In high-voltage (HV) transmission networks, the fault current of high-resistance ground faults is weak and difficult to detect, which poses a huge challenge to existing research. To address this challenge, this paper proposes a deep learning-based high-resistance ground fault detection method to enhance the performance of single-phase ground protection. We select three advanced sequence modeling architectures, namely long short-term memory (LSTM) recurrent network, one-dimensional convolutional neural network (CNN), and Transformer, and apply them to time series voltage and current data to achieve accurate detection of high-resistance ground faults. To verify the effectiveness of the proposed method, these models are trained and evaluated on an open source high-resistance fault dataset. Experimental results show that the proposed system performs well in distinguishing high-resistance ground faults from normal conditions and other transient events, with high detection accuracy.