Abstract: Traditional power grid monitoring systems in mines face problems such as one-sided fault information perception, insufficient positioning accuracy, and slow real-time response. Therefore, an intelligent monitoring and fault location system for mine power grids that integrates multi-source heterogeneous information and improved deep learning technology is proposed. First, the topology of the mine power grid and its operational characteristics under harsh conditions must be analyzed to determine the system’s detection requirements and create a three-layer collaborative monitoring architecture of “intelligent sensing - edge computing - cloud platform”. This enables comprehensive perception of electrical and non-electrical parameters, rapid processing, and intelligent decision-making. Next, an improved 1D-CNN-GAM-GRU hybrid feature extraction model is planned, using a Global Attention Mechanism （GAM） to enhance critical features in the channel dimension and local fault patterns in the spatial dimension, while a Gated Recurrent Unit （GRU） precisely captures the temporal correlations of transient fault signals. A multi-source information fusion positioning algorithm based on rough set theory for selection and D-S evidence theory for fusion is proposed. This algorithm combines large amounts of information, including electrical quantities, traveling wave signals, equipment states, and environmental parameters, while eliminating redundant interference evidence. A 6 kV mine power grid simulation model is created in MATLAB/Simulink, and various typical faults are introduced for validation. Experimental results show that the system reduces the average positioning error to below 0.35%. Even under high-resistance grounding, branch line faults, or strong noise environments, the positioning accuracy remains at 99.4%. Its stability and adaptability are far superior to traditional traveling wave methods, impedance methods, and pure deep learning methods, providing strong technical support for the safe and stable operation of mine power grids.