Discrimination of water inrush source in coal mine based on PCA-GA-ET

Abstract

In recent years, mine water inrush has hindered coal mining, and accurate identification of the source of water inrush is essential for the prevention and control of mine water disasters. To study the source of water inrush in Pingdingshan coalfield, water source samples of 124 groups of main water inrush aquifers were collected. The main water-filled aquifers in Pingdingshan coalfield include surface water (I), Quaternary pore water (II), Carboniferous limestone karst water (III), Permian sandstone water (IV) and Cambrian limestone karst water (IV).In this paper, the conventional hydrochemical discriminant ions Na++K+, Ca2+, Mg2+, Cl-, SO42-, and HCO3- of observed water samples are extracted to build a water source discriminant model. According to the Pearson correlation coefficient, Na++K+has a strong correlation with HCO3-. To avoid the impact of information overlap on the accuracy of the model, principal component analysis (PCA) is used to extract the main indicators. Based on the adaptive differential evolution genetic algorithm (GA) to optimize the super parameters of the extreme tree (ET), the PCA-GA-ET water source identification model is constructed. To further verify the reliability of the model, PCA-GA-ET is compared with grid search-random forest (GS-RF), artificial neural network (MLP), and particle swarm optimization-support vector machine (PSO-SVM). Finally, using the trained model to discriminate 25 groups of validation samples, it can be concluded that the discrimination results of 24 groups of samples are consistent with the observation results, and the accuracy of PCA-GA-ET is 96%. Research shows that the use of principal component analysis can reduce the information overlap of data, and the extreme tree (ET) model optimized by genetic algorithm (GA) greatly improves the efficiency and accuracy of water inrush source identification.