基于改进的BP神经网络学习算法的变压器故障诊断

Transformer Fault Diagnosis Based on Improved BP Neural Network Algorithm

电力变压器的故障除了给其自身带来重大损失外,还对电力系统的安全造成很大影响。利用BP神经网络对变压器故障进行诊断,针对BP神经网络学习率的缺点,提出了一种跟踪型自适应学习率的确定方法,该方法仅需整定一个参数,有效地提高了BP神经网络的收敛性和训练时间,进而通过构建变压器故障诊断训练样本集,验证了该方法的可行性,获得了更精确的诊断结果。

Failure of power transformers will bring significant losses to itself and cause great impact on the power system security.This paper improves fault diagnosis method of transformer based on BP neural network.Aiming at the disadvantage of learning rate of BP neural network,a self-adaptive tracking method is proposed to gain learning rate with only determining one parameter,which improves convergence and training time of BP neural network.Then the training sample set of transformer fault diagnosis is established to verify the feasibility of the proposed method.Finally,the accurate diagnosis results is obtained.