本文针对自适应自动重合闸过程中如何判别故障性质的问题，提出了基于人工神经网络的方法，并设计了一个三层神经网络来建立自适应单相自动重合闸的方案。采用了改进的ＢＰ学习算法，并对特征输入量的提取作了探讨。结合ＥＭＴＰ仿真的训练和测试结果表明，所设计的神经网络是可以解决自适应单相重合闸问题的。 In this paper, aiming at the problem of how to judge the nature of the fault in the course of adaptive auto-reclosure, a method based on artificial neural network is proposed and a three-layer neural network is designed to establish an adaptive single-phase auto-reclosure scheme. The improved BP learning algorithm is adopted, and the extraction of feature input is discussed. The training and test results based on EMTP simulation show that the designed neural network can solve the problem of adaptive single-phase reclosing.