基于优化深度神经网络的配电网单相接地故障诊断

Diagnosis for Single-phase Grounding Fault in Distribution Networks Based on the Optimized Deep Neural Network

随着我国配电网的不断扩大,当配电网发生单相接地故障时,迅速找出并切除故障线路是供电可靠性的保证之一。提出了一种暂态故障特征和稳态故障特征相结合,并采用交叉熵损失函数和改进学习率优化的深度神经网络对故障进行辨别的方法。结果表明,方法有效地减少了深度神经网络的迭代次数,提高了学习效率。在辨别单相接地故障时,采用交叉熵损失函数和改进学习率优化的深度神经网络方法比未优化的深度神经网络准确率高,抗干扰性好。

With continuous expansion of China's distribution network,in case of a single-phase grounding fault in the distribution network,rapid detection and removal of the fault line is one of the guarantees for power supply reliability.In this paper,a fault identification method based on the combination of transient and steady-state fault characteristics was introduced,and fault identification was implemented through the deep neural network adopting cross-entropy loss function and optimized improved learning rate.Experimental results indicated that this method greatly reduced the number of iterations of the deep neural network and improved learning efficiency.For identification of single-phase grounding faults,the presented approach of deep neural network adopting cross-entropy loss function and optimized improved learning rate achieved a higher accuracy and better anti-interference performance than the unoptimized deep neural network.