Analysis of Potential Water Inflow Rates at an Underground Coal Mine Using a WOA-CNN-SVM Approach

Abstract

The water yield of aquifers increases the risk of water inflow, threatens the safe production of coal mines, and even causes geological disasters and construction hazards. To predict water yield quickly and accurately, multiple composite factors are used to invert unit water inflow rates to judge water yield grade. Taking the typical representative of north China-type coal fields as an example, six factors are selected: aquifer thickness, the radius of influence, normalized drawdown, permeability coefficient, the core rate of drilling holes, and the proportion of clay thickness to the thickness of the lower group. The whale optimization algorithm (WOA)–convolutional neural network (CNN)–support vector machine (SVM) model is established with the unit water inflow rate as the forecast target, and different models are selected for comparison. The water yield zoning map is obtained by bringing the borehole data into the model for prediction. The findings indicate that the root mean square error and average absolute error of the composite predictive model models are 0.0318 and 0.0268, respectively, and the model outperforms alternative models. The predicted water yield zoning aligns well with the actual conditions, offering a novel paradigm for water yield assessment.