PDF(4596 KB)
PDF(4596 KB)
PDF(4596 KB)
基于TF-PARNet的有源配电网轻量化故障诊断
A Lightweight Fault Diagnosis Method for TF-PARNet-Based Active Distribution Networks
【目的】针对有源配电网并网点故障信号非平稳特征提取困难，以及现有深度学习模型参数冗余难以满足边缘智能终端实时部署需求的难题，提出一种基于轻量化时频双分支金字塔注意力网络（time-frequency dual-branch pyramid attention residual network， TF-PARNet）的故障诊断方法。【方法】首先，构建时域与频域双分支并行架构，并引入差异化多头自注意力机制，分别捕获暂态波形突变与全局频谱依赖；其次，设计基于门控机制的智能自适应融合模块，实现异构时频特征的动态加权与重标定；最后，利用深度可分离卷积重构金字塔残差单元，解耦空间与通道相关性以实现模型极致压缩。【结果】在PSCAD/EMTDC平台上搭建IEEE 13节点系统进行仿真。测试结果表明，在叠加高斯噪声与时序偏移的复杂工况下，TF-PARNet 的平均准确率达96.14%，模型参数量仅为151.0 ×103个，在纯CPU环境下的单样本推理时间低至3.8 ms。【结论】该方法在保证高保真特征解析的同时大幅削减了计算开销，兼具故障辨识的高准确性与边缘计算的高效性，为资源受限的配电网就地化智能保护提供了可靠的工程方案。
[Objective] To address the challenges of extracting non-stationary features from fault signals at the Point of Common Coupling in active distribution networks， and to overcome the redundancy of existing deep learning models that hinders real-time deployment on edge intelligent terminals， a fault diagnosis method based on a lightweight time-frequency dual-branch pyramid attention residual network （TF-PARNet） is proposed. [Methods] First， a time-frequency dual-branch parallel architecture was constructed， incorporating a differentiated multi-head self-attention mechanism to capture transient waveform mutations and global spectral dependencies， respectively. Second， an intelligent adaptive fusion module based on a gating mechanism was designed to achieve dynamic weighting and recalibration of heterogeneous time-frequency features. Finally， the pyramid residual unit was reconstructed using depthwise separable convolutions to decouple spatial and channel correlations， achieving extreme model compression. [Results] Simulations were conducted on an IEEE 13-node system using the PSCAD/EMTDC platform. The test results demonstrate that under complex working conditions with superimposed Gaussian noise and temporal offsets， TF-PARNet achieves a mean accuracy of 96.14%. The model contains only 151.0 × 103 parameters， and the single-sample inference time in a pure CPU environment is as low as 3.8 ms. [Conclusion] The proposed method reduces computational overhead while ensuring high-fidelity feature analysis. It combines high diagnostic accuracy with edge computing efficiency， providing a reliable engineering solution for the localized intelligent protection of resource-constrained distribution networks.
有源配电网 / 故障诊断 / 时频分析 / 自适应融合 / 轻量化网络
active distribution network / fault diagnosis / time-frequency analysis / adaptive fusion / lightweight network
| [1] |
侯思祖, 金雅慧. 基于谐波特性与ConvNeXt网络的配电网故障区段定位方法[J]. 供用电, 2025, 42(11): 67-76.
|
| [2] |
何小龙, 高红均, 王仁浚, 等. 基于图深度强化学习的有源配电网故障恢复方法[J]. 电网技术, 2025, 49(10): 4342-4352.
|
| [3] |
陈继明, 陈文淙, 仉志华, 等. 基于波形相似性的柔性互联配电网短路故障区段定位方法[J]. 电力工程技术, 2025, 44(1): 104-114.
|
| [4] |
林湘宁, 刘胜福, 翁汉琍, 等. 可实现故障快速响应和全面覆盖的有源高压配电网单端暂态量阶段式保护方案[J]. 广东电力, 2024, 37(10): 121-134.
|
| [5] |
徐小健, 于飞. 基于输出电压轨迹的三相逆变器开关管开路故障诊断[J]. 中国电机工程学报, 2024, 44(3): 1106-1116.
|
| [6] |
|
| [7] |
齐郑, 黄朝晖, 陈艳波. 基于零序分量的阻抗法配电网故障定位技术[J]. 电力系统保护与控制, 2023, 51(6): 54-62.
|
| [8] |
|
| [9] |
付华, 王婧羽. 基于SR-VMD的微弱故障行波检测方法[J]. 电力系统保护与控制, 2021, 49(1): 156-162.
|
| [10] |
|
| [11] |
吴佳享, 孙云莲, 陈楚昭. 基于MRSVD-VMD的多分支配电网故障定位[J]. 电力系统及其自动化学报, 2022, 34(11): 58-65.
|
| [12] |
张广斌, 陈柏宇, 束洪春, 等. 基于时域波形特征认知的输电线路近端故障辨识与定位[J]. 电力系统自动化, 2024, 48(5): 146-156.
|
| [13] |
周军, 曲振国, 战紫昂. 基于CPO-VMD的输电线路行波故障测距方法[J]. 电力科学与技术学报, 2025, 40(2): 30-41.
|
| [14] |
张帝, 宋兴荣, 唐海国, 等. 基于自适应VMD和WVD的配电网故障行波检测方法[J]. 电力科学与技术学报, 2024, 39(2): 80-90.
|
| [15] |
张大海, 张晓炜, 孙浩, 等. 基于卷积神经网络的交直流输电系统故障诊断[J]. 电力系统自动化, 2022, 46(5): 132-140.
|
| [16] |
|
| [17] |
翁楦乔, 文成林. 基于深度特征聚类和RNN的电网故障诊断[J]. 控制工程, 2022, 29(1): 175-181.
|
| [18] |
胡登宇, 王宝华, 刘晋宏. 基于改进图神经网络的含源配电网故障诊断方法及效果[J]. 科学技术与工程, 2025, 25(21): 8936-8944.
|
| [19] |
胡国栋, 马宏忠, 孙维, 等. 融合双向时序特征与多头注意力机制的电缆故障定位方法[J]. 广东电力, 2025, 38(12): 33-43.
|
| [20] |
|
| [21] |
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.
|
| [22] |
|
| [23] |
石昱烜, 席燕辉, 张伟杰. 基于GAF-ResNet50的配电网故障区段定位[J]. 电力科学与技术学报, 2025, 40(2): 122-130, 149.
|
| [24] |
叶远波, 王吉文, 邵庆祝, 等. 配电网故障识别Transformer-联邦迁移学习算法设计[J]. 电力系统及其自动化学报, 2025, 37(10): 120-128.
|
| [25] |
张彼德, 邱杰, 娄广鑫, 等. 基于CNN和Transformer的轻量化电能质量扰动识别模型[J]. 电力工程技术, 2025, 44(1): 69-78.
|
| [26] |
秦苏亚, 薛永端, 刘砾钲, 等. 有源配电网小电流接地故障暂态特征及其影响分析[J]. 电工技术学报, 2022, 37(3): 655-666.
|
| [27] |
王宾, 崔鑫, 董新洲. 配电线路弧光高阻故障检测技术综述[J]. 中国电机工程学报, 2020, 40(1): 96-107, 377.
|
| [28] |
黄昕飞, 王有鹏, 刘凤, 等. 利用行波主频分量幅值信息的多分支配电网故障定位方法[J]. 电力系统及其自动化学报, 2025, 37(12): 15-23.
|
| [29] |
张小凤, 李昂, 周雷金, 等. 基于AIBPSO算法的含DG主动配电网分区故障定位[J]. 供用电, 2024, 41(12): 93-101.
|
| [30] |
余波, 高学军, 王灿, 等. 基于行波特征与KOA-CNN-BiGRU-AM的柔直输电线路故障诊断[J]. 电力工程技术, 2025, 44(2): 185-196.
|
| [31] |
孟宏宇, 张建良, 蔡兆龙, 等. 基于CNN-BiLSTM-Attention的直流微电网故障诊断研究[J]. 中国电机工程学报, 2025, 45(4): 1369-1380.
|
| [32] |
白星振, 宋昭杉, 葛磊蛟, 等. 含分布式电源的复杂配电网相间故障定位等效解耦模型[J]. 电力建设, 2022, 43(2): 45-53.
分布式电源规模化接入,使配电网拓扑结构和组成元素越来越复杂,传统矩阵算法由于馈线开关处故障电流方向难以确定已不再适用于配电网相间故障定位。为有效将复杂配电网的拓扑结构简化,基于深度优先原则对配电网进行解耦,使其变成由若干个树干状结构组成的网络;针对复杂配电网故障定位矩阵算法计算量大、步骤复杂的难题,提出了一种新的复杂配电网故障定位矩阵算法,与传统矩阵算法相比,该方法对配电网T型区段及末梢区段故障的定位更加精确。最后,通过典型园区配电网模型与改进的 IEEE 33系统案例,验证了所提方法既能实现配电网的故障定位,又能简化复杂配电网结构,与以往的行波法相比,文章提出的方法缩短了定位时间,具有较好的工程应用价值。
Since distributed power supply is connected to distribution networks on a large scale, its topology and components are becoming more and more complex. As the direction of fault current at feeder switches is difficult to determine, a new matrix algorithm for equivalent decoupling fault location in complex distribution networks is proposed in this paper. In order to simplify the topology of complex distribution network effectively, the distribution network is decoupled according to the principle of depth priority to make it a network composed of tree trunks. A new fault-location matrix algorithm for complex distribution network is presented to solve the problem of large computation and complex steps in fault-location matrix algorithm for complex distribution network. Compared with traditional matrix algorithm, this method is more accurate for T-section and terminal section of power network fault location. Finally, through the typical park distribution network model and the improved IEEE 33-node system case, the method proposed in this paper can not only locate the faults in the distribution network, but also simplify the complex distribution network structure. Compared with the traveling wave method, the proposed method shortens the locating time and has better application value. |
| [33] |
周军, 曲振国, 战紫昂. 融合小波去噪与VMD-NTEO的输电线路故障定位[J]. 电力系统及其自动化学报, 2025, 37(7): 141-150.
|
| [34] |
|
| [35] |
|
| [36] |
|
| [37] |
|
| [38] |
|
| [39] |
When a fault arises in an active distribution network, both the grid upstream of the fault point and the distributed generation (DG) downstream of the fault point contribute short-circuit currents, posing challenges for existing centralized feeder automation (FA) technologies in accurately locating and isolating the fault. Moreover, the radial network structure of the distribution network can lead to power outages for downstream loads when the DG downstream of the fault point ceases operation due to islanding effects. To tackle this issue, this paper introduces a communication-based distributed FA strategy that employs interlocking signals between adjacent switches to pinpoint and isolate faults in active distribution networks, effectively minimizing the scope of fault-induced power outages. Furthermore, to guarantee uninterrupted power supply to the non-fault zones located downstream of the fault point, an inverter control strategy integrating voltage negative feedback control, reactive power negative feedback control, and quasi-synchronization grid-connection control is proposed. This strategy allows for seamless switching between grid-following and grid-forming inverter control modes, enabling adaptive grid-connection/islanding control of DGs. Consequently, non-fault areas downstream of the fault point maintain uninterrupted power supply, significantly enhancing the reliability of power supply in distribution networks. Simulation models have been constructed using MATLAB/Simulink to validate the efficacy of the proposed approach.
|
| [40] |
舒佳蕾, 陈依林, 曹虹, 等. 基于图注意力网络的配电网故障行波定位方法[J]. 电力科学与技术学报, 2025, 40(1): 85-91.
|
| [41] |
|
| [42] |
|
| [43] |
|
| [44] |
Non-predictive or inaccurate weather forecasting can severely impact the community of users such as farmers. Numerical weather prediction models run in major weather forecasting centers with several supercomputers to solve simultaneous complex nonlinear mathematical equations. Such models provide the medium-range weather forecasts, i.e., every 6 h up to 18 h with grid length of 10–20 km. However, farmers often depend on more detailed short-to medium-range forecasts with higher-resolution regional forecasting models. Therefore, this research aims to address this by developing and evaluating a lightweight and novel weather forecasting system, which consists of one or more local weather stations and state-of-the-art machine learning techniques for weather forecasting using time-series data from these weather stations. To this end, the system explores the state-of-the-art temporal convolutional network (TCN) and long short-term memory (LSTM) networks. Our experimental results show that the proposed model using TCN produces better forecasting compared to the LSTM and other classic machine learning approaches. The proposed model can be used as an efficient localized weather forecasting tool for the community of users, and it could be run on a stand-alone personal computer.
|
| [45] |
|
| [46] |
顾衍璋, 阳浩, 雷一勇, 等. 面向轻量级芯片化终端的配电网态势感知增强方法[J/OL]. 电力系统及其自动化学报, 2026-03-06. https://doi.org/10.19635/j.cnki.csu-epsa.001804.
|
| [47] |
|
| [48] |
The high-level proliferation of inverter-interfaced distributed generators (IIDGs) in modern distribution networks (DNs) has changed system topologies and fault current signatures, which compromises the protective relays in DNs. Investigating IIDG fault behaviors-based protection scheme will benefit the grid’s safety and stability. This paper proposes a novel current differential protection (CDP) scheme that considers the delay behaviors of positive- and negative-sequence component extractors for IIDGs in DNs. A frequency-domain analytical model of the fault current for a grid-connected IIDG with the PQ control strategy and a low-voltage ride-through (LVRT) capability is investigated. The dynamic behavior of the IIDGs considering the sequence-component extractor based on the Pade approximation is presented, where the T/4 delay extractor of the IIDGs causes a two-stage behavior in the fault transient process. It is found that a 5 ms error between the measured and actual values after the fault will affect the transient characteristics of the IIDGs. The transient current generated by the IIDGs during grid faults contains a large number of low-order harmonic components within the range of 0–200 Hz, which is significantly different to the current provided by the power grid. Therefore, the proposed CDP scheme uses protective relays at both terminals to obtain the required transient electric quantity using the Prony method. By constructing the frequency-characteristics ratio (FCR) and the exchanging FCR between two terminal relays, the developed protection criteria are implemented. The accuracy of the fault analysis method, whose maximum computational error is below 0.1%, and the feasibility of the proposed protection scheme are demonstrated by using a 10 kV DN in a PSCAD/EMTDC simulation, which can be applied to various fault conditions and traditional DNs without IIDGs.
|
| [49] |
Effective diagnosis of distribution network faults is crucial to ensuring the reliability of power systems. However, the bidirectional power flow caused by the integration of new energy limits the effectiveness of traditional detection methods. Although data-driven approaches are not restricted by power flow direction, their performance is heavily dependent on the quantity and quality of training samples. In addition, factors such as measurement noise, variable fault impedance, and volatile photovoltaic output complicate fault information. To address this, we present a new fault diagnosis model named the dynamic, adaptive, and coupled dual-field-encoding graph neural network (DACDFE-GNN), which introduces a dynamic aggregation module to assign different weights to reduce noise interference and fully integrates information from observable nodes. On this basis, the coupled dual-field-encoding module is proposed, which encodes topological information and physical–electrical domain information as part of the initial features, thereby capturing fault features and learning the law of feature propagation. The experimental results for the IEEE 34- and IEEE 123-node feeder systems indicate that the proposed model surpasses recent fault diagnosis methods in detection performance, particularly regarding its low training sample rate.
|
所有作者声明不存在利益冲突。
储云迪负责研究构思与研究指导；贺梦洁负责模型实现、实验验证、数据分析及论文初稿撰写；侯世玺负责研究监督及论文审阅与修订；吕朋蓬、徐挺负责工程场景分析与结果验证；史朋飞负责研究指导及论文审定。所有作者均阅读并同意了论文终稿内容。
AI小编
/
| 〈 |
|
〉 |
AI 导读
