Identification and prediction of mixed water sources in adjacent limestone aquifers based on conventional hydrochemistry and strontium isotopes

Water gushing in mines is one of the most threatening geological disasters in the process of coal mine production, so the key to preventing water-gushing disasters in mines is to identify the water-gushing sources (WGS) quickly and effectively. Given the problem that it is difficult to effectively identify the WGS from adjacent limestone aquifers by conventional technical approaches, a typical coal mine in Huainan coalfield, the Group A coal seams in the Panji-2 coal mine, was selected as the object of investigation in this study. A total of 60 water samples were collected from the Taiyuan Formation limestone aquifer and the Ordovician limestone aquifer five times by using underground hydrological long-observation holes. The conventional ion hydrogeochemical analysis of Taiyuan Formation limestone water (TLW) and Ordovician limestone water (OLW) samples, in conjunction with the Sr concentration of rocks and limestone water with their isotope test results, indicated that there is a difference between Taiyuan limestone water and Ordovician limestone water through comprehensive analyses. On the basis of five algorithms, principal component analysis (PCA), Fisher identification analysis model (Sr-F), Distance discriminant analysis model (Sr-D), BP neural network analysis model (Sr-B) and Grey relational analysis model (Sr-G), four identification models are constructed for the training dataset by combining the Sr isotope values. The prediction results showed that Sr-B had the highest prediction accuracy of 95% compared to the other three models. A comparative study was carried out on the prediction dataset to further explore the accuracy and stability of the four models. The results revealed that the Sr-B model has the highest prediction accuracy and good stability, and can be used to identify the TLW and OLW. However, the Sr-D model shows better stability but lacks accuracy and needs to be used with caution in practical applications.