半月刊

ISSN 1000-1026

CN 32-1180/TP

+高级检索 English
变压器励磁涌流多角度时频特征综合辨识方法
DOI:
作者:
作者单位:

1.长沙理工大学电气与信息工程学院;2.国网益阳供电公司;3.湖南大学电气与信息工程学院;4.广东电网有限责任公司中山供电局

摘要:

配电网中电力电子器件的不断接入,导致系统谐波电流日益加大,传统变压器二次谐波制动的差动保护面临挑战;同时,单一特征辨识方法受分布式电源类型和合闸角影响,无法准确区分不同场景下的故障电流和励磁涌流。为了提高励磁涌流的辨识准确率,本文提出多角度时频分析方法,全面整合时域、频域和时频域特征,利用贝叶斯算法优化极度梯度上升(extreme gradient boosting, XGBoost)的分类参数,提高模型的泛化能力,实现不同容量、不同类型分布式电源接入下的故障电流与励磁涌流的准确辨识;采用shapley additive explanations(SHAP)值分析方法,揭示各特征值在辨识模型中的贡献度。通过PSCAD/EMTDC仿真数据以及现场实测数据对所提方法进行验证,在本文提供的数据样本内,多角度时频分析下的贝叶斯-XGBoost算法励磁涌流辨识准确率接近100%,优于文中所对比的几种常见分类算法。

关键词:

基金项目:

湖南省自然科学基金优秀青年项目(2023JJ20039),国家自然科学基金(52007009),南方电网公司科技项目资助(GDKJXM20231017)

通信作者:

作者简介:


Identification Method of Multi-perspective Time-frequency Characteristic Synthesis for Transformer Excitation Inrush Current
Author:
Affiliation:

Abstract:

The continuous integration of power electronic devices in distribution networks has led to an increasing level of harmonic currents, posing challenges for traditional transformer secondary harmonic restraint differential protection. Simultaneously, single-feature identification methods are influenced by distributed energy resource types and closing angles, making it difficult to accurately distinguish fault currents and excitation inrush currents in different scenarios. To enhance the accuracy of excitation inrush current identification, this paper proposes a multi-angle time-frequency analysis method that comprehensively integrates time-domain, frequency-domain, and time-frequency-domain features. It utilizes Bayesian optimization of XGBoost (extreme gradient boosting) classification parameters to improve model generalization, enabling accurate identification of fault currents and excitation inrush currents under various capacities and types of distributed energy resource integration. The SHAP (shapley additive explanations) value analysis method is employed to reveal the contribution of each feature value in the identification model. The proposed method was verified through PSCAD/EMTDC simulation data and field measured data. Within the data samples provided in this article, the Bayesian-XGBoost algorithm under multi-angle time-frequency analysis has an accuracy of identification of excitation inrush current close to 100%, which is better than several common classification algorithms compared in this paper.

Keywords:

Foundation:
Natural Science Foundation for Excellent Youth of Hunan Province, China (No.2023JJ20039); National Natural Science Foundation of China (No.52007009); Science and Technology Project Foundation of Southern Power Grid Corporation of China (GDKJXM20231017)
引用本文
[1]陈春,占露昕,曹伯仲,等.变压器励磁涌流多角度时频特征综合辨识方法[J/OL].电力系统自动化,http://doi. org/[doi].
Chen Chun, ZHAN Luxin, CAO Bozhong, et al. Identification Method of Multi-perspective Time-frequency Characteristic Synthesis for Transformer Excitation Inrush Current[J/OL]. Automation of Electric Power Systems, http://doi. org/[doi].
复制
支撑数据及附录
分享
历史
  • 收稿日期:2024-07-23
  • 最后修改日期:2024-11-23
  • 录用日期:2024-11-25
  • 在线发布日期:
  • 出版日期: