一种改进卷积神经网络的逆变器故障诊断

针对传统二极管钳位式三电平逆变器故障诊断方法存在的诊断效率低且准确率不高的问题,将一种自适应正则化系数引入卷积神经网络CNN(convolutional neural network),对逆变器进行故障诊断。在传统CNN模型引入正则化去拟合中,正则化系数常采用全局统一的常数型参数,训练过程中需不断试错且效果甚微,针对此提出根据目标损失函数梯度变化,自适应调整正则化系数的CNN模型,能够加快其在逆变器故障诊断中的收敛速度,增强模型泛化能力,提高故障识别准确率。实验表明,与传统BP神经网络和原始CNN模型相比,改进的CNN模型能对逆变器复杂故障做出实时准确诊断。

Improved Inverter Fault Diagnosis Based on Convolutional Neural Network

In view of the low diagnostic efficiency and accuracy of the traditional diode clamping three-level inverter fault diagnosis method, an adaptive regularization coefficient is introduced into the convolutional neural network (CNN) for fault diagnosis of the inverter. In the conventional CNN model introduced regularization and de-fitting, the regularization coefficient often using global unified constant regularization coefficient model parameters, in the training process, it is necessary to continuously try and little effect. So, this paper proposes adaptive adjustment regularization according to the target loss function gradient change to speed up the convergence of the CNN model in inverter fault diagnosis, enhance the generalization ability of the model, and improve the accuracy of fault identification. Experiments show that the improved convolutional neural network model can make real-time and accurate diagnosis of the complex fault of the inverter compared with the traditional BP neural network and the original CNN model.