Abstract: Any mine water hazard will show different omens. In order to consolidate the foundation of intelligent early warning of water disaster, the connotation of system construction of perception, identification, evaluation, prediction and the interlogical relation are clarified. The mechanism of water inrush is varred for different types of water hazards, we designed three types of multi-mode water inrush typical scenarios and established the corresponding water inrush criteria. In addition, so we proposed two kinds of forecasting methods, namely the precise prediction of deterministic theory and the non-deterministic trend projections including big data and deep learning, which laid a solid theoretical foundation for the prediction, warning criteria and threshold setting of the intelligent early warning system. Taking Tingnan Coal Mine of Binchang mining area in Shaanxi Province as an example, we established an index system of dynamic information, static information and related information, which integrated the ground hydrological dynamic monitoring unit, underground hydrological environment monitoring unit and mining dynamic monitoring unit of mining face, and built an in-situ acquisition and water-inrush element prediction perception system. Based on the dynamic monitoring of electrical parameters of key layers and the joint arrangement of single point or multi-point and multi-parameter monitoring of key parts, the accurate acquisition scheme of water inrush precursor information was implemented. The deterministic simulation model and the non-deterministic intelligent model were adopted to realize the prediction and early warning function of water disasters. According to the multi-source data fusion and spatial linkage analysis technology, the early warning system realized the visual display of "one map" of the whole space water disaster early warning on surface and underground. The results prove that the monitoring and early warning platform has a solid theoretical foundation, and the prediction and early warning effect is remarkable.