1 China Energy Shendong Coal Group Co., Ltd., Technology Research Institute, Yulin 719315, China
2 State Key Laboratory of Water Resource Protection and Utilization in Coal Mining, National Institute of Clean and Low Carbon Energy, Beijing 102209, China
3 Shangwan Coal Mine, National Energy Group Shendong Coal Group Corporation, Ordos 017209, China
Accurate prediction of mine water inflow is a key technology to ensure safe production and ecological protection of mines. In view of the fact that traditional prediction methods are difficult to handle complex spatiotemporal characteristics, and the existing data-driven models have problems such as single structure and insufficient parameter optimization, a temporal convolutional network-bidirectional long short-term memory network (TCN-BiLSTM) joint prediction model was proposed based on simulated annealing algorithm optimization. This model combines the multi-scale feature extraction capability of the temporal convolutional network (TCN) and the context-dependent modeling advantages of the bidirectional long short-term memory network (BiLSTM), and combines the simulated annealing algorithm (SA) for global optimization of hyperparameters to construct an efficient SA-TCN-BiLSTM prediction framework. Based on the mine water inflow data and hydrological and meteorological observation data of Shenhua Shangwan Mining Area from 2014 to 2023, the model is compared with benchmark models such as long short-term memory (LSTM), BiLSTM, and TCN. The results show that the water inflow in the mining area shows significant seasonal changes from 2014 to 2023, and precipitation, temperature and potential evaporation have non-balanced contributions to the prediction of water inflow. The proposed SA-TCN-BiLSTM model is significantly better than the traditional method in prediction accuracy, with an mean square error(MSE) of 2.10 and a mean absolute percentage error(MAPE) of 0.34%. Compared with the unoptimized TCN-BiLSTM model, the prediction accuracy is improved by 97.5%. The model predicts that the water inflow in the next 12 months will fluctuate in the range of 250~430 m3/h and reach the annual peak during July and September. This prediction trend is highly consistent with historical observation data and regional hydrological characteristics. The research results provide reliable technical support and scientific basis for the prediction of water inflow in mining areas and the formulation of water prevention and control plans.
HE An-min, HU Yu-tian, ZHANG Hai-qin, JIANG Bin-bin, HAO Jing-wei, HE Rui-min, LI Xue-jia, LI Yu.
Application of TCN-BiLSTM Model Based on Simulated Annealing Optimization in Water Inflow Prediction in Shendong Mining Area[J].
Science Technology and Engineering, 2026, 26(5): 1927-1937 DOI:10.12404/j.issn.1671-1815.2501885
DuZ, WuQ, ZhaoY, et al. A multi-constraint and multi-objective optimization layout method for a mine water inrush monitoring network[J]. Scientific Reports, 2023, 13(1): 11817.
ZengYifan, WuQiang, ZhaoSuqi, et al. Characteristics, causes, and prevention measures of coal mine water hazard accidents in China[J]. Coal Science and Technology, 2023, 51(7): 1-14.
ZhouYanan, CuiYibin. Uncertainty analysis and risk response research of coal mine dynamic forecast water inflow[J]. China Energy and Environmental Protection, 2024, 46(11): 281-286.
[6]
HuW, ZhaoC. Evolution of water hazard control technology in China’s coal mines[J]. Mine Water and the Environment, 2021, 40(2): 334-344.
WuJian, ZhouZhifang, LiMingwei, et al. Advance on the methods for predicting water inflow into tunnels[J]. Journal of Engineering Geology, 2019, 27(4): 890-902.
SunGangyou, HuQingzhen, KangQinrong, et al. Comparative analysis of water inflow prediction of a mine based on large well method and GMS numerical simulation method[J]. Science Technology and Engineering, 2023, 23(21): 9024-9031.
LuRui, YinShangxian, WangYuguo, et al. Prediction of water inflow in deep coal seam mining face based on GMS[J]. Safety in Coal Mines, 2025, 56(1): 164-170.
[16]
BianJ, HouT, RenD, et al. Predicting mine water inflow volumes using a decomposition-optimization algorithm-machine learning approach[J]. Scientific Reports, 2024, 14(1): 17777.
[17]
SunJ, HuL, LiD, et al. Data-driven models for accurate groundwater level prediction and their practical significance in groundwater management[J]. Journal of Hydrology, 2022, 608: 127630.
[18]
SalmasiF, NouriM, SihagP, et al. Application of SVM, ANN, GRNN, RF, GP and RT models for predicting discharge coefficients of oblique sluice gates using experimental data[J]. Water Supply, 2020, 21(1): 232-248.
HuangCunhan, FengTao, WangWeijun, et al. Mine water inrush prediction based on fractal and support vector machines[J]. Journal of China Coal Society, 2010, 35(5): 806-810.
YaoDongze. Research on mine water inflow prediction model based on deep learning Self-Attention[D]. Yangling: Northwest A & F University, 2024.
[23]
GongX, LiB, YangY, et al. Construction and application of optimized model for mine water inflow prediction based on neural network and ARIMA model[J]. Scientific Reports, 2025, 15(1): 2009.
[24]
YaoD, ChenS, DongS, et al. Modeling abrupt changes in mine water inflow trends: a CEEMDAN-based multi-model prediction approach[J]. Journal of Cleaner Production, 2024, 439: 140809.
HouEnke, XiaBingbing, WuZhangtao, et al. Mine water inflow prediction based on CEEMDAN-BO-BiGRU[J]. Science Technology and Engineering, 2023, 23(28): 12012-12019.
[27]
ShiJ, WangS, QuP, et al. Time series prediction model using LSTM-Transformer neural network for mine water inflow[J]. Scientific Reports, 2024, 14(1): 18284.
[28]
YuY, SiX, HuC, et al. A review of recurrent neural networks: LSTM cells and network architectures[J]. Neural Computation, 2019, 31(7): 1235-1270.
[29]
MaoY, ZhangY, JiaoL, et al. Document-level sentiment analysis using attention-based bi-directional long short-term memory network and two-dimensional convolutional neural network[J]. Electronics, 2022, 11(12): 1906.
[30]
FanJ, ZhangK, HuangY, et al. Parallel spatio-temporal attention-based TCN for multivariate time series prediction[J]. Neural Computing and Applications, 2023, 35(18): 13109-13118.
DingYingying, YinShangxian, LianHuiqing, et al. Prediction of mine water inflow along mining faces using the SSA-CG-Attention multifactor model[J]. Coal Geology & Exploration, 2024, 52(4): 111-119.
[33]
JiangK, PanZ, PanF, et al. Influence patterns of soil moisture change on surface-air temperature difference under different climatic background[J]. Science of The Total Environment, 2022, 822: 153607.
[34]
FriedmanJ H. Greedy function approximation: a gradient boosting machine[J]. The Annals of Statistics, 2001, 29(5): 1189-1232.
[35]
JukiĆD, DeniĆ-JukiĆV. Groundwater balance estimation in karst by using a conceptual rainfall-runoff model[J]. Journal of Hydrology, 2009, 373(3): 302-315.
[36]
WolkersdorferC. Water management at abandoned flooded underground mines: fundamentals, tracer tests, modelling, water treatment[M]. Springer: Springer Science & Business Media, 2008.
[37]
Mazor E. Chemical and isotopic groundwater hydrology[M]. New York: CRC Press, 2003.
[38]
XiaQ, HeJ, LiB, et al. Hydrochemical evolution characteristics and genesis of groundwater under long-term infiltration (2007—2018) of reclaimed water in Chaobai River, Beijing[J]. Water Research, 2022, 226: 119222.
[39]
ZhangZ, WangW, GongC, et al. Salix psammophila afforestations can cause a decline of the water table, prevent groundwater recharge and reduce effective infiltration[J]. Science of the Total Environment, 2021, 780: 146336.
[40]
AtawnehD A, CartwrightN, BertoneE. Climate change and its impact on the projected values of groundwater recharge: a review[J]. Journal of Hydrology, 2021, 601: 126602.
[41]
SangiorgioM, DercoleF. Robustness of LSTM neural networks for multi-step forecasting of chaotic time series[J]. Chaos, Solitons & Fractals, 2020, 139: 110045.
[42]
WangH, ZhangY, LiangJ, et al. DAFA-BiLSTM: deep autoregression feature augmented bidirectional LSTM network for time series prediction[J]. Neural Networks, 2023, 157: 240-256.
[43]
WeiH, ZhangQ, GuY. Remaining useful life prediction of bearings based on self-attention mechanism, multi-scale dilated causal convolution, and temporal convolution network[J]. Measurement Science and Technology, 2023, 34(4): 045107.
[44]
LiuY, HeidariA A, CaiZ, et al. Simulated annealing-based dynamic step shuffled frog leaping algorithm: optimal performance design and feature selection[J]. Neurocomputing, 2022, 503: 325-362.