煤炭能源是我国的主体能源，在维持国民经济稳健发展上具有重要作用。在煤矿的生产过程中，深部开采趋势增加使得工作面上更容易产生矿压灾害，对安全生产产生影响，给煤矿工人的生命安全带来了严重的威胁。绝大多数矿压灾害发生在来压期间，因此，预测工作面周期来压对预防矿压灾害、提供支护指导具有重要的意义。论文对工作面周期来压预测方法主要的研究内容如下：

提出了基于Adamax-GRU的工作面周期来压预测方法。对矿压原始数据进行预处理，采用均方误差做损失函数，应用Adamax优化算法对门控循环单元进行优化，建立了基于 Adamax-GRU的工作面周期来压预测模型。实验结果表明：相较于BPNN，LSTM和GRU，Adamax-GRU模型在测试集上误差分别减少了27.8%、14.1%和10.9%，预测精度提高，预测结果表现出来压的周期性，符合煤矿现场实际，能反映来压最大值。

提出了基于堆叠GRU的工作面周期来压时空预测方法。使用灰色关联度分析工作面矿压时序序列之间的空间关联关系，构造时空序列输入；采用堆叠式网络结构提高运算精度；构造一种基于堆叠GRU的工作面周期来压时空预测模型。实验结果表明：相较于单序列Adamax-GRU模型，堆叠GRU时空预测模型在测试集上误差减少了84.77%；相较于堆叠LSTM模型误差减少了21.2%。

将提出的工作面周期来压预测算法实际应用，设计并实现了煤矿动态矿压监测及智能预警分析系统，实现了系统的智能预警功能。在陕西某矿上进行工程实测，系统操作流畅，取得了较好的监测预警功能，能有效为煤矿工作面生产提供支护指导，对减少矿压灾害有实际作用。

Coal energy is the main source lor='red'>of energy in our country and plays an important role in maintaining the steady development lor='red'>of the national economy. In the production process lor='red'>of coal mines, the increasing trend lor='red'>of deep mining makes mine pressure hazards more likely to occur on the working surface, which has an impact on safe production and has brought serious threats to the lives lor='red'>of coal mine workers. The vast lor='red'>of mine pressure disasters occur during the pressure period, therefore, predicting the working face cycle to prevent mine pressure disasters and provide support guidance is lor='red'>of great significance. The main research contents lor='red'>of the paper on the prediction method lor='red'>of working face periodic pressure are as follows:

A method for forecasting the working face periodic pressure based on Adamax-GRU is proposed. Perform data preprocessing on the original mine pressure data, use the mean square error as the loss function, and apply the Adamax optimization algorithm to optimize the gated recurrent unit, and establish a working face periodic pressure forecasting model based on Admax-GRU. The experimental results show that: Compared with BPNN, LSTM and GRU, the Adamax-GRU model reduces the errors lor='red'>of the test set by 27.8%, 14.1% and 10.9%, respectively, and the prediction accuracy is improved. The prediction results reflect the periodicity, which is in line with the actual coal mine site, can reflect the maximum pressure.

A method based on stacked GRU working face periodic pressure spatio-temporal prediction is proposed. The gray correlation degree is used to analyze the spatial correlation between the mining pressure time series lor='red'>of the working face, and the input lor='red'>of the time-space sequence is constructed; the stacked network structure is used to improve the calculation accuracy; and a working face periodic pressure time-space prediction model based on the stacked GRU is constructed. The experimental results show that: compared with the single-sequence Adamax-GRU model, the RMSE lor='red'>of the stacked GRU spatiotemporal prediction model test set is reduced by 84.77%; compared with the stacked LSTM model test set RMSE is reduced by 21.2%.

Practical application lor='red'>of the proposed algorithm for forecasting working face periodic pressure is designed and implemented for coal mine dynamic pressure monitoring and intelligent early-warning analysis system, which realizes the system’s intelligent early-warning function. The project was tested in a mine in Shaanxi, and the system was operated smoothly and achieved good monitoring and early warning functions. It can effectively provide support and guidance for coal mining face production and has a practical effect on reducing mine pressure disasters.

[1] 王双明. 对我国煤炭主体能源地位与绿色开采的思考[J].中国煤炭,2020,46(02):11-16.

[2] 付华. 煤矿瓦斯灾害特征提取与信息融合技术研究[D]. 辽宁工程技术大学, 2006.

[3] 孙继平,钱晓红. 2004—2015 年全国煤矿事故分析[J]. 工矿自动化, 2016 (11): 1-5.

[4] 连清旺. 矿井顶板(围岩)状态监测及灾害预警系统研究及应用[D].太原理工大学,2012.

[5] 张国恩,解振华,史洪恺. 过空巷群支护技术与矿压显现规律特征分析[J].煤矿安全,2020,51(11):248-252.

[6] 刘航舵. 基于循环神经网络的瓦斯灾害预测预警研究[D].西安科技大学,2020.

[7] 窦林名,杨思先. 煤矿开采冲击矿压灾害防治[M]. 中国矿业大学出版社, 2006.

[8] 袁义鲁,魏建青,梁广锋等. 采用智能型数值模型对63_上10采面矿压及巷道变形作超前趋势分析──矿压显现的智能预测预报方法之一[J].煤矿开采,1996(S1):23-30.

[9] 屈世甲,李鹏. 基于支架工作阻力大数据的工作面顶板矿压预测技术研究[J].矿业安全与环保,2019,46(02):92-97.

[10] 徐刚,于健浩,范志忠等. 国内典型顶板条件工作面矿压显现规律研究[J/OL].煤炭学报:1-10[2021-04-08].

[11] 魏世明,王耀. 采场围岩支承压力分布特征研究现状分析[J].煤炭科学技术,2018,46(S1):18-22.

[12] 王延生,张勋,戚旭鹏等. 山丘地形综采面矿压显现多因素敏感性分析[J].辽宁工程技术大学学报(自然科学版),2017,36(03):238-242.

[13] 梁沙平,陆银龙,郭鹏等. 特厚煤层坚硬顶板初次破断特征的力学分析[J].煤矿安全,2020,51(08):245-250.

[14] 杨俊哲,郑凯歌,王振荣等. 坚硬顶板动力灾害超前弱化治理技术[J].煤炭学报,2020,45(10):3371-3379.

[15] 乔懿麟,王苏健,吴学明等. 基于模糊综合评价的冲击矿压危险性信息平台研究[J].煤炭技术,2016,35(11):253-255.

[16] 惠艳荔. 电磁辐射技术在采煤工作面矿压预测中的应用[J].煤炭技术,2013,32(12):73-75.

[17] 郭辉,王新萍,芦盛亮等. 基于微震监测的高强度开采工作面煤岩破坏规律研究[J]. 矿业研究与开发, 2018, 38(12):56-60.

[18] 李静,何江,巩思园等. 上覆遗留煤柱作用下冲击矿压预测预警案例研究[J].煤炭学报,2016,41(S2):305-310.

[19] 刘昆轮,闫瑞兵. 基于地音监测的近直立煤层冲击地压前兆特征研究[J].煤炭工程,2020,52(04):48-51.

[20] Li Y, Shi X. Mine Pressure Prediction Study Based on Fuzzy Cognitive Maps[J]. International Journal of Computational Intelligence and Applications, 2020, 19(03): 2050023.

[21] 刘卫东,孟晓静,丁恩杰等. 模糊逻辑在矿井冲击矿压预测中的应用[J].煤矿安全,2008(01):66-68.

[22] Toderas M, Moraru R I. Mine pressure and rock displacement temporal variation analysis for Jiu Valley directional drifts in the context of" n" stability criterion[J]. Archives of Mining Sciences, 2014, 59.

[23] 贺超峰,华心祝,杨科等. 基于BP神经网络的工作面周期来压预测[J].安徽理工大学学报(自然科学版),2012,32(01):59-63.

[24] Wang K, Zhuang X, Zhao X, et al. Roof Pressure Prediction in Coal Mine Based on Grey Neural Network[J]. IEEE Access, 2020, 8: 117051-117061.

[25] 尹光志,代高飞,闫河等. 冲击地压预测的遗传神经网络方法[J].岩土力学,2003(06):1016-1020.

[26] 王雨虹,刘璐璐,付华等. 基于改进BP神经网络的煤矿冲击地压预测方法研究[J].煤炭科学技术,2017,45(10):36-40.

[27] 杨硕,王胄. 用粒子群优化BP神经网络的顶板矿压预测[J].陕西煤炭,2010,29(04):83-86.

[28] 崔铁军,马云东. 基于泛函网络的周期来压预测方法研究[J].计算机科学,2013,40(S1):243-246.

[29] 吕鹏飞,邱林. 基于PSO-LSSVM的矿井冲击地压分级预测研究[J].矿业安全与环保,2021,48(01):120-125.

[30] 赵毅鑫,杨志良,马斌杰等. 基于深度学习的大采高工作面矿压预测分析及模型泛化[J/OL].煤炭学报: 2020,45(01):54-65.

[31] Ying G . A weighted cluster kernel PCA prediction model for multi-subject brain imaging data[J]. Statistics & Its Interface, 2010, 3(1):103-112.

[32] Hinton G E, Srivastava N, Krizhevsky A , et al. Improving neural networks by preventing co-adaptation of feature detectors[J]. Computer Science, 2012, 3(4):págs. 212-223.

[33] 那守范著. 顶板[M].北京:煤炭工业出版社,1983.09.

[34] 徐刚,于健浩,范志忠等. 国内典型顶板条件工作面矿压显现规律研究[J/OL].煤炭学报:1-10[2021-03-05].

[35] 秋丰岐,朱磊,古文哲等. 浅埋煤层大采高工作面矿压显现规律及支架工作阻力研究[J].煤矿安全,2020,51(11):243-247.

[36] 李青锋,彭跃金,吴昊等. 基于微震与矿山压力监测的工作面周期来压步距与机理分析[J].采矿与安全工程学报,2020,37(03):571-577.

[37] 黄庆享. 浅埋煤层的矿压特征与浅埋煤层定义[J]. 岩石力学与工程学报, 2002, 21(008):1174-1177.

[38] 逄锦伦. 近距离煤层群顶板结构及破断模型分析[J].矿业安全与环保,2020,47(03):105-109.

[39] Zeki-Suac M, Has A, Kneevi M. Predicting Energy Cost of Public Buildings by Artificial Neural Networks, CART, and Random Forest[J]. Neurocomputing, 2021.

[40] Jain A, Zamir A R, Savarese S, et al. Structural-RNN: Deep Learning on Spatio-Temporal Graphs[C]// IEEE Conference on Computer Vision & Pattern Recognition. IEEE, 2015.

[41] 杨丽,吴雨茜,王俊丽等. 循环神经网络研究综述[J].计算机应用,2018,38(S2):1-6+26.

[42] Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural Computation (S0899-7667), 1997, 9(8): 1735-1780.

[43] Cho, Kyunghyun, van Merrienboer, Bart, Gulcehre, Caglar. Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation[J]. Computer science, 2014.

[44] Yanhua Guo,Lei Sun,Zhihong Zhang,Hong He. Algorithm Research on Improving Activation Function of Convolutional Neural Networks [C]. 第31届中国控制与决策会议,2019:5.

[45] H Chen, Jiang L , H Yang, et al. An Efficient Hardware Architecture with Adjustable Precision and Extensible Range to Implement Sigmoid and Tanh Functions[J]. Electronics, 2020, 9(10):1739.

[46] STURSA D, DOLEZEL P. Comparison of ReLU　and Linear Saturated Activation Functions in　Neural Network for Universal　Approximation[C]// International Conference on Process Control,2019:146-151.

[47] D.P. Kingma, J. Ba. Adam: A Method for Stochastic Optimization. The International Conference on Learning Representations (ICLR) , 2015.

[48] Richard S Sutton. Two Problems with Backpropagation and Other Steepest-Descent Learning Procedures for Networks, 1986.

[49] Neyshabur B , Salakhutdinov R , Srebro N . Path-SGD: Path-Normalized Optimization in Deep Neural Networks[J]. MIT Press, 2015.

[50] Kingma D , Ba J . Adam: A Method for Stochastic Optimization[J]. Computer Science, 2014.

[51] Poudel Prasis,Bae Sang Hyun,Jang Bongseog. Comparison of Different Deep Learning Optimizers for Modeling Photovoltaic Power[J]. Journal of the Chosun Natural Science,2018,11(4).

[52] 秦玉金,苏伟伟,姜文忠等. 我国矿井瓦斯涌出量预测技术研究进展及发展方向[J].煤矿安全,2020,51(10):52-59.

[53] Nam H, Kim S, Jung K. Number sequence prediction problems for evaluating computational powers of neural networks[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2019, 33: 4626-4633.

[54] Xing Y , Yue J , Chen C , et al. Dynamic Displacement Forecasting of Dashuitian Landslide in China Using Variational Mode Decomposition and Stack Long Short-Term Memory Network[J]. Applied Sciences, 2019, 9(15):2951.

[55] 山世光,阚美娜,刘昕等. 深度学习:多层神经网络的复兴与变革[J].科技导报,2016,34(14):60-70.

[56] 煤炭工业部安全司. 矿井安全监测监控原理与应用[M]. 徐州: 中国矿业大学出版社,1996.