我国煤炭资源十分丰富，但开采条件并不理想，煤矿安全事故时有发生，其中顶板事故位居我国各类煤矿事故危害之首。采煤工作面矿压预测是顶板灾害预警的重要手段，对保障煤矿安全、经济、高效开采具有重要意义。现有矿压预测方法存在两方面问题：①在构建训练样本集时没有深入分析矿压数据间的时空关联性。②部分时刻矿压预测值偏差极大。这些问题严重影响了预测精度的提升。为了实现采煤工作面高精度的矿压预测，本文主要做了以下研究工作：

第一，提出一种基于时空关联分析的样本集构建方法。首先基于灰色关联度对矿压数据进行时空关联分析，识别出最优的时间窗口以及紧密关联液压支架群。其次根据时空关联分析的结果构建训练样本集，训练LSTM预测模型。最后以平均绝对误差作为评价指标对预测模型进行评估。实验结果表明，与传统依赖人工检验构造样本集的方法相比，基于时空关联分析构建样本集的方法使LSTM矿压预测模型的预测误差平均降低了4.3%。

第二，提出一种基于条件熵的来压概率估计方法，对工作面进行来压概率预测。首先对原始矿压数据进行来压划分，生成来压状态序列。其次基于条件熵计算最佳历史状态条件数。再次找出长度为最佳历史状态条件数的所有历史状态序列，统计其下一个时刻来压概率，构建来压概率预测树。最后根据来压概率预测树实现工作面的来压概率预测。

第三，提出一种基于来压概率估计的矿压预测修正方法，利用来压概率的估计结果对时空关联分析的矿压预测结果进行修正。对于未来某一时刻，同时进行矿压预测和来压概率预测，当两者预测结果出现矛盾时，计算矿压的期望值作为最终的矿压预测值，以此来修正预测结果。实验结果表明，与时空关联分析方法相比，本方法的预测误差平均降低了4.1%。

最后，在前面的研究基础上，设计并开发了采煤工作面矿压特征分析及预测系统。该系统实现了矿压特征分析和预测功能，并将运行结果进行直观展示。

China is very rich in coal resources, but the mining conditions are not ideal. Coal safety accidents occur from time to time, among which the roof accident ranks first in the hazards of various coal mine accidents in our country. Coal mine pressure prediction of working face is an important means of early warning of roof disasters, and it is of great significance to ensure safe, economical and efficient mining of coal mines. There are two problems in the existing coal mine pressure prediction methods: ①There is no in-depth analysis of the spatiotemporal correlation among the mine pressure data when constructing the training sample set. ②In some moments, the predicted value of mine pressure has great deviation. These problems seriously affect the improvement of prediction accuracy. In order to achieve high-precision mine pressure prediction of coal mining face, this paper mainly does the following work:

Firstly, a sample set construction method based on spatiotemporal correlation analysis is proposed. First, based on the grey correlation degree, the spatiotemporal correlation analysis of the rock pressure data is carried out, and the optimal time window and the closely related hydraulic support group are identified. Secondly, a training sample set is constructed according to the results of the spatiotemporal correlation analysis, and the LSTM prediction model is trained. Finally, the prediction model is evaluated with the mean absolute error as the evaluation index. The experimental results show that, compared with the traditional LSTM model, the prediction error of the LSTM model constructed by this method is reduced by 4.3% on average.

Secondly, an approach probability estimation method based on conditional entropy is proposed to predict the approach probability of the working face. First, the original rock pressure data is divided into pressure to generate a series of pressure state. Second, the optimal historical state condition number is calculated based on the conditional entropy. Third, the best historical state condition number is found out, and the probability prediction tree is constructed. Finally, according to the probability prediction tree, the probability prediction of the working face is realized.

Thirdly, a method for revising coal mine pressure prediction based on probability estimation of incoming pressure is proposed, which uses the prediction result of the probability of arrival to revise the prediction result of the spatiotemporal correlation analysis. For a certain moment in the future, the prediction of the mine pressure and the probability of the coming pressure are carried out at the same time. When the two prediction results are contradictory, the expected value of the mine pressure is calculated as the final predicted value of the mine pressure, so as to correct the prediction result. The experimental results show that, compared with the spatiotemporal correlation analysis method, the prediction error of this method is reduced by 4.1% on average.

Finally, on the basis of the previous research, a characteristic analysis and prediction system of rock pressure in coal mining face is designed and developed. The system realizes the features of rock pressure analysis and prediction, and visually displays the operation results.

参考文献

[1]张维宸.能源安全视角下煤炭大国开采政策对比[J].国土资源情报,2021(03):16-27.

[2]中华人民共和国自然资源部,中国矿产资源报告2020,北京:地质出版社,2020.

[3]苗琦,孟刚,陈敏,等.我国煤炭资源可供性分析及保障研究[J].能源与环境,2020(02):6-8+23.

[4]牛超,施龙青,肖乐乐,等.2001-2013年煤矿生产事故分类研究[J].煤矿安全,2015,46(03): 208-211.

[5]杜龙龙,宋宜猛,冯宇峰.煤矿顶板事故分析与防治对策研究[J].中国煤炭,2020,46(10): 50-54.

[6]康红普,张镇,黄志增.我国煤矿顶板灾害的特点及防控技术[J].煤矿安全,2020,5(10): 24-33,38

[7]宁小亮.2013—2018年全国煤矿事故规律分析及对策研究[J].工矿自动化,2020,46(07): 34-41.

[8]侯刚.煤矿智能化开采系统功能及典型模式分析研究[J].中国煤炭,2022,48(02):55-60.

[9]王海波,贺耀宜.深耕煤矿安全生产领域 助力煤矿智能化建设——中煤科工集团常州研究院服务煤矿智能化建设之路[J].智能矿山,2022,3(02):2-10.

[10]王国法,任怀伟,赵国瑞,巩师鑫,杜毅博,薛忠新,庞义辉,张潇.智能化煤矿数据模型及复杂巨系统耦合技术体系[J].煤炭学报,2022,47(01):61-74.

[11]许日杰,杨科,吴劲松,阚磊.麻地梁煤矿智能化开采研究[J].工矿自动化,2021,47(11):9-15.

[12]武福生,卜滕滕,王璐.煤矿安全智能化及其关键技术[J].工矿自动化,2021,47(09):108-112.

[13]虎晓龙,殷华.煤矿智能化开采技术创新与发展研究[J].工矿自动化,2021,47(S2):10-12.

[14]Yang Y, Ma Y, Ji C, et al. Effect of mining thickness on overburden movement and underground pressure characteristics for extrathick coal seam by sublevel caving with high bottom cutting height. Advances in Civil Engineering, 2018, 1-15.

[15]钱鸣高,缪协兴,何富连,刘长友.采场支架与围岩耦合作用机理研究[J].煤炭学报,1996(01):40-44.

[16]宋振骐.实用矿山压力控制[M]徐州:中国矿业大学出版社,1988.

[17]曾庆田,吕珍珍,石永奎,等.基于Prophet+LSTM模型的煤矿井下工作面矿压预测研究[J].煤炭科学技术,2021,49(07):16-23.

[18]Whittaker B N, Potts E L. Appraisal of strata control practice. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 1974, 11(11), 27-28.

[19]Alehossein H, Poulsen B A. Stress analysis of long wall top coal caving. International Journal of Rock Mechanics & Mining Sciences, 2010, 47(1), 30-41.

[20]Song X, Gu T, Yan Z. Effects of increasing working face’s length on underground pressure behaviors of mining super-high faces under shallow coal seam. Chinese Journal of Rock Mechanics and Engineering, 2016, 2007(S2), 4007-4012.

[21]Zhao Y, Wang S, et al. Pressure-arching characteristics in roof blocks during shallow coal mining. Advances in Civil Engineering, 2018, 1-12.

[22]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,(8):245-250.

[23]Chen J, Gao J, Pu Y, et al. Machine learning method for predicting and warning of rock burst. Journal of Minging and Strata Control Engineering, 2021,3(1), 57-68.

[24]谷栓成,李昂.灰色预测理论在回采工作面周期来压步距中的应用[J].煤矿安全,2010(10):61-64.

[25]贺超峰,华心祝,杨科,马菁花.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.

[26]张洋,马云东,崔铁军.基于小波和混沌优化LSSVM的周期来压预测[J].安全与环境学报,2014,14(04):63-66.

[27]Zhang T, Zhao Y, Zhu G, et al. A multi-coupling analysis of mining-induced pressure characteristics of shallow-depth coal face in Shendong mining area. Journal of China Coal Society, 2016, 41(S2), 287-296.

[28]Zhang T, Zhao Y, Zhu G, et al. A multi-coupling analysis of mining-induced pressure characteristics of shallow-depth coal face in Shendong mining area. Journal of China Coal Society, 2016, 41(S2), 287-296.

[29]弓仲标,冯晓斌,贺斌.特厚煤层下分层综放面初采期间矿压显现规律分析[J].陕西煤炭,2021,40(05):41-45.

[30]李云飞,解振华,杨龙龙.综采工作面不规律生产期间矿压显现规律分析[J].陕西煤炭,2021,40(05):14-16+25.

[31]闫吉太,梁广锋,安满林,卫建清,冯增强.“孤岛”综采放顶煤工作面矿压预测预报[J].中国矿业大学学报,1996(04):98-103.

[32]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.

[33]冀汶莉,刘艺欣,柴敬,王斌.基于随机森林的矿压预测方法[J].采矿与岩层控制工程学报,2021,3(03):71-81.

[34]柴敬,王润沛,雷武林.基于遗传-支持向量机的分布式光纤监测矿压时序预测[J].科学技术与工程,2020,20(32):13137-13142.

[35]尹希文,徐刚,刘前进,卢振龙,于秋鸽,张震.基于支架载荷的矿压双周期分析预测方法[J].煤炭学报,2021,46(10):3116-3126.

[36]Qian M G. Mining pressure and strata control, China University of Mining and Technology: Xuzhou, China, 2010, 100-119.

[37]Tan T, Yang Z, Chang F, et al. Prediction of the first weighting from the working face roof in a coal mine based on a GA-BP neural network[J]. Applied Sciences, 2019, 9(19): 4159.

[38]Yu J, Li Y. Intelligent optimization for support pressure zone algorithm based on fuzzy neural network damage identification. EURASIP Journal on Wireless Communications and Networking, 2018, 2018(1).

[39]Liu Y, Ouyang Y, Zhu P, et al. A Distributed Optical Fiber Monitoring Mine Pressure Prediction Model Based on Convolutional Neural Network[C]//2021 International Conference on Intelligent Transportation, Big Data & Smart City (ICITBS). IEEE, 2021:166-169.

[40]Feng X, Wang Y, Yao J. A neural network model on real-time prediction of roof pressure in coal mines. International Journal of Rock Mechanics and Mining Science, 1996, 33(6), 647-653.

[41]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.

[42]Sak H, Senior A, Beaufays F. Long short-term memory based recurrent neural network architectures for large vocabulary speech recognition. Computer Science, 2014, 338-342.

[43]Wu Y, Yuan M, Dong S, et al. Remaining useful life estimation of engineered systems using vanilla LSTM neural networks. Neurocomputing, 2018, 275, 167-179.

[44]Kong W, Dong Z, Jia Y, et al. Short-term residential load forecasting based on LSTM recurrent neural network. IEEE Transactions on Smart Grid, 2017, 10(1), 841-851.

[45]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].煤炭学报,2020,45(01):54-65.

[46]贾澎涛,苗云风.基于堆叠LSTM的多源矿压预测模型分析[J].矿业研究与开发,2021,41(08):79-82.

[47]刘思峰.灰色系统理论及其应用[M].北京:科学出版社,2010:17-29.

[48]李泽萌.基于LSTM的采煤工作面矿压预测方法研究[D].西安科技大学,2020.

[49]黄林,王电钢,刘萧等. 基于LSTM的网络流量预测方法[J].计算机应用研究, 2020,37(S1):264-265+272.

[50]张知宇,孔祥领,刘传世,李雪芳.绕流柱体问题的量纲归一化模拟方法[J].应用力学学报,2021,38(05):2004-2011.

[51]韩普,张展鹏,张伟.基于多任务学习和多态语义特征的中文疾病名称归一化研究[J].情报学报,2021,40(11):1234-1244.

[52]刘子坤.某矿综放工作面矿压规律及煤柱尺寸优化研究[D].中国矿业大学,2021.

[53]阮吉寿,张华.信息论基础[M]. 机械工业出版社, 2007.11:7-12.