基于改进卷积神经网络的电力通信网故障诊断研究

针对当前电力通讯网络故障诊断方法及时性差、准确率低和自我学习能力差等缺陷,提出基于改进卷积神经网络的电力通信网故障诊断方法,结合ReLU和Softplus两个激活函数的特点,对卷积神经网络原有激活函数进行改进,使其同时具备光滑性与稀疏性;采用ReLU函数作为作为卷积层与池化层的激活函数,改进激活函数作为全连接层激活函数的结构模型,基于小波神经网络模型对告警信息进行加权操作,得到不同告警类型和信息影响故障诊断和判定的权重,进一步提升故障诊断的准确率;最后通过仿真试验可以看出,改进卷积神经网络相较贝叶斯分类算法与卷积神经网络具有较高的准确率和稳定性,故障诊断准确率达到99.1%,准确率标准差0.915%,为今后电力通讯网智能化故障诊断研究提供一定的参考。

Research on Fault Diagnosis of Electric Power Communication Network Based on Improved Convolutional Neural Network

Aiming at the shortcomings of current power communication network fault diagnosis methods,which is poor timeliness,low accuracy,and poor self-learning ability and so on,based on improved convolutional neural networks,a power communication network fault diagnosis method is proposed,combines the characteristics of two activation functions of ReLU and Softplus.The original activation function of the product neural network is improved to simultaneously make it both smooth and sparse.The ReLU function is used for the activation function of the convolutional layer and the pooling layer,and the activation function is improved as the structural model of the activation function of the fully connected layer.The wavelet neural network model performs a weighted operation on the alarm information,and obtains the weights of different alarm types and information that affect the fault diagnosis and judgment,and further improves the accuracy of the fault diagnosis.Finally,through the simulation test,it can be presented that the improved convolutional neural network is compared with the Bayesian.The classification algorithm and convolutional neural network have high accuracy and stability.The accuracy of fault diagnosis reaches by 99.1,and the standard deviation of accuracy is 0.915%.It provides a certain reference for future research on intelligent fault diagnosis of electric power communication network.