基于特征分析和模式识别的端子箱故障预警算法与模型  

Fault early warning algorithm and model for terminal boxes based on feature analysis and pattern recognition

作　　者：万雄彪 梁仕斌 吴桂鸿 沐润志 伍阳阳 张瑀明 杨文睿 WAN Xiongbiao;LIANG Shibin;WU Guihong;MU Runzhi;WU Yangyang;ZHANG Yuming;YANG Wenrui(Yunnan Electric Power Test&Research Institute(Group)Co.,Ltd.,Kunming 651027,China)

机构地区：[1]云南电力试验研究院(集团)有限公司,云南昆明651027

出　　处：《电气应用》2025年第4期120-125,共6页Electrotechnical Application

摘　　要：随着电力系统规模的不断扩展,传统的端子箱故障检测方法已难以满足现代电网运行的需求。提出了一种基于特征分析与模式识别的端子箱故障预警算法,结合多模态数据融合、时频分析和深度学习技术,能够有效地识别端子箱的故障类型。首先,提取端子箱状态数据中的故障特征,利用加权数据融合方法生成综合特征向量。接着,结合自适应卷积神经网络(ACNN)与时序优化长短期记忆网络(T-LSTM)进行多模态特征的时空建模,捕捉数据中的动态变化。然后,引入了自适应动态阈值调整机制,使得预警系统能够根据实时数据的波动,动态调整故障判断阈值。最后,通过仿真实验,评估了模型在不同故障类型下的性能,结果表明该模型具有较高的准确率(96.2%)和召回率(94.5%),显著提高了故障预警的准确性与实用性。With the continuous expansion of power systems,traditional fault detection methods for terminal boxes are no longer sufficient for modern grid operations.This paper proposes a fault early warning algorithm for terminal boxes based on feature analysis and pattern recognition,which integrates multi-modal data fusion,time-frequency analysis,and deep learning techniques to effectively identify fault types.First,fault features are extracted from terminal box state data,and a weighted data fusion method is applied to generate a comprehensive feature vector.Then,the combination of Adaptive Convolutional Neural Network(ACNN)and Temporal-Optimized Long Short-Term Memory Network(T-LSTM)models is used to perform spatiotemporal modeling of multi-modal features,capturing dynamic changes in the data.An adaptive dynamic threshold adjustment mechanism is introduced,allowing the warning system to adjust the fault detection threshold based on real-time data fluctuations.Through simulation experiments,the performance of the model was evaluated for different fault types,and the results demonstrate that the model achieves a high accuracy rate(96.2%)and recall rate(94.5%),significantly improving fault warning accuracy and practical applicability.

关 键 词：端子箱 故障预警 特征分析 模式识别 

分 类 号：TM63[电气工程—电力系统及自动化]