A topology identification method based on one-dimensional convolutional neural network for distribution network
Jielong Ni, Zao Tang, Jia Liu, Pingliang Zeng, Chimeddorj Baldorj
Abstract
Distribution network (DN) topology identification (TI) is the basis of distribution network state estimation. However, the connection of high-penetration renewable energy makes TI of DN more challenging. Thus, a TI method of active distribution network (ADN) based on one-dimensional convolutional neural network is proposed in this manuscript. Based on the sensitivity of node voltage to topology changing of DN, the characteristics of nodes are analyzed to select the key nodes to place the distribution network phasor measurement unit (DPMU), which can save investment and reduce the redundancy of model training. Several tests are carried out with the modified IEEE-33 bus DN with photovoltaic (PV) units. The results show that the proposed distribution network topology identification method can realize the high accuracy TI in ADN under limited DPMU measurement.