基于栈式自编码器和Softmax分类器的电力变压器故障诊断

Fault diagnosis for power transformer using stacked auto-encoders and Softmax regression

为更加有效地解决电力变压器故障诊断时面临的数据提取、局部最优、梯度消散等问题,提出了一种基于栈式自编码器(stacked auto-encoders,SAE)与Softmax分类器的电力变压器故障诊断新方法。所提方法首先基于SAE与Softmax分类器理论,建立电力变压器故障诊断模型;然后基于k步对比散度算法,利用大量无标签样本对故障诊断模型中的每个受限玻尔兹曼机(restricted Boltzmann machine,RBM)进行逐层无监督训练,并使用有监督算法对模型参数进行调优;最后结合Softmax分类器对故障类型进行判断。算例分析证明,与基于支持向量机(support vector machine,SVM)和反向传播神经网络算法的故障诊断方法相比,所提方法在电力变压器评估方面具有较好的稳定性及更高的准确率。

To solve the availability of data extraction,better local optimum,gradient to dissipate more efficiently,this paper presents a new method of power transformer fault diagnosis based on stacked auto-encoders(SAE)and Softmax classifier.The model of power transformer fault diagnosis is established based on SAE and Softmax regression,then each restricted Boltzmann machine(RBM)of fault diagnosis model is optimized by the unsupervised raining with massive unlabeled samples based on k-step contrastive divergence and adjusts parameters of fault diagnosis model by the supervised algorithm.Finally,the power transformer fault types are determined by Softmax regression.Test results show that the proposed method has higher diagnosis accuracy and better adaptability than those based on the back propagation neural network and the support vector machine(SVM).