Abstract: In view of the problems of complex hydrogeological conditions, insufficient accuracy of the traditional zoning method and others in mining areas, this study takes Zuoquan Xinshun Coal Industry as the research object and proposes a type of hydrogeological zoning method based on multi-source monitoring data. This method constructs a comprehensive data acquisition system that includes drilling, geophysical exploration, and hydrological monitoring, and develops a multi-source data fusion and zoning evaluation model based on D-S evidence theory, Kalman filtering, fuzzy comprehensive judgment, multi-task learning framework, and spatiotemporal attention network. In the practical application, this method successfully divides five hydrogeological types of the mining area, providing a scientific basis for the layout of coal mining faces and the management of water prevention and control. After 12 months of practical application, this method enables the accuracy rate of mine water inflow prediction to reach 92%, effectively guiding the safe production. This research results provide valuable practical references for hydrogeological works in coal mines under complex conditions.