基于CNN_BiLSTM的矿井瓦斯涌出量预测模型  

Prediction model of mine gas emission based on CNN_BiLSTM

作　　者：解恒星 张雄 董锦洋 刘晓东 姚小兵 毕振彪 李磊 XIE Hengxing;ZHANG Xiong;DONG Jinyang;LIU Xiaodong;YAO Xiaobing;BI Zhenbiao;LI Lei(Guizhou Jinsha Longfeng Coal Industry Co.,Ltd.,Bijie Guizhou 551700,China;School of Safety and Emergency Management Engineering,Taiyuan University of Technology,Jinzhong Shanxi 030024,China;Shanxi Energy Holdings Coal Industry Group,JinCheng Coal Mine Ventilation Department,Jincheng Shanxi 048000,China)

机构地区：[1]贵州金沙龙凤煤业有限公司,贵州毕节551700 [2]太原理工大学安全与应急管理工程学院,山西晋中030024 [3]晋能控股煤业集团晋城煤炭通风部,山西晋城048000

出　　处：《中国安全生产科学技术》2024年第11期53-59,共7页Journal of Safety Science and Technology

基　　金：太原市关键核心技术攻关“揭榜挂帅”项目(2024TYJB0139)。

摘　　要：为了实现对瓦斯涌出量准确预测,从而有效预防瓦斯灾害。提出1种结合卷积神经网络(convolutional neural network,CNN)和双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)的瓦斯涌出量预测模型,采用CNN在时间序列上提取瓦斯涌出量及其影响因素的局部关键特征,有效捕捉数据的局部时序相关性;BiLSTM模型利用这些特征,通过其前向和后向处理能力,全面捕捉时间序列中长期依赖性和复杂模式。研究结果表明:该模型预测准确率达93.6%,均方误差显著低于CNN、BPNN、LSTM、BiLSTM、CNN_LSTM、CNN_BiLSTM 6个模型,决定系数接近1,表明其出色的预测能力和解释力。研究结果可有效预测瓦斯涌出量波动,有助于提高矿井瓦斯风险预警能力,提升矿井安全管理水平。In order to achieve the accurate prediction of gas emission quantity and effectively prevent gas disaster,a prediction model of gas emission quantity combining the convolutional neural network(CNN)and bidirectional long short-term memory(BiLSTM)network was proposed.The gas emission quantity and key local features of its influencing factors were extracted from time series by using CNN,and the local temporal correlation of data was effectively captured.These features were used by the BiLSTM model,which leveraged its forward and backward processing capabilities to comprehensively capture the long-term dependency and complex patterns in the time series.The results show that the model has an accuracy rate of 93.6%,with a significantly lower mean square error than the CNN,back propagation neural network(BPNN),long short-term memory(LSTM),BiLSTM,convolutional neural network-long short-term memory(CNN_LSTM),and CNN_BiLSTM models.The coefficient of determination is close to 1,indicating its excellent predictive ability and explanatory power.The research results can effectively predict the fluctuation of gas emission quantity,contribute to improve the gas risk warning capability,and enhance the safety management level of mines.

关 键 词：瓦斯涌出量预测模型 卷积神经网络 双向长短时记忆网络 反向神经网络 基线对比 

分 类 号：X936[环境科学与工程—安全科学]