Prediction-assessment-optimization of water influx in underground coal gasification: A systematic method

Underground coal gasification (UCG) is a crucial innovation for clean coal development and large-scale hydrogen production. Water influx affects gasification and gas quality, but managing it remains challenging due to the complex interactions of various processes. This paper presents a UCG water influx management framework that integrates water influx prediction, gasification safety assessment, and optimal water injection design. A prediction model combining convolutional neural networks (CNN), bidirectional long short-term memory networks (BiLSTM), the attention mechanism, and support vector machines (SVM) is proposed for water influx prediction. New safety assessment criteria for gasification site selection and an optimization model for water injection volume are established, considering the dynamic water influx effect. The results show that the CNN-BiLSTM-Attention-SVM model achieves the highest prediction accuracy compared with CNN-BiLSTM-Attention, CNN-BiLSTM-SVM, CNN-SVM, and SVM models. Taking the X experimental area in Xinjiang's Santanghu Basin, China, as an example, the dynamic water influx during gasification is below the upper limit, indicating its suitability for gasification. When the gasification water demand is 0.8 m³ per ton of coal, the optimal strategy is no injection in the initial stage and 121.68 m³/d in the steady stage. Hydrodynamics indicates that a gradual and slow pressure-drop scheme could benefit the hydrogen production in the initial stage of gasification. This study provides theoretical support for the selection of UCG sites and the optimization of injection schemes. • A water influx management framework that integrates prediction, evaluation, and optimization is developed. • The proposed CNN-BILSTM-Attention-SVM model achieves long-term water influx predictions. • Safety assessment criteria for gasification site selection are established based on static and dynamic parameters. • An optimization model for water injection volume is developed based on the water injection-production balance equation.