Abstract: In response to issues such as incomplete fault information perception, low positioning accuracy, and poor real-time performance in traditional coal mine power grid monitoring systems, this study proposes an intelligent monitoring and fault location method for coal mine power grids based on multi-source information fusion and improved deep learning. First, an intelligent monitoring architecture integrating smart sensing, edge computing, and cloud platform layers is constructed, enabling panoramic perception and collaborative transmission of both electrical and non-electrical parameters. Second, a fault feature extraction model based on an improved parallel structure of one-dimensional convolutional neural network （1D-CNN） and long short-term memory （LSTM） network is proposed to effectively capture spatiotemporal features in transient fault signals. Furthermore, a multi-source information fusion fault location algorithm based on Bayesian inference and D-S evidence theory is designed, integrating electrical quantities, traveling wave signals, and switch status information, significantly improving positioning reliability and accuracy in complex distribution network structures. Finally, a simulation model is built using MATLAB/Simulink, and various types of faults are injected for validation. Experimental results demonstrate that compared to traditional traveling wave ranging and impedance methods, the proposed approach reduces the average positioning error to below 0.5%, while maintaining high positioning accuracy and robustness in high-noise environments, providing effective technical support for the safe and reliable operation of coal mine power grids.