| 引用本文: | 曲朝阳,刘谊豪,曲 楠,等.基于声纹脊线化和元学习的变压器故障诊断方法[J].电力系统保护与控制,2025,53(13):163-174.[点击复制] |
| QU Zhaoyang,LIU Yihao,QU Nan,et al.Transformer fault diagnosis method based on acoustic ridging and meta-learning[J].Power System Protection and Control,2025,53(13):163-174[点击复制] |
|
| 摘要: |
| 针对变压器声纹检测中信号易受干扰且足量样本获取困难的问题，提出一种融合声纹脊线化与元学习的变压器声纹诊断方法。首先，基于脊线化特征处理，对优化后的变压器声纹时频谱图进行物理特征筛选与形态特征压缩。然后，搭建选择性编码器(selective encoder, SE)加深时频与形态表征的关联度，提升模型收敛速度。最后，构造元学习网络评估变压器状态，并引入基于OD-Reptile的一阶梯度更新策略，通过内外循环优化机制增强参数泛化性，从而实现少样本、信息干扰条件下的高精度声纹诊断。相较于R-WDCNN、LSTM、CNN等传统深度学习信号诊断方法，该方法在低样本、高噪声环境下(SNR为-12 dB)，收敛轮数减少10轮以上。同时，准确率分别提高6.35%，12.1%和16.93%。实验结果显示，所提方法在准确性、抗噪性、鲁棒性以及泛化性方面均有显著提升。 |
| 关键词: 声纹 小样本 脊线化 时频谱图 选择性编码 元学习 故障诊断 |
| DOI：10.19783/j.cnki.pspc.241186 |
| 投稿时间：2024-09-07修订日期：2024-11-30 |
| 基金项目:国家自然科学基金项目资助(52377081) |
|
| Transformer fault diagnosis method based on acoustic ridging and meta-learning |
| QU Zhaoyang1,2,LIU Yihao1,QU Nan3,JIANG Tao1,XU Xiaoyu1 |
| (1. Northeast Electric Power University, Jilin 132012, China; 2. Jilin Electric Power Big Data Intelligent Processing Engineering
Technology Research Center, Jilin 132012, China; 3. State Grid Nanjing Power Supply Company, Nanjing 210000, China) |
| Abstract: |
| To address the challenges in transformer voiceprint detection, namely, signal susceptibility to interference and difficulty in obtaining sufficient training samples, a transformer voiceprint fault diagnosis method that integrates acoustic pattern ridging and meta-learning is proposed. First, based on the ridge feature processing, the optimized transformer voice pattern time spectrum is selected for physical feature screening and morphological feature compression. A selective encoder (SE) is then constructed to enhance the correlation between time-frequency and morphological representations, improving model convergence speed. Finally, a meta-learning network is designed for transformer state evaluation. An OD-Reptile-based first-order gradient update strategy is introduced, and an inner-outer loop optimization mechanism is used to enhance parameter generalization. This enables accurate voiceprint diagnosis under limited sample and noisy conditions. Compared with traditional deep learning methods, such as R-WDCNN, LSTM, and CNN, the proposed method reduces the number of convergence iterations by more than 10 rounds in low-sample, high-noise environments (SNR is -12 dB). Accuracy is improved by 6.35%, 12.1%, and 16.93%, respectively. Experimental results demonstrate significant improvements in accuracy, noise immunity, robustness, and generalization. |
| Key words: voiceprint small samples ridging time-frequency spectrograms selective encoding meta-learning fault diagnosis |
