Work

Application of the CPO-CNN-BILSTM Hybrid Model for Evaluation of Water Abundance of the Roof Aquifer—A Case Study of WoBei Mine in Huaibei Coalfield, China

Year: 2025

Type: article

Abstract: With the gradual increase in coal production capacity, the problem of water damage from the coal seam roof is becoming more and more prominent. Neogene loose strata overlie coal seams in eastern China, and pressurized aquifers commonly lie at the bottom of the loose strata. The aquifers are mainly composed of unconsolidated sand, gravel, and weakly consolidated marl, which has strong permeability and an extremely unfavorable impact on safe production. Identifying the target area to prevent and control roof water damage can reduce the likelihood of water damage accidents in mines. This study takes the 85 mining district of Wobei mine as an engineering case. The discriminant indexes are selected for aquifer thickness, gradation coefficient, marlstone thickness, permeability, grouting quantity, and grouting termination pressure. A model integrating the newly proposed Crowned Porcupine Optimization (CPO, 2024), Convolutional Neural Network (CNN), and Bidirectional Long Short-Term Memory (BiLSTM) was constructed to predict unit water influx. A zonal map was generated based on the expected unit water influx of the fourth aquifer after grouting. In addition, the prediction results are compared with those from other models. Results indicate that the CPO-CNN-BiLSTM model achieves a higher accuracy and fewer errors in water abundance prediction, with an RMSE of 2.58 × 10−5 and an R2 of 0.982 for the testing dataset. According to the prediction result, the fourth aquifer after grouting in the 85 mining district is divided into five water abundance zones. The strong and medium–strong water abundance zones are mainly distributed in the study area’s eastern region. A small portion of them is distributed in the northwestern and northern areas. This study provides a new insight for predicting the water abundance of thick loose aquifers and a theoretical basis for safe mining under thick loose aquifers.

Source: Applied Sciences

DOI: 10.3390/app152111816

Authors: Yuchu Liu, Qiqing Wang, Jingzhong Zhu, Dongding Li, Wenping Li

Institutions: China University of Mining and Technology, Ministry of Emergency Management of the People's Republic of China

Language: English

Topic: Geoscience and Mining Technology

Subfield: Safety, Risk, Reliability and Quality

Field: Engineering

Domain: Physical Sciences

SDG: Clean water and sanitation

Open Access status: gold

Estimated APC paid: $2,896