论文部分内容阅读
变压器振动基频幅值大小是分析和评判变压器运行状态和故障诊断的重要依据。为此,提出一种基于广义回归神经网络(GRNN)的变压器振动基频幅值计算方法,用于计算正常运行状态下的变压器表面振动基频幅值。所提方法考虑变压器振动产生机理和影响因素,先根据变压器运行电压、负载电流、油温等历史运行工况数据以及表面振动历史数据进行GRNN训练,保存训练好的GRNN网络即可根据实时运行工况数据计算变压器表面振动基频幅值。某台在运变压器表面振动实测信号的计算结果表明:所提方法比现有方法计算误差大幅下降约50%。研究结果可为变压器振动在线监测提供重要参考。
Transformer vibration frequency amplitude is an important basis for analysis and evaluation of transformer operation and fault diagnosis. For this reason, a method of calculating fundamental frequency of transformer vibration based on generalized regression neural network (GRNN) is proposed to calculate the fundamental frequency amplitude of transformer surface vibration during normal operation. The proposed method considers the mechanism and influencing factors of transformer vibration. Firstly, it conducts GRNN training according to historical operating conditions such as transformer operating voltage, load current and oil temperature as well as surface vibration history data. The trained GRNN network can be saved according to real-time operation Condition Data Calculate the fundamental frequency amplitude of the transformer surface. The calculation results of the measured signal of a certain transformer on the surface of transformers show that the proposed method can reduce the error greatly by about 50% compared with the existing method. The research results provide an important reference for on-line monitoring of transformer vibration.