基于对比歧化和深度置信网络的配电网故障类型识别

针对配电网数据分支多、设备类型多样、现有故障诊断方法精度低的问题,提出基于深度置信网络的配电网故障诊断方法。该方法建立了4种不同层数的深度置信网络,将配电网的实际监测数据分为训练和测试数据导入到深度置信模型,采用对比歧化算法优化初始参数选择和加速模型训练,测试模型对样本的识别精度,建立改进BP神经网络和Petri网对比故障识别精度。结果表明,深度置信网络可以通过实时分析配电网实时监测数据,准确辨识配电网故障类型,提高了配网故障诊断的准确率和速度。

Fault Identification of Distribution Network Based on Contrast Disproportionate Algorithm and Deep Belief Network

To solve the problems of multiple data branches, various equipment types and low accuracy of the existing fault diagnosis methods in the power distribution network, this paper proposes a fault diagnosis method for the distribution network based on deep belief networks. With this method, the deep confidence network is built for four kinds of layers of different numbers, and the actual monitoring data of the distribution network are divided into training and test data, which are imported into the deep confidence model. The contrast disproportionate algorithm is used to optimize the initial parameter selection and acceleration model training, and test the accuracy of the sample identification of the model. Simultaneously the improved BP neural network and Petri net are established to compare fault recognition accuracies. The results show that the deep belief network can analyze the distribution network real-time monitoring data, accurately calculate the distribution network fault type, and improve the accuracy and speed of distribution network fault diagnosis.