| 引用本文: | 王有鹏,曾祥君,刘 丰,等.基于行波全频带特征的配电网故障行波波头标定方法[J].电力系统保护与控制,2025,53(1):171-180.[点击复制] |
| WANG Youpeng,ZENG Xiangjun,LIU Feng,et al.Fault traveling wave head calibration method for a distribution network based on the full band characteristics of a traveling wave[J].Power System Protection and Control,2025,53(1):171-180[点击复制] |
|
| 摘要: |
| 针对配电网行波波头标定方法易受噪声、波头畸变影响的问题，提出一种基于行波全频带特征的配电网故障行波波头标定方法。首先，根据行波高频段分量包含奇异点特征、行波中低频段分量不受噪声干扰的特点，提出利用行波全频带分量特征来标定行波，并分析了不同工况下利用行波全频带分量特征标定波头的优势。然后，设计并搭建基于目标检测模型的卷积神经网络(convolutional neural network, CNN)，以行波全频带分量作为特征输入量，利用一维卷积核提取行波信号的波头特征。最后，结合特征金字塔网络与路径聚合网络结构，融合行波波头高中低频带特征，实现行波到达时刻的准确标定。与传统方法相比，所提方法在短线路、强噪声情况下具有较强的适应性，并且在微弱故障行波场景下也能够实现波头标定，具有良好的现场应用效果。 |
| 关键词: 配电网 波头标定 行波全频带 目标检测模型 |
| DOI：10.19783/j.cnki.pspc.240597 |
| 投稿时间：2024-05-14修订日期：2024-09-17 |
| 基金项目:国家自然科学基金联合基金重点支持项目资助(U22B20113)；国家自然科学基金项目资助(52407078)；南方电网公司数字研究院有限公司科技项目资助(210002KK 52222011) |
|
| Fault traveling wave head calibration method for a distribution network based on the full band characteristics of a traveling wave |
| WANG Youpeng1,ZENG Xiangjun1,LIU Feng1,LIU Feng1,JIANG Zhuang1,YU Kun1,XIE Liwei1,LI Xiaobo2 |
| (1. State Key Laboratory of Disaster Prevention & Reduction for Power Grid (Changsha University of Science and Technology),
Changsha 410114, China; 2. China Southern Power Grid Digital Research Institute Co., Ltd., Guangzhou 510700, China) |
| Abstract: |
| There is a problem in that the traveling wave head calibration method of a distribution network is easily affected by noise and wave head distortion. Thus a fault traveling wave calibration method based on the full-band characteristics of the traveling wave is proposed. From the singularities of the high frequency component and the noise immunity of the mid-low frequency component of the traveling wave, the full frequency component of the traveling wave is used to calibrate the traveling wave, and the advantages of using the full frequency component of the traveling wave in different working conditions are analyzed. Then, a convolutional neural network (CNN) based on a target detection model is designed and built. The full-band component of the traveling wave is taken as the feature input, and the wave head features of the traveling wave signal are extracted using a one-dimensional convolution kernel. Finally, the feature pyramid network and path aggregation network structure are combined to integrate the high, middle and low band features of the traveling wave head, and the accurate calibration of the arrival time of the traveling wave is realized. Compared with the traditional method, the proposed method has stronger adaptability in the case of short line and strong noise, and can also realize wave head calibration in the weak fault traveling wave scenario, and has good field application. |
| Key words: distribution network wave head calibration traveling wave full band object detection model |
