Abstract: For single-phase ground fault routing in small-current grounding systems, traditional methods commonly use a routing model based on one-dimensional signals, which suffers from low routing accuracy and weak noise immunity. This paper proposes a line selection method of improved variational mode decomposition and ConvNeXt single-phase ground faults for small current grounding systems. First, the ant-lion algorithm is introduced to optimize the variational mode decomposition algorithm. The ant-lion algorithm is used to automatically select the appropriate number of decompositions and penalty factors, calculate the distribution entropy of each component obtained by the decomposition, filter out the noise components, and remove the remaining effective ones. The components are linearly reconstructed to obtain the denoised zero-sequence current signals. Moreover, the one-dimensional zero-sequence current signal after denoising is converted into a two-dimensional image signal through the Gram angle field, and a fault line selection data set is prepared. Then, the pre-trained ConvNeXt model is introduced. According to the data model characteristics of this article, the model parameters are fine-tuned based on its existing weights, thereby improving the accuracy of the model and forming the final line selection model. Finally, the absolute mean error and root mean square error are introduced as evaluation indicators to verify the effectiveness of the proposed noise reduction algorithm. On the premise of adding noise or not, the proposed model is compared with the three line selection models. The experimental results show that the proposed model has the highest accuracy and better noise immunity performance. Besides, the algorithm applied can reach 99.82% accuracy and maintain an accuracy of more than 91% under different noise conditions. This performance is higher than that of other line selection models and overcomes the problems of low accuracy and poor noise immunity of traditional fault line selection methods.