基于数字孪生的优化概率神经网络变压器故障诊断

Fault Diagnosis of Transformers Based on Optimal Probabilistic Neural Network Based on Digital Twin

针对油浸式电力变压器无法实时监测其运行情况、故障诊断精度低、速度慢等问题,提出一种基于数字孪生技术的优化概率神经网络的故障诊断方法。首先,根据变压器结构、运行等特点建立基于数字孪生的故障诊断模型,采用差分进化算法优化概率神经网络(PNN)中的平滑因子,再将优化后的平滑因子赋给PNN,最终得到优化后的故障诊断模型,进而构建高精度变压器数字孪生体进行实时故障诊断分析。优化结果表明,与优化前以及RBF和BP网络相比,变压器故障诊断的精度明显提高且收敛速度快,基于数字孪生技术能够实现实时诊断变压器故障。

An optimized probabilistic neural network(PNN)fault diagnosis method based on digital twin technology is proposed to solve the problems of oil-immersed power transformers such as the inability to monitor their operation in real time,low accuracy and slow speed.First of all,according to the structure and operation characteristics of transformer fault diagnosis model based on the number of twin,using differential evolution algorithm to optimize the PNN smooth factors,then the optimal smoothing factor is assigned to PNN,finally obtained the optimized fault diagnosis model,and then build a high precision digital transformer analyzes real-time fault diagnosis of the twin.The optimization results show that compared with RBF and BP networks before optimization,the accuracy of transformer fault diagnosis is obviously improved and the convergence speed is fast.