Aiming at the problems that single-phase grounding fault line selection of mine power grid is greatly interfered by underground environment, and the fault line selection speed and accuracy are low, A new method for single-phase grounding fault line selection of mine power grid based on arithmetic optimization algorithm and improved continuous variational mode decomposition and attention mechanism convolutional neural network is proposed. Firstly, the parameters of continuous variational mode decomposition are optimized by arithmetic optimization algorithm, and the zero-sequence current sequence is divided into natural mode functions of different frequencies. Secondly, the data preprocessing method of relative position matrix is introduced to convert one-dimensional sequence into two-dimensional image, and the time-frequency characteristic diagram of zero-sequence current signal is obtained. Finally, the attention mechanism is embedded into the CNN classification algorithm model to realize fault line selection.. The simulation and experimental results show that the proposed method can accurately select fault lines under the conditions of strong noise and asynchronism of sampling time, and can meet the requirements of accuracy and reliability of line selection in mine power grid.