Water gushing in mines is one of the most threatening geological disasters in the process of coal mine production, so the key to prevent water gushing disaster in mines is to quickly and effectively identify the water-gushing sources (WGS). In view of the problem that it is difficult to effectively identify the WGS from adjacent limestone aquifer by conventional technical approaches, the Group A coal seams in Panji-2 coal mine, a typical coal mine in Huainan coalfield, was selected as the object of investigation in this study. A total of 60 water samples from two adjacent aquifers (Taiyuan formation limestone aquifer and Ordovician dolomitic-limestone aquifer) were systematically collected by using underground hydrological long-distance observation hole. On the basis of analyzing the hydro-chemical properties of the water samples, the hydrochemistry types of each aquifer were obtained. The results indicate that there is a certain hydraulic connection between Taiyuan formation limestone water (TLW) and Ordovician limestone water (OLW). In order to characterize and trace the similarities and differences between the two types of water sources, the concentration of strontium (γSr2+) and its isotope value (87Sr/86Sr, δ87Sr) were first determined, and it was found that γSr2+ had a significant discrimination between the two types of water sources, and accounted for the major contribution rate in the results of principal component analysis (PCA). Therefore, in conjunction to PCA, four identification models of WGS based on strontium isotope were established successively (Fisher discrimination analysis model (δ87Sr-F model), Distance discrimination analysis model (δ87Sr-D model), BP neural network analysis model (δ87Sr-B model) and Grey relational analysis model (δ87Sr-G model)). First, the stability and reliability of four models were trained according to the data of 40 water samples, and then the trained model was used to identify the source of the remaining 20 water samples. From the results, the δ87Sr-B model has the best discrimination effect, and its accuracy rate reaches 95%. In the actual production process of Panji-2 coal mine in the future, it is suggested to adopt the δ87Sr-B model to carry out the theoretical model of WGS identification of Group A coal seams floor, which provides a guarantee for coal mine safety production.