摘要
由于变压器有载分接开关操作频繁且工作恶劣,导致出现多种故障问题,影响变压器的正常运行和稳定性。为此,提出基于改进神经网络的变压器有载分接开关故障自动化诊断系统。依据系统的基本要求与实现功能,设计四层架构,并采用加速度传感器和工控机作为故障诊断系统的硬件部分,基于此,通过振动信号时频域转换和分解,提取故障特征量,并引入神经网络算法构建故障诊断模型,并对其进行参数优化与改进,将提取的故障特征量作为输入数据,通过概率计算输出变压器有载分接开关的故障类型,以此实现故障自动化诊断。实例应用结果显示,所设计的系统可以较为准确地诊断变压器有载分接开关的故障类型,故障诊断误报率低于2.5%,诊断精度较高。
Due to frequent and harsh operation of on load tap changers in transformers,various fault problems have occurred,affecting the normal operation and stability of transformers.Therefore,an automated diagnosis system for transformer on load tap changer faults based on improved neural networks is proposed.Based on the basic requirements and implementation functions of the system,a four layer architecture is designed,and acceleration sensors and industrial control computers are used as the hardware parts of the fault diagnosis system.Based on this,fault feature quantities are extracted through time-frequency domain conversion and decomposition of vibration signals,and neural network algorithms are introduced to construct a fault diagnosis model.Parameter optimization and improvement are carried out,and the extracted fault feature quantities are used as input data to output the fault type of the on load tap changer of the transformer through probability calculation,thereby achieving automatic fault diagnosis.The application results of the example show that the designed system can accurately diagnose the fault types of on load tap changers in transformers,with a false alarm rate of less than 2.5%and high diagnostic accuracy.
作者
李朋宇
国伟辉
闫帅
李强
刘子恩
何山
LI Pengyu;GUO Weihui;YAN Shuai;LI Qiang;LIU Zien;HE Shan(School of Electric Engineering,Xinjiang University,Urumqi 830049 China;State Grid Anhui Electric Power Co.,Ltd.Bozhou Power Supply Company,Bozhou,Anhui 236800,China;SGCC Key Laboratory for Sulfur Hexafluoride Gas Analysis and Purification,Hefei 230022,China)
出处
《自动化与仪器仪表》
2025年第7期68-72,共5页
Automation & Instrumentation
关键词
改进神经网络
变压器有载分接开关
故障诊断
诊断模型
模型优化
improve neural networks
transformer on load tap changer
fault diagnosis
diagnostic model
model optimization
