ORIGINAL RESEARCH article
Front. Earth Sci.
Sec. Geohazards and Georisks
Volume 13 - 2025 | doi: 10.3389/feart.2025.1620487
This article is part of the Research TopicNatural Disaster Prediction Based on Experimental and Numerical MethodsView all 24 articles
Aquifer water yield property prediction based on a hybrid neural network model: A case of Yili No.4 colliery, Xinjiang
Provisionally accepted
- China University of Mining and Technology, Xuzhou, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
With the gradual increase of coal production capacity, the mining-induced roof water damage has become increasingly prominent. Accurately and effectively predicting the water yield property of the roof aquifer based on the hydrogeological data is of great significance for preventing and controlling mine water damage. In this study, we select six evaluation factors, including aquifer thickness (AT), permeability coefficient (PC), coring rate (CR), rock brittleness-plasticity ratio (RBPR), equivalent thickness of sandstone (ETS), and fold undulation (FU). A hybrid model is proposed, integrating the convolutional neural networks (CNN) with long short-term memory (LSTM) optimized by the Attention module to improve the model's performance. The model is applied to predict the water yield property of the Paleocene aquifer in the Yili No. 4 colliery by collecting 100 hydrogeological datasets. The model is trained to predict the unit water inflow (UWI) of the roof aquifer, reflecting the water yield property. Besides, comparative analysis with the CNN, LSTM, and the CNN-LSTM models demonstrates that the prediction performance of the CNN-LSTM-Attention model outperforms the three contrastive models. The CNN-LSTM-Attention model has higher accuracy and lower error with the real value, and the RMSE of the testing dataset is 1.98e-3, and the prediction accuracy is 0.9725. Based on the CNN-LSTM-Attention model validated, the UWI is predicted, and then the water yield property zone (WYPZ)of the Paleocene aquifer is identified using the predicted UWI. Additionally, a comparison is conducted between the WYPZ results derived from predicted UWI and those based on actual UWI. The findings demonstrate that the CNN-LSTM-Attention hybrid model achieves higher prediction accuracy. This study proposes a scientifically robust evaluation method for delineating WYPZ in mining areas with limited hydrogeological exploration data.
Keywords: water yield property, Hydrogeological data, CNN-LSTM-Attention, prediction performance, Unit water inflow
Received: 29 Apr 2025; Accepted: 23 Jul 2025.
Copyright: © 2025 Jielin, Li, Zhu and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence:
Wenping Li, China University of Mining and Technology, Xuzhou, China
Jingzhong Zhu, China University of Mining and Technology, Xuzhou, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.