1 Jinneng Holding Equipment Manufacturing Group, Jincheng 048000, China
2 College of Artificial Intelligence and Computer Science, Xi’an University of Science and Technology, Xi’an 710054, China
3 Shanxi Jinmei Group Technology Research Institute Co., Ltd., Jincheng 048000, China
4 College of Safety Science and Engineering, Xi’an University of Science and Technology, Xi’an 710054, China
Show less
文章历史+
Received
Published
Revised
2025-09-11
2026-08-28
2026-01-12
Online Date
2026-09-04
PDF (7590K)
摘要
针对现有瓦斯浓度预测模型在多尺度时频特征提取与长程依赖建模方面的不足,提出一种基于TFSA-Net(time-frequency and stabilized attention network),TFSA-Net的工作面瓦斯浓度预测模型。该模型首先通过小波变换卷积提取多尺度时频特征,增强对不同时间尺度信息的捕获能力;进而利用稳定化注意力门控循环单元(stabilized attentive gate recurrent unit,SA-GRU)学习序列的时间依赖关系,并通过层归一化与多头注意力机制提升训练稳定性和长程依赖信息的利用效率;最后引入自适应频谱模块,在频域中对特征进行快速傅里叶变换(fast Fourier transform,FFT),并借助自适应高频掩码增强关键频率分量,抑制噪声干扰。实验结果表明:TFSA-Net在平均绝对误差(mean absolute error,MAE)和均方根误差(root mean square error,RMSE)指标上优于Nlinear、LDlinear、长短期记忆网络(long short-term memory,LSTM)、双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)、门控循环单元网络(gate recurrent unit,GRU)、双向门控循环单元(bidirectional gated recurrent unit,BiGRU)、多头注意力机制(multi-head attention,MHA)-BiGRU、Crossformer和Informer等对比模型,降幅分别为6.09%~43.47%和5.42%~36.45%;相较于表现最优的对比模型MHA-BiGRU,其误差指标分别降低约0.6%和5.42%,且R2提高约0.54%。工程应用验证表明,TFSA-Net模型在预测精度与泛化性能方面均具有优势,可为瓦斯灾害智能预警提供可靠技术支撑。
Abstract
To overcome the limitations of existing gas concentration prediction models in capturing multi-scale time-frequency features and modeling long-range temporal dependencies, time-frequency and stabilized attention network (TFSA-Net), a novel deep learning framework for forecasting working-face gas concentrations was proposed. The architecture first leveraged wavelet transform-based convolution to extract multi-scale time-frequency features, thereby enhancing the model’s ability to represent dynamic patterns across diverse temporal scales. Next, a stabilized attention-enhanced gated recurrent unit (SA-GRU) was incorporated to capture sequential dependencies effectively, with layer normalization and multi-head attention mechanisms jointly improving training stability and enabling more efficient exploitation of long-range contextual relationships. Finally, an adaptive spectral module was introduced to transform features into the frequency domain via the fast Fourier transform (FFT), where an adaptive high-frequency masking strategy selectively strengthened salient frequency components while attenuating noise interference. The experimental results show that TFSA-Net consistently outperforms state-of-the-art models such as Nlinear, LDlinear, long short-term memory (LSTM), bidirectional long short-term memory (BiLSTM), gate recurrent unit (GRU), bidirectional gated recurrent unit (BiGRU), multi-head attention (MHA)-BiGRU, Crossformer, and Informer in terms of mean absolute error (MAE) and root mean square error (RMSE) metrics, with reductions ranging from 6.09% to 43.47% and 5.42% to 36.45%, respectively. Compared to the best-performing comparative model MHA-BiGRU, its error metrics are reduced by approximately 0.6% and 5.42%, while the R2 score improves by about 0.54%. Real-world engineering validation demonstrates that TFSA-Net delivers superior prediction accuracy and generalization performance, providing reliable and robust technical support for intelligent early-warning systems in coal-mine gas hazard prevention.
WANG Jin-feng, JIA Peng-tao, LI Ai-jun, WANG Yi-ming, WANG Run-chao, YANG Hong-yu, WANG Wei-feng.
Gas Concentration Prediction Model for Coal Mining Faces Based on the TFSA-Net Framework[J].
Science Technology and Engineering, 2026, 26(24): 10342-10350 DOI:10.12404/j.issn.1671-1815.2506555
LiGuofu, LiChao, ZhangBichuan, et al. Gas drainage and utilization in coal mines in China: history, technological advances, and prospects[J]. Coal Geology & Exploration, 2025, 53(1): 77-91.
ZhangJianguo, WangWenchang, WeiFengqing, et al. Prediction of gas concentration in deep heading face based on the frequency spectrum acoustic method[J]. Journal of Mining Science and Technology, 2025, 10(5): 821-833.
FuHua, FuYu, ZhaoJuncheng, et al. Prediction of gas emission based on KPCA-ARIMA algorithm[J]. Journal of Liaoning Technical University (Natural Science), 2022, 41(5): 406-412.
WangJianhui, WangFujun, ChenWen, et al. Research on influencing factors of coal seam gas content based on feature derivation and machine learning[J]. Science Technology and Engineering, 2025, 25(23): 9748-9758.
QiaoWeihao, AnWeipeng, ZhaoXuehan, et al. Gas Prominence prediction based on GRA-BSMOTE-RF under unbalanced samples[J]. Coal Technology, 2024, 43(2): 121-125.
YangBaolin, LiuJie, LouLaizong. A prediction and early warning method for gas overrun based on PSO optimized BP neural network and SVM[J]. Science Technology and Engineering, 2025, 25(26): 11098-11106
QinYan, ShengWu. Research on multi-parameter coal mine gas concentration prediction and early warning based on PCA-GRU-SVM model[J]. Safety and Environmental Engineering, 2023, 30(6): 81-88.
LüJianli. Study on prediction of gas concentration in coal mine underground based on RNN algorithm[J]. China Energy and Environmental Protection, 2023, 45(9): 84-87.
LiuHaidong, LiXingcheng, ZhangWenhao. Research on the application of improved Adam training optimizer ingas emission prediction[J]. Journal of Mine Automation, 2023, 49(12): 25-32.
LanYongqing, QiaoYuandong, ChengHongming, et al. Gas concentration prediction model based on SSA-LSTM[J]. Journal of Mine Automation, 2024, 50(2): 90-97.
YangJunhui, MiaoChangsheng, ZhangJida, et al. Adaptive prediction of main gas concentration based on low-concentration gas direct current oxidation system[J]. Science Technology and Engineering, 2025, 25(24): 10182-10192.
WangXiangqian, WangQin, XuNingke. Prediction model of coal mine face gas concentration interval based on GRU and non-parametric kernel density estimation[J]. Mining Safety & Environmental Protection, 2025, 52(5): 8-15.
QinJiaxin, GeShuwei, LongFengqi, et al. Spatiotemporal distribution prediction of gas concentration based on GCN-GRU[J]. Journal of Mine Automation, 2023, 49(5): 82-89, 111.
ShengWu, FanBinbin. Application of the EMD GRU combined model for predicting gas volume fraction in coal mines[J]. Journal of Safety and Environment, 2025, 25(4): 1339-1348.
XieHengxing, ZhangXiong, DongJinyang, et al. Prediction model of mine gas emission based on CNN-BiLSTM[J]. Journal of Safety Science and Technology, 2024, 20(11): 53-59.
RongTongrui, HouEnke, XiaBingbing. Research on prediction method of coal mining face gas outflow based on quadratic decomposition and BO-BiLSTM combination model[J]. Safety in Coal Mines, 2024, 55(5): 83-92.
LiYupeng, JiaPengtao, LiuPing, et al. A gas concentration prediction model for working faces based on MHA-BiGRU[J]. China Mining Magazine, 2024, 33(12): 180-188.
HuQingsong, ZhengShuo, LiShiyin. An intelligent prediction method of gas concentration in coal mines based onimproved TCN TimeGAN[J]. Coal Science and Technology, 2024, 52(S2): 321-330.
WenTingxin, ChenSiyu. Prediction model of time series gas concentration based on IAPSO-Holt-TCN[J]. Journal of Safety Science and Technology, 2023, 19(7): 57-62.
[45]
ShahafE F, RoyA, EranT, et al. Wavelet convolutions for large receptive fields[C]// European Conference on Computer Vision. Cham: Springer. 2025: 363-380.
[46]
EmadeldeenE, MohamedR, ChenZ H, et al. TSLANet: rethinking transformers for time series representation learning[C]// Proceedings of the 41st International Conference on Machine Learning. Cambridge, MA: PMLR, 2024: 12409-12428.