随着煤炭开采深度的持续增加，因矿井围岩应力增高而造成的煤炭生产事故发生强度和频度也不断增加，严重威胁着煤炭安全生产。因此，为了避免此类事故的发生，必须要对围岩应力采取必要的监测预警手段，从而保障矿井安全高效生产。论文针对矿井围岩应力监测数据分析中存在的可靠性低、预测精度不满足现场需求的问题，基于深度学习的理论对矿井围岩应力预测方法展开研究，主要研究内容如下：

（1）提出了基于多传感器支持度与自适应加权的围岩应力融合算法SDAWS（data fusion algorithm of surrounding rock stress based on support degree and adaptive weighted）。该算法针对矿井围岩应力监测数据来源多、数据可靠性低的问题，将多个位置的工作面支架阻力进行数据融合。通过长短期记忆网络模型分别对经过SDAWS算法融合前后的围岩应力数据进行预测，实验结果表明：相对于未融合的数据，经过SDAWS算法融合后的围岩应力数据进行预测，在训练集和测试集的均方误差分别降低了31.16%，33.77%。

（2）提出了基于Attention-CNN-HN（Highway Network Integrated with Attention mechanism and Convolutional Neural Network）的矿井围岩应力预测方法。构建基于highway神经网络的围岩应力预测模型，采用均方误差做损失函数，确定模型的超参数。采用CNN自动提取围岩应力的特征信息，再将提取特征输入到attention层，为不同的输入特征分配相应的权重。然后应用highway神经网络对围岩应力进行时序性数据预测。实验结果表明：相对于Attention-CNN-LSTM和Attention-CNN-GRU两种模型，Attention-CNN-HN在测试集上的均方误差分别降低了37.06%，12.17%，且运行效率分别提高了17.36%，5.85%。

（3）采用面向对象的思想，在JAVA平台下，基于SDAWS围岩应力融合算法与Attention-CNN-HN矿井围岩应力预测模型，设计并实现了矿井围岩应力预警系统。系统具有动态围岩应力监测、数据统计查询、围岩应力预警、系统设置四个功能。通过对系统的分析、设计、实现、测试等多个环节，实现了系统的应用。系统达到了界面美观、操作简便的目标。现场应用效果表明：系统具有较好的应用价值，为井下安全生产提供了良好保障。

With the increasing deepening of the coal mining depth, the intensity and frequency of coal mine production accidents caused by the enhancement of coal mine surrounding rock stress, which seriously threatens the safety of coal production. Therefore, in order to avoid such accidents, it is necessary to take essential means to the monitor and early-warning of the surrounding rock stress and ensure the safe and efficient production of the coal mine. This thesis is developed on the prediction methods of coal mine surrounding rock stress based on deep learning in order to solve the problems of low data reliability and unsatisfactory prediction precision in mine surrounding rock stress monitoring. The main researching content of this thesis is as follows:

(1) A data fusion algorithm of surrounding rock stress based on support degree and adaptive weighted (SDAWS) of multi sensors is proposed in this thesis. Aiming at the problems of the mine surrounding rock stress monitoring data with multi-source and low data reliability, the algorithm fused the data of working yield of support in working faces in multiple positions. The surrounding rock stress data before fusion and after fusion were predicted respectively through the long-term and short-term memory network model. The results indicate that the mean square error is reduced by 31.16% and 33.77% in the training set and the test set, respectively.

(2) The stress prediction method of coal mine surrounding rock based on Highway Network Integrated with Attention mechanism and Convolutional Neural Network is proposed in this thesis. A prediction model of surrounding rock stress based on highway neural network is established, and the mean square error is used as the loss function to determine the hyper parameters of the model. Firstly, Convolutional Neural Network is used to automatically extract the characteristic information of surrounding rock stress. Then, the extracted features are input into the attention layer to assign weights to different input features. Finally, the model of surrounding rock stress prediction is established by the application of highway neural network to the time series data prediction of surrounding rock stress. The experimental results show that Attention-CNN-HN model achieves less mean square error on the test set by 37.06% and 12.17% respectively, compared with Attention-CNN-LSTM and Attention-CNN-GRU model. In addition, the operating efficiency of Attention-CNN-HN model is increased by 17.36% and 5.85% respectively.

(3) The early-warning system of mine surrounding rock stress based on SDAWS algorithm and Attention-CNN-HN model is designed and implemented in this thesis. The system is analyzed and designed with the idea of object-oriented. Under the Java platform, four functions of dynamic surrounding rock stress monitoring, data statistical query, early-warning of surrounding rock stress and system setting are realized. Through the analysis, design, realization and test of the early-warning system of mine surrounding rock stress, the application of the system has been realized. The system has the features of beautiful interface and simple operation. The field application effect shows that the software has good application value and provides a good guarantee for the safety of underground production.

[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] 杨英明,孙建东,李全生.我国能源结构优化研究现状及展望[J].煤炭工程,2019,51(02):149-153.

[3] 杜贝举,刘长友,吴锋锋,杨敬轩.深井高应力软弱围岩巷道变形机理及控制研究[J].采矿与安全工程学报,2020,37(06):1123-1132.

[4] 范志忠,潘黎明,徐刚,尹希文.智能化高强度开采超长工作面围岩灾变预警技术[J].煤炭科学技术,2019,47(10):125-130.

[5] Zhipeng Shen, Yuanming Zhang, Jiawei Lu, Jun Xu, Gang Xiao. A novel time series forecasting model with deep learning[J]. Neurocomputing,2020,396.

[6] 张旭. 基于循环神经网络的时间序列预测方法研究[D].南京大学,2019.

[7] 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.

[8] 鹿天柱,钱晓超,何舒,谭振宁,刘飞.一种基于深度学习的时间序列预测方法[J].控制与决策,2021,36(03):645-652.

[9] 钱青,唐桂忠,张广明,邓歆,尹海培.基于AR-DBN的建筑分项能耗短期预测[J].计算机工程,2019,45(06):290-296.

[10] 邵良杉,周玉.基于MIV-MA-KELM模型的岩爆烈度等级预测[J].中国安全科学学报,2018,28(02):34-39.

[11] Paul W. Talbot,Cristian Rabiti,Andrea Alfonsi,Cameron Krome,M. Ross Kunz,Aaron Epiney,Congjian Wang,Diego Mandelli. Correlated synthetic time series generation for energy system simulations using Fourier and ARMA signal processing[J]. International Journal of Energy Research,2020,44(10).

[12] Tripti Dimri,Shamshad Ahmad,Mohammad Sharif. Time series analysis of climate variables using seasonal ARIMA approach[J]. Journal of Earth System Science: Published by the Indian Academy of Sciences,2020,129(3).

[13] Box G E P , Jenkins G M . Time series analysis: forecasting and control[J]. Journal of Time, 2010, 31(3).

[14] Thao-Tsen Chen,Shie-Jue Lee. A weighted LS-SVM based learning system for time series forecasting[J]. Information Sciences,2015,299.

[15] Time series prediction of mining subsidence based on a SVM[J]. Mining Science and Technology,2011,21(04):557-562.

[16] Rubén Loaiza-Maya,Michael Stanley Smith. Variational Bayes Estimation of Discrete-Margined Copula Models With Application to Time Series[J]. Journal of Computational and Graphical Statistics,2019,28(3).

[17] Eitan Greenshtein,Ariel Mantzura,Ya’acov Ritov. Nonparametric empirical Bayes improvement of shrinkage estimators with applications to time series[J]. Bernoulli,2019,25(4B).

[18] Xueheng Qiu,Le Zhang,Ponnuthurai Nagaratnam Suganthan,Gehan A.J. Amaratunga. Oblique random forest ensemble via Least Square Estimation for time series forecasting[J]. Information Sciences,2017,420.

[19] Hristos Tyralis,Georgia Papacharalampous. Variable Selection in Time Series Forecasting Using Random Forests[J]. Multidisciplinary Digital Publishing Institute,2017,10(4).

[20] Li Hui,Li Fan,Jia Rong,Zhai Fang,Bai Liang,Luo Xingqi. Research on the Fault Feature Extraction of Rolling Bearings Based on SGMD-CS and the AdaBoost Framework[J]. Energies,2021,14(6).

[21] Biao Xiong,Ruiping Li,Dong Ren,Huigang Liu,Tao Xu,Yingping Huang. Prediction of flooding in the downstream of the Three Gorges Reservoir based on a back propagation neural network optimized using the AdaBoost algorithm[J]. Natural Hazards,2021(prepublish).

[22] Omer Berat Sezer,Mehmet Ugur Gudelek,Ahmet Murat Ozbayoglu. Financial time series forecasting with deep learning : A systematic literature review: 2005–2019[J]. Applied Soft Computing Journal,2020,90.

[23] Silvio Barra,Salvatore Mario Carta,Andrea Corriga,Alessandro Sebastian Podda,Diego Reforgiato Recupero.Deep Learning and Time Series-to-Image Encoding for Financial Forecasting[J].IEEE/CAA Journal of Automatica Sinica,2020,7(03):683-692.

[24] Zahra Karevan,Johan A.K. Suykens. Transductive LSTM for time-series prediction: An application to weather forecasting[J]. Neural Networks,2020,125.

[25] Mingming Gao,Jianjing Li,Feng Hong,Dongteng Long. Day-ahead power forecasting in a large-scale photovoltaic plant based on weather classification using LSTM[J]. Energy,2019,187.

[26] Wang Y, Li L, Xu X. A piecewise hybrid of ARIMA and SVMs for short-term traffic flow prediction[C]. International Conference on Neural Information Processing. Guangzhou: Springer, 2017:493-502.

[27] Tao Ma,Constantinos Antoniou,Tomer Toledo. Hybrid machine learning algorithm and statistical time series model for network-wide traffic forecast[J]. Transportation Research Part C,2020,111.

[28] Amin Morid Mohammad,Liu Sheng Olivia R.,Kawamoto Kensaku,Abdelrahman Samir. Learning Hidden Patterns from Patient Multivariate Time Series Data Using Convolutional Neural Networks: A Case Study of Healthcare Cost Prediction[J]. Journal of Biomedical Informatics,2020,111(prepublish).

[29] Hong Shenda,Wang Can,Fu Zhaoji. Gated temporal convolutional neural network and expert features for diagnosing and explaining physiological time series: A case study on heart rates[J]. Computer Methods and Programs in Biomedicine,2020.

[30] Maleki Mohsen,Mahmoudi Mohammad Reza,Wraith Darren,Pho KimHung. Time series modelling to forecast the confirmed and recovered cases of COVID-19.[J]. Travel medicine and infectious disease,2020,37.

[31] Hongchao Qi,Shuang Xiao,Runye Shi,Michael P. Ward,Yue Chen,Wei Tu,Qing Su,Wenge Wang,Xinyi Wang,Zhijie Zhang. COVID-19 transmission in Mainland China is associated with temperature and humidity: A time-series analysis[J]. Science of the Total Environment,2020,728.

[32] Nikhil Ketkar. Deep Learning with Python[M].Apress, Berkeley, CA:2017-01-01.

[33] Jü,rgen Schmidhuber. Learning Complex, Extended Sequences Using the Principle of History Compression[J]. Neural Computation,1992,4(2).

[34] Lecun, Y, Bottou, L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 86(11):P.2278-2324.

[35] Wang Rui,Chen Shiwen,Lu Jing. Electric short-term load forecast integrated method based on time-segment and improved MDSC-BP[J]. Systems Science & Control Engineering,2021,9(S1).

[36] 曹卫东,许志香,王静.基于深度生成模型的半监督入侵检测算法[J].计算机科学,2019,46(03):197-201.

[37] 王雅思,姚鸿勋,孙晓帅,许鹏飞,赵思成.深度学习中的自编码器的表达能力研究[J].计算机科学,2015,42(09):56-60+65.

[38] Alberto Mozo,Bruno Ordozgoiti,Sandra Gómez-Canaval. Forecasting short-term data center network traffic load with convolutional neural networks[J]. PLOS ONE,2018,13(2).

[39] Boubacar Doucoure,Kodjo Agbossou,Alben Cardenas. Time series prediction using artificial wavelet neural network and multi-resolution analysis: Application to wind speed data[J]. Renewable Energy,2016,92.

[40] Mahlagha Afrasiabi,Hassan khotanlou,Muharram Mansoorizadeh. DTW-CNN: time series-based human interaction prediction in videos using CNN-extracted features[J]. The Visual Computer,2019(prepublish).

[41] Xin Wei,Lulu Zhang,Hao-Qing Yang,Limin Zhang,Yang-Ping Yao.Machine learning for pore-water pressure time-series prediction:Application of recurrent neural networks[J].Geoscience Frontiers,2021,12(01):453-467.

[42] Xindi Cai,Nian Zhang,Ganesh K. Venayagamoorthy,Donald C. Wunsch. Time series prediction with recurrent neural networks trained by a hybrid PSO–EA algorithm[J]. Neurocomputing,2005,70(13).

[43] Bengio, Y. Learning Deep Architectures for AI[J]. Foundations & Trends® in Machine Learning, 2(1):1-127.

[44] Jian Cao,Zhi Li,Jian Li. Financial time series forecasting model based on CEEMDAN and LSTM[J]. Physica A: Statistical Mechanics and its Applications,2018,519.

[45] Alaa Sagheer,Mostafa Kotb. Time Series Forecasting of Petroleum Production using Deep LSTM Recurrent Networks[J]. Neurocomputing,2018,323.

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

[47] Xu Zhang,Furao Shen. Time Series Forecasting Using GRU Neural Network with Multi-lag After Decomposition[J]. Springer, Cham, 2017.

[48] Qianting Li,Tong Xu. VS-GRU: A Variable Sensitive Gated Recurrent Neural Network for Multivariate Time Series with Massive Missing Values[J]. Applied Sciences,2019,9(15):3041.

[49] Chiou-Jye Huang,Ping-Huan Kuo. A Deep CNN-LSTM Model for Particulate Matter (PM2.5) Forecasting in Smart Cities[J]. Sensors,2018,18(7).

[50] 蔡舒凌,李二兵,陈亮,高磊,濮仕坤,段建立,谭跃虎.基于FA-NAR动态神经网络的隧洞围岩变形时序预测研究[J].岩石力学与工程学报,2019,38(S2):3346-3353.

[51] 黄敏,唐绍辉,黄英华,吴亚斌.基于ANSYS-FLAC~(3D)的采空区稳定性分析数值模拟研究[J].矿业研究与开发,2017,37(06):78-83.

[52] 王初步,谭富生,王宇,李智,李真.基于数值模拟的复杂采空区稳定性分析[J].矿业研究与开发,2018,38(01):96-100.

[53] 郭延辉,侯克鹏.深埋炭质千枚岩巷道松动圈测试及主被动联合支护研究[J].地下空间与工程学报,2021,17(01):214-221.

[54] 刘叶玲, 张海燕, 来兴平. 基于SVM的煤岩破裂与失稳预测模型[J]. 煤田地质与勘探, 2007(03):65-68.

[55] 王茂盛,朱晔,李永亮,郭东明,李炜煜,王茂源.基于模糊聚类方法的煤巷围岩稳定性分类研究[J].矿业科学学报,2018,3(03):238-245.

[56] 赵汝星.基于随机森林的回采巷道围岩稳定性分类[J].煤矿安全,2014,45(11):200-202+206.

[57] 赵毅鑫,杨志良,马斌杰,宋红华,杨东辉.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].煤炭学报,2020,45(01):54-65.

[58] 赵铭生,刘守强,纪润清,刘德民,曹欢,胡棉舒.基于遗传算法优化BP神经网络的华北型煤田矿压破坏带深度预测[J].矿业研究与开发,2020,40(06):89-93.

[59] Yue Pan,Liang Chen,Ju Wang,Hongsu Ma,Shuling Cai,Shiku Pu,Jianli Duan,Lei Gao,Erbing Li. Research on deformation prediction of tunnel surrounding rock using the model combining firefly algorithm and nonlinear auto-regressive dynamic neural network[J]. Engineering with Computers,2019(prepublish).

[60] 王军号. 基于物联网感知的煤矿安全监控信息处理方法研究[D].安徽理工大学,2013.

[61] 孙田川,刘洁瑜.基于支持度和自适应加权的MEMS陀螺信息融合算法[J].传感技术学报,2016,29(10):1548-1552.

[62] 敬如雪,高玉琢.基于多传感器的数据融合算法研究[J].现代电子技术,2020,43(10):10-13.