基于贝叶斯正则化深度信念网络的电力变压器故障诊断方法

传统的深度信念网络规模大、难度大、训练时间长,导致其故障诊断的时间较长。针对该问题,提出了一种基于贝叶斯正则化深度信念网络的电力变压器故障诊断方法。采用贝叶斯正则化算法改进传统深度信念网络的训练性能函数,在保证网络精度的同时快速提高计算速度,从而提高网络的收敛速度。实验结果表明,经过贝叶斯正则化改进后,深度信念网络训练的泛化能力得到了提高,同时故障诊断的准确率也得到了保证。

Fault diagnosis of power transformer based on BR-DBN

The traditional DBN(Deep Belief Network) has large scale, great difficulty and long training time, which leads to the long time of fault diagnosis. Aiming at this problem, a fault diagnosis method for power transformer based on BR-DBN(Bayesian Regularization-Deep Belief Network) is proposed. The Bayesian regularization is used to improve the function training performance of traditional DBN, which not only maintains the network accuracy but also quickly improves the calculation speed, so as to improve the convergence speed of the network. The experimental results show that, the generalization ability of the whole DBN enhanced by Bayesian regularization is improved, meanwhile its fault diagnosis accuracy is guaranteed.