基于边缘计算和深度学习的有限信息配电网单相接地故障区段定位

目前围绕量测条件受限的配电网展开的故障定位研究较少，且传统的主站集中式故障定位系统在实时性与安全性等方面存在不足。针对上述问题，提出一种基于边缘计算和深度学习的单相接地故障区段定位方法。首先，构建基于分区修正的边缘计算单元配置多目标优化模型。该模型通过分区修正方法降低了故障定位系统的通信时延，提升了数据传输安全性，进而保障配电网安全运行。其次，将基于数据驱动的智能算法应用于配电网故障区段定位，选择易获取的相电流稳态有效值在故障前后的变化量作为故障特征，利用全连接型深度神经网络学习样本特征与标签间的映射关系，得到离线训练好的定位模型并储存在边缘节点以实现快速故障定位。最后，以IEEE33节点系统为例进行仿真。算例结果表明该模型在分布式电源接入、高阻故障、噪声干扰以及拓扑改变等情况下均具有良好表现。

Single-phase ground fault section location in distribution networks with limited information based on edge computing and deep learning

At present, there are few studies on fault location in a distribution network with limited measurement conditions. In addition, the traditional centralized fault location system of a master station has shortcomings in real-time and security. Thus, a single-phase ground fault section location method based on edge computing and deep learning is proposed. First, a multi-objective optimization model of edge computing unit configuration based on partition correction is constructed. The model reduces the communication delay of a fault location system and improves the security of data transmission by the partition correction method, thus ensuring the safe operation of the network. Second, a data-driven intelligent algorithm is applied to the fault section location. The variation of the steady-state effective value of the phase current before and after the fault is selected as the fault feature. A fully connected deep neural network is used to learn the mapping relationship between sample features and labels, and an offline trained location model is obtained and stored at the edge nodes to achieve fast fault location. Finally, the IEEE 33-bus system is taken as an example for simulation. The example shows that the model performs well with distributed generation access, high resistance fault, noise interference and topology change.