| 基于RMCA-CNN和同步相量的风电场次/超同步振荡参数智能辨识方法 |
| 点此下载全文 |
| 引用本文：潘中昊1，2,冯双1，2,陆友文2,梅悦2,陈力1,洪希1.基于RMCA-CNN和同步相量的风电场次/超同步振荡参数智能辨识方法[J].电网与清洁能源,2025,41(4):34~42 |
| 摘要点击次数: 6 |
| 全文下载次数: 0 |
|
| 基金项目:国家自然科学基金资助项目（52377084）；电网运行风险防御技术与装备全国重点实验室资助项目（SGNR0000KJJS2302237） |
|
| 中文摘要:基于同步相量正频谱的传统辨识方法由于频谱混叠无法辨识超同步振荡的参数。为此，基于理论分析证明基于同步相量正负频谱能够同时辨识次/超同步分量。提出了基于残差多通道注意力卷积神经网络模型和同步相量正负频谱分量的次/超同步振荡参数辨识方法。将多通道注意力机制嵌入卷积神经网络中提高模型对多振荡参数的学习能力，引入残差模块解决深层卷积神经网络的梯度消失和网络退化问题，通过对预训练模型进行迁移学习，在较少样本下拓宽了参数辨识模型的辨识频段。仿真结果表明，所提方法能够准确辨识同时包含次、超同步振荡分量的同步相量参数。 |
| 中文关键词:卷积神经网络 同步相量 次/超同步振荡 参数辨识 迁移学习 多通道注意力 |
| |
| An Intelligent Identification Method of Wind Farm Sub-Synchronous/Super-Synchronous Oscillation Parameters Based on RMCA-CNN and Synchrophasor |
|
|
| Abstract:The traditional identification method based on positive spectra of synchrophasor has difficulty in identifying the parameters of super-synchronous oscillations due to spectral aliasing. To address this， the theoretical analysis in this paper firstly proves that it is possible to identify the sub/super-synchronous components simultaneously based on the positive and negative spectral components of the synchrophasor. On this basis， the paper proposes a sub/super-synchronous oscillation parameters identification method based on residual multi-channel attention convolutional neural network and the positive and negative spectral components of the synchrophasor， which embeds the multi-channel attention mechanism into the convolutional neural network to improve the model's learning ability， introduces the residual module to address the gradient disappearance and network degradation of the deep convolutional neural network in the updating of the weight parameters， and finally broadens the band of the parameters identification with fewer samples by migrating the knowledge which have learned from the pre-trained model.The simulation results show that the proposed method is able to accurately identify the parameters of the synchrophasor which contains both sub-synchronous and super-synchronous oscillation components. |
| keywords:convolutional neural networks synchrophasor sub/super-synchronous oscillations parameter discrimination transfer learning multi-channel attention |
| 查看全文 查看/发表评论 下载PDF阅读器 |
