究对于提升电力 系统可靠性具有重要的理论价值和实践意义。

本文基于搭建的高压断路器实验平台，系统采集了高压断路器在正常运行状态及典 型故障工况下的振动信号，从振动信号的特征提取与故障识别方面进行研究，主要研究 工作内容包括以下三个方面：

（1）针对高压断路器振动信号的故障特征提取较为困难的问题，提出一种基于 CPO-VMD-GPLVM 的特征提取方法，采用冠豪猪优化算法（CPO）对变分模态分解 （VMD）的关键参数（分解层数和惩罚因子）进行智能寻优；完成 VMD 自适应分解 后，根据各模态分量与原信号的皮尔逊相关系数大小进行筛选，选取相关系数最大的模 态分量作为敏感分量，提取敏感分量的时域、频域特征和复合多尺度样本熵（CMSE） 构成 16 维的多域特征集，最后采用高斯过程潜在变量模型（GP-LVM）进行特征降维， 提取出关键的特征量。

（2）针对当前高压断路器故障诊断中存在的模型精度不足及样本关联性缺失问题， 提出一种基于图注意力网络（GAT）的高压断路器故障诊断方法，将提取出的特征向量 构成 k 近邻图，通过图建立了故障样本之间的联系，充分挖掘样本之间的潜在相似性特 征并利用 GAT 的注意力机制以提高模型精度。实验结果表明，GAT 故障诊断模型在测 试集上的准确率可达 99.37%，与其他故障诊断模型相比，GAT 的准确率明显高于其他 模型，且抗噪性更好。

（3）针对在工程实际中由于条件因素限制难以采集到足够的样本数据，导致故障 诊断模型性能下降或失效的问题，提出一种基于元学习的高压断路器小样本故障诊断方法。该方法将 k 近邻图输入到图编码器中提取出特征向量，运用基于度量的元学习方法 实现小样本故障诊断。实验结果表明，所提方法在有限样本条件下仍能保持优异的诊断 性能，为解决断路器故障诊断中的小样本难题提供了一种可行的方案。

综上所述，本文研究针对高压断路器故障诊断中振动信号特征提取困难、诊断模型 精度不足及小样本问题导致诊断模型性能大幅下降甚至失效等问题，提出了相应的解决 方案，为更好的推动高压断路器故障诊断的发展提供参考。

High-voltage circuit breakers, as one of the core components in modern power systems, play an irreplaceable dual role in power system operation: they are responsible for performing load switching operations under normal operating conditions while simultaneously fulfilling the critical protective mission of rapidly isolating faulted lines. When failures occur in high-voltage circuit breakers, they not only compromise the normal operation of adjacent electrical equipment but also pose significant risks to the overall security and stability of the power grid. Therefore, conducting research on fault diagnosis methodologies for high-voltage circuit breakers holds substantial theoretical significance and practical implications for enhancing the reliability of power systems.

Based on the established experimental platform for high-voltage circuit breakers, this study systematically collected vibration signals under both normal operating conditions and typical fault scenarios. The research focuses on feature extraction and fault identification from the vibration signals, with the primary research efforts encompassing the following three aspects:

(1) To address the challenge of fault feature extraction from vibration signals in high-voltage circuit breakers, this paper proposes the CPO-VMD-GPLVM-based feature extraction methodology. Firstly, the Crested Porcupine Optimizer (CPO) is employed for intelligent optimization of critical parameters in Variational Mode Decomposition (VMD), specifically the decomposition layer number and penalty factor. Subsequently, after completing the adaptive VMD decomposition, modal components are filtered based on Pearson correlation coefficients between each component and the original signal. The component with the maximum correlation coefficient is selected as the sensitive component. the 16-dimensional multi-domain feature set is then constructed by extracting time-domain features, frequency-domain features, and Composite Multi-scale Sample Entropy (CMSE) from the sensitive component. Finally, Gaussian Process Latent Variable Model (GP-LVM) is implemented for feature dimensionality reduction to extract essential characteristic quantities.

(2) To address the issues of insufficient model accuracy and lack of sample correlation in current high-voltage circuit breaker fault diagnosis, this paper proposes the fault diagnosis method based on Graph Attention Network (GAT). The extracted feature vectors are constructed into a k-nearest neighbor graph, which establishes connections between fault samples. This approach fully exploits the potential similarity features among samples and utilizes GAT's attention mechanism to improve model accuracy. Experimental results show that the GAT fault diagnosis model achieves an accuracy of 99.37% on the test set. Compared with other fault diagnosis models, the GAT model demonstrates significantly higher accuracy and better noise immunity.

(3) To address the problem of performance degradation or failure of fault diagnosis models due to insufficient sample data collection under practical engineering constraints, this paper proposes the meta-learning-based few-shot fault diagnosis method for high-voltage circuit breakers. The method inputs the k-nearest neighbor graph into a graph encoder to extract feature vectors and employs metric-based meta-learning to achieve few-shot fault diagnosis. Experimental results demonstrate that the proposed method maintains excellent diagnostic performance even with limited samples, providing a feasible solution to the fewshot challenge in circuit breaker fault diagnosis.

In summary, this study addresses critical challenges in high-voltage circuit breaker fault diagnosis, including difficult vibration signal feature extraction, insufficient diagnostic model accuracy, and significant performance degradation or failure caused by few-sample problems. Corresponding solutions are proposed, providing valuable references for advancing the development of high-voltage circuit breaker fault diagnosis technologies.

[1] 任西壁. 基于分合闸线圈电流信号的高压断路器操作机构故障诊断研究[D]. 电子科技大学, 2023.

[2] 钟建英, 陈刚, 谭盛武, 等. 高压开关设备关键技术及发展趋势[J]. 高电压技术, 2021, 47(08): 2769-2782.

[3] Roy Billinton, Juan Yu, Jun Zhong,et al. Incorporating a Condition Monitoring Based Aging Failure Model of a Circuit Breaker in Substation Reliability Assessmen[J]. IEEE Transactions on Power Delivery,2018,8(2):365-374.

[4] 刘幸. 高压断路器机械特性的非接触式测量及其缺陷诊断[D]. 西安科技大学, 2022.

[5] 盖曜麟, 葛丽娟, 窦旭萌. 高压断路器故障诊断技术研究综述[J]. 电力学报, 2020,35(02): 180-186.

[6] 韩翔宇, 纽春萍, 何海龙, 等. 电磁式断路器状态监测与智能评估技术综述[J]. 电工技术学报, 2023, 38(08): 2191-2210.

[7] 冯颖姣, 汪黎明, 张盛, 等. 基于分合闸线圈电流的断路器状态在线监测系统研究及应用[J]. 电工技术, 2021(22): 97-99.

[8] 张龙, 黄新波, 蒋波涛, 等. 基于触头行程测量的断路器机械特性在线监测[J]. 西安工程大学学报, 2018, 32(03): 302-310.

[9] 孙曙光, 杨飞龙, 陈静, 等. 基于振动-电流广域特征与软共享机制的断路器多故障诊断[J]. 仪器仪表学报, 2024, 45(01): 46-59.

[10]Razi.Kazemi A A, Niayesh K. Condition Monitoring of High Voltage Circuit Breakers: Past to Future[J]. IEEE Transactions on Power Delivery, 2021, 36(2): 740-750.

[11]林婧, 张佳灏, 黎旭, 等. 基于振动信号的高压断路器螺栓松动故障诊断研究[J]. 高压电器, 2023, 59(03): 28-33.

[12]张艳飞, 邵阳, 公维炜, 等. 考虑小波包-灰度共生矩阵的高压断路器弹簧疲劳故障程度诊断研究[J]. 机械科学与技术, 2024, 43(02): 274-281.

[13]王振浩, 顾欣然, 孙福军. 基于 EMD 近似熵的高压断路器故障诊断[J]. 高压电器,2018, 54(10): 151-156+163.

[14]赵书涛, 马莉, 朱继鹏, 等. 基于 CEEMDAN 样本熵与 FWA-SVM 的高压断路器机

械故障诊断[J]. 电力自动化设备, 2020, 40(03): 181-186.

[15]陈佳豪, 吴浩, 胡潇涛, 等. 基于 LMD 和 DE-PNN 的高压断路器机械故障识别方法[J]. 计算机应用与软件, 2023, 40(06): 57-62+165.

[16]王国东, 马莉, 李科云, 等. 基于 LMD 边际谱能量熵的高压断路器机械故障诊断[J].仪器仪表与分析监测, 2024, (01): 17-22.

[17]王茜, 田慕琴, 宋建成, 等. 基于经验小波变换的振动信号特征量提取[J]. 振动与冲击, 2021, 40(16): 261-266.

[18]李永刚, 丁其, 赵书涛. 基于变分模态分解的断路器机械故障诊断[J]. 电测与仪表,2019, 56(20): 82-86+107.

[19]赵科, 杨景刚, 马速良, 等. 基于神经网络集成模型在高压断路器机械故障诊断中的应用[J]. 高压电器, 2018, 54(07): 217-223.

[20]张佳, 陈志英, 陈丽安, 等. 基于粒子群优化极限学习机的断路器故障诊断方法研究[J]. 高压电器, 2020, 56(06): 181-188.

[21]邵阳, 武建文, 马速良, 等. 用于高压断路器机械故障诊断的 AM-ReliefF 特征选择下集成 SVM 方法[J]. 中国电机工程学报, 2021, 41(08): 2890-2901.

[22]黄迅迪, 庞雄文. 基于深度学习的智能设备故障诊断研究综述[J]. 计算机科学, 2023,50(05): 93-102.

[23]陈欣昌, 冯玎, 林圣. 基于深度自编码网络的高压断路器操作机构机械故障诊断方法[J]. 高电压技术, 2020, 46(09): 3080-3088.

[24]曹宇鹏, 罗林, 王乔, 等. 基于卷积深度网络的高压真空断路器机械故障诊断方法[J].电力系统保护与控制, 2021, 49(03): 39-47.

[25]姜楠, 罗林, 王乔, 等. 结构优化深度网络的高压断路器机械故障诊断[J]. 辽宁石油化工大学学报, 2023, 43(03): 91-96.

[26]赵荣超, 吴百礼, 陈祝云, 等. 多尺度时空信息融合驱动的图神经网络故障诊断方法[J]. 华南理工大学学报(自然科学版), 2023,51(12): 42-52.

[27]谭启瑜, 马萍, 张宏立. 基于图卷积神经网络的滚动轴承故障诊断[J]. 噪声与振动控制, 2023, 43(06): 101-108+116.

[28]赵志宏, 张然, 刘克俭, 等. 一种改进原型网络的小样本轴承故障诊断方法[J]. 振动与冲击, 2023, 42(20): 214-221.

[29]宋业栋, 马光伟, 裴国斌, 等. 基于深度迁移学习的柴油机故障诊断研究[J]. 振动与冲击, 2023, 42(21): 219-226.

[30]徐琛苑. 基于高压断路器振动信号的特征提取与故障诊断研究[D]. 南京理工大学,2021.

[31]Li H, Liu T, Wu X, et al. An optimized VMD method and its applications in bearing fault diagnosis[J]. Measurement, 2020, 166: 108185.

[32]Abdel-Basset M, Mohamed R, Abouhawwash M. Crested Porcupine Optimizer: A new nature-inspired metaheuristic[J]. Knowledge-Based Systems, 2024, 284: 111257.

[33]杨远鹏, 陈志刚, 余志红, 等. 基于 WOA-VMD 与 PSO-SVM 的滚动轴承故障诊断

[J]. 制造技术与机床, 2025, (02): 23-29+42.

[34]李敬一, 苏翔. GWO 优化 CNN-BiLSTM-Attenion 的轴承剩余寿命预测方法[J]. 振动与冲击, 2025, 44(02): 321-332.

[35]Liu B, Liu C, Zhou Y, et al. A chatter detection method in milling based on gray wolf optimization VMD and multi-entropy features[J]. The International Journal of Advanced Manufacturing Technology, 2023, 125(1): 831-854.

[36]Tan Y, Hu E, Liu Y, et al. Review of digital vibration signal analysis techniques for fault diagnosis of high-voltage circuit breakers[J]. IEEE Transactions on Dielectrics and Electrical Insulation, 2023, 31(1): 404-418.

[37]Liu Z, Liu H. A novel hybrid model based on GA-VMD, sample entropy reconstruction and BiLSTM for wind speed prediction[J]. Measurement, 2023, 222: 113643.

[38]Tan H, Xie S, Liu R, et al. Fatigue condition diagnosis of rolling bearing based on normalized balanced multiscale sample entropy[J].

International Journal of Fatigue, 2023, 172: 107642.

[39]李伟, 王付广, 王东生. 基于改进复合多尺度样本熵的行星齿轮箱故障诊断[J]. 航空

动力学报, 2024, 39(09): 336-345.

[40]Watanabe K, Maeda K, Ogawa T, et al. A Gaussian Process Decoder with Spectral Mixtures and a Locally Estimated Manifold for Data Visualization[J]. Applied Sciences, 2023, 13(14): 8018.

[41]刘兆赓. 面向样本间潜在关系挖掘的图学习方法研究[D]. 吉林大学, 2024.

[42]苏易礌, 李卫军, 刘雪洋, 等. 基于图神经网络的文本分类方法研究综述[J]. 计算机工程与应用, 2024, 60(19): 1-17.

[43]周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(06):1229-1251.

[44]Bhatti U A, Tang H, Wu G, et al. Deep learning with graph convolutional networks: An overview and latest applications in computational intelligence[J]. International Journal of Intelligent Systems, 2023, 2023(1): 8342104.

[45]Zhang S, Tong H, Xu J, et al. Graph convolutional networks: a comprehensive review[J].Computational Social Networks, 2019, 6(1): 1-23.

[46]Tang S, Li B, Yu H. ChebNet: efficient and stable constructions of deep neural networks with rectified power units via Chebyshev approximation[J]. Communications in Mathematics and Statistics, 2024: 1-27.

[47]袁科研, 兰媛, 黄家海, 等. 基于 GCN 的轴向柱塞泵故障诊断方法[J]. 振动.测试与诊断, 2025, 45(01): 140-145+205.

[48]Velickovic P, Cucurull G, Casanova A, et al. Graph attention networks[J]. stat, 2017, 1050(20): 10-48550.

[49]Ding C, Sun S, Zhao J. MST-GAT: A multimodal spatial–temporal graph attention network for time series anomaly detection[J]. Information Fusion, 2023, 89: 527-536.

[50]Vrahatis A G, Lazaros K, Kotsiantis S. Graph attention networks: a comprehensive review of methods and applications[J]. Future Internet, 2024, 16(9): 318.

[51]谭启瑜, 马萍, 张宏立, 等. 基于图注意力网络的风力发电机齿轮箱故障诊断[J]. 太阳能学报, 2024, 45(01): 265-274.

[52]梁秀. 基于小样本学习的有限数据轴承故障诊断研究[D]. 石家庄铁道大学,2024.

[53]李金科. 元学习在旋转机械故障诊断中的应用综述[J]. 科技与创新, 2023,(18): 176-178.

[54]Zhang A, Li S, Cui Y, et al. Limited data rolling bearing fault diagnosis with few-shot learning[J]. Ieee Access, 2019, 7: 110895-110904.

[55]Li C, Li S, Zhang A, et al. Meta-learning for few-shot bearing fault diagnosis under complex working conditions[J]. Neurocomputing, 2021, 439: 197-211.

[56]Liu Z, Chen Y, Zhang Y, et al. Diagnosis of arrhythmias with few abnormal ECG samples

using metric-based meta learning[J]. Computers in biology and medicine, 2023, 153: 106465.

[57]Sireesha Chamarthi, Katharina Fogelberg, Jakob Gawlikowski, et al. Few-shot learning for skin lesion classification: A prototypical networks approach[J]. Informatics in Medicine Unlocked, 2024, 48: 101520.

[58]胡宏俊. 基于 ACMD 和元学习的柱塞泵故障诊断方法研究[D]. 中北大学, 2024.

[59]Chicco D. Siamese neural networks: An overview[J]. Artificial neural networks, 2021: 73-94.

[60]王泽栋. 融合匹配网络与原型网络的小样本关系抽取研究与应用[D]. 华东师范大学, 2023.