本文针对陕北榆神府矿区的近浅埋煤层的综采工作面，在随工作面推进过程中可能出现的顶板不规律来压造成的安全隐患，通过矿压理论分析、工作面数据采集、数据深度挖掘、建立初步预测模型及优化的模型算法，来实现榆神府矿区近浅埋煤层综采工作面的顶板来压规律预测。其主要研究成果包括：

在榆神府矿区采集100组综采工作面的地质条件、开采工艺以及基本顶来压的矿压数据作为预测模型的训练、测试、验证样本数据集，并采用灰色系统理论提出的关联度分析各输入因素对顶板来压的影响程度。灰度理论分析结果表明：工作面长度、推进速度、采高、基本顶条件，直接顶厚度，埋深，煤层倾角为显著因素，煤层厚度为次要因素；并利用主成分分析理论进一步在显著因素中提取了5个主成分因子，作为预测模型的输入特征数据。

在主成分分析中的100组测试集样本数据中，随机选取90组数据作为训练样本，10组数据作为测试集样本，建立工作面顶板来压的标准支持向量机预测模型。其中，选择主成分分析中的五组主成分因子作为其输入特征，初次来压步距、初次来压强度、周期来压步距、周期来压强度作为输入输出特征。标准支持向量机预测结果得出，四个输出结果的均方误差分别为15.6%，11.7%，8.7%，13.2%。结果表明：标准支持向量机预测模型对于综采工作面顶板来压规律具有一定的参考价值。

为进一步优化预测模型，基于标准支持向量机在建立模型过程中出现的超参选择困难问题，建立基于粒子群群智优化的支持向量机预测模型。对比标准支持向量机预测模型，初次来压步距、初次来压强度、周期来压步距、周期来压强度的均方误差分别提高12%，10%，5.6%，12.1%。结果表明：基于粒子群优化的支持向量机顶板来压预测模型无论在精准度还是拟合优度都明显优于标准支持向量机预测模型。

最后，我们通过察哈素煤矿31201采煤工作面收集到的数据验证粒子群群智优化支持向量机预测模型为脚本的人工交互系统。结果表明：初次来压步距、初次来压强度、周期来压步距、周期来压强度的均方误差分别较真实理论数据，误差分别为：5.0%，6.1%，4.4%，1.1%。均小于10%。因此，基于粒子群优化的支持向量机具有较强的鲁棒性，能够较精准预测类似地质条件、开采工艺的工作面顶板来压规律，消除顶板来压带来的安全隐患。

In this paper, aiming at the fully mechanized mining face of the near-shallow buried coal seam in the Yushenfu mining area in northern Shaanxi, the potential safety hazard caused by the irregular roof pressure in the process of advancing with the working face, through the theoretical analysis of mine pressure, working face Data collection, in-depth data mining, establishment of preliminary prediction model and optimization model algorithm are used to realize the prediction of the roof pressure law of the fully mechanized mining face of the near-shallow buried coal seam in the Yushenfu mining area. Its main research results include:

In the Yushenfu mining area, 100 groups of fully mechanized mining face geological conditions, mining technology and basic mining pressure data were collected as training and testing samples for the prediction model, and the correlation degree proposed by the grey system theory was used to analyze the impact of each input factor on the The degree of influence of the pressure from the top plate. The grayscale theoretical analysis results show that: working face length, working face advancing speed, mining height, basic roof conditions, direct roof thickness, burial depth, coal seam dip angle are significant factors, and coal seam thickness is a secondary factor; and the principal component analysis theory is used to further Five principal component factors were extracted from the significant factors as the number of features for the prediction model.

Among the 80 sets of test set sample data in the principal component analysis, 70 sets of data were randomly selected as training samples and 10 sets of data as test set samples, and a support vector machine prediction model for the pressure of the working surface roof was established. Among them, the seven principal component factors in the principal component analysis are selected as input data, and the initial pressure step, the initial pressure intensity, the periodic pressure step, and the periodic pressure intensity are the output targets. The standard support to the machine prediction results shows that the mean square errors of the four output results are 15.6%, 11.7%， 8.7%， and 13.2%, respectively. The results show that the standard support vector machine prediction model has a certain reference value for the roof pressure law of fully mechanized mining face.

In order to further optimize the prediction model, a support vector machine prediction model based on particle swarm optimization was established based on the hyperparameter selection mining accident problem that occurred during the model building process of the standard support vector machine. Compared with the standard support vector machine prediction model, the mean square errors of the initial pressure step, the initial pressure intensity, the periodic pressure step distance, and the periodic pressure intensity increased by 12%, 10%, 5.6%, and 12.1%, respectively. The results show that the prediction model based on particle swarm optimization is obviously better than the standard support vector machine in terms of accuracy and goodness of fit.

Finally, we verify the particle swarm optimization support vector machine model as a scripted manual interactive system through the data collected from the 31201 coal face of the Chahasu coal mine. The results show that the mean square errors of the initial pressure step distance, the initial pressure intensity, the periodic pressure step distance, and the periodic pressure intensity are compared with the real theoretical data, and the errors are: 5.0%, 6.1%, 4.4%, and 1.1%, are less than 10%. Therefore, the support vector machine based on particle swarm optimization has strong robustness, and can be more precise and similar to the geological conditions and mining process of the working face roof pressure law, and eliminate the safety hazards caused by the roof pressure.

[1] Su Wen H, Peng Syd, Hsiung,SM, Inieraetions in multiple-Seam mining, Soc of Mining Engineers, 1986, 31-44.

[2] Peng SS, Chiang HS. Longwall mining [M]. New York John Wiley sons, 1984.

[3] Bodrac B B. Rock pressure features of Moscow Suburb coal-field [J]. Coal,1998(2),38-49.

[4] B.霍勃尔瓦依特.浅部长壁法开采效果的地质技术评价煤炭科研参考资料，1985,3.

[5] HOLLA L and BUIZEN M. The ground movement, strata fracturing and changes in permeability due to deep longwall mining [J]. International Journal of rock mechanics, 1991, 28(2-3): 207-217.

[6] HOLLA L. Some aspects of strata movement related to mining under water bodies in New South Wales, Australia. Proceedings of the Fourth I.M.W.A. Congress. Lubljana, Australia, 1991[C]. 1991.

[7] 钱鸣高,缪协兴.采场上覆岩层结构的形态与受力分析[J].岩石力学与工程学报, 1995 (02):97-106.

[8] 缪协兴,钱鸣高.采场围岩整体结构与砌体梁力学模型[J].矿山压力与顶板管理,1995 (Z1):3-12+197.

[9] 宋振骐.实用矿山压力控制[M].徐州：中国矿业大学出版社，1988.

[10] 张顶立,钱鸣高,翟明华,等.综放工作面覆岩结构型式及矿压显现[J].矿山压力与顶板管理,1994(04):13-17+80.

[11] 周金龙,黄庆享.浅埋大采高工作面顶板关键层结构稳定性分析[J].岩石力学与工程学报,2019,38(07):1396-1407.

[12] 宋选民,顾铁凤,闫志海.浅埋煤层大采高工作面长度增加对矿压显现的影响规律研究[J].岩石力学与工程学报,2007(S2):4007-4012.

[13] 侯忠杰,黄庆享.松散层下浅埋薄基岩煤层开采的模拟[J].陕西煤炭技术,1994, (02):38-41+65.

[14] 侯忠杰,张杰.厚松散层浅埋煤层覆岩破断判据及跨距计算[J].辽宁工程技术大学学报,2004(05):577-580.

[15] 张杰,陈诚,张建辰,等.浅埋厚层间距煤层过上覆采空区煤柱矿压显现机理研究[J].煤矿安全,2020,51(12):63-68.

[16] 许家林,朱卫兵,王晓振.浅埋煤层覆岩关键层结构分类[J].煤炭学报,2009,34(7):865- 870

[17] 朱卫兵,许家林,施喜书.覆岩主关键层运动对地表沉陷影响的钻孔原位测试研究[J].岩石力学与工程学报,2009,28(2):403-409.

[18] 屠世浩,窦凤金,万志军.浅埋房柱式采空区下近距离煤层综采顶板控制技术[J].煤炭学报,2011,36(3):366-370.

[19] 刘洪磊,杨天鸿,张博华,等.西部煤层开采覆岩垮落及矿压显现影响因素研究[J].煤炭学报,2017,42(02):460-469.

[20] 于斌,杨敬轩,刘长友,等.大空间采场覆岩结构特征及其矿压作用机理[J].煤炭学报,2019,44(11):3295-3307.

[21] 王国法,张德生.煤炭智能化综采技术创新实践与发展展望[J].中国矿业大学学报, 2018,47(03):459-467.

[22] Tingjiang Tan,Zhen Yang,Feng Chang,Ke Zhao. 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).

[23] 冯夏庭,王泳嘉,姚建国.煤矿顶板矿压显现实时预报的自适应神经网络方法[J].煤炭学报,1995(05):455-460.

[24] 吴洪词.基于神经网络的采场底板分类与顶板来压预报[J].贵州工学院学报,1996 (04):32-36.

[25] 张丽华,蔡美峰.顶板来压识别与预测的复合小波神经网络方法[J].煤炭科学技术, 2003(07):41-43.

[26] 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.

[27] 程敬义,万志军,PENG Syd S,等.基于海量矿压监测数据的采场支架与顶板状态智能感知技术[J].煤炭学报,2020,45(06):2090-2103.

[28] 左凌云.基于支架工作阻力大数据的工作面区域矿压分析研究[J].煤炭工程,2019, 51(11):60-64.

[29] 巩师鑫，任怀伟，杜毅博，等.基于MRDA-FLPEM集成算法的综采工作面矿压迁移预测[J/OL].煤炭学报：1-11[2021-01-18].

[30] 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.

[31] 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.

[32] 赵铭生,刘守强,纪润清,等.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.

[33] 窦林名,何学秋,REN Ting,等.动静载叠加诱发煤岩瓦斯动力灾害原理及防治技术[J].中国矿业大学学报,2018,47(01):48-59.

[34] 付玉平,宋选民,邢平伟.浅埋煤层大采高超长工作面垮落带高度的研究[J].采矿与安全工程学报,2010,27(02):190-194.

[35] 钱鸣高,何富连,李全生,刘双跃.综采工作面端面顶板控制[J].煤炭科学技术,1992(01):41-46+59.

[36] 钱鸣高,殷建生,刘双跃.综采工作面直接顶的端面冒落[J].煤炭学报,1990(01):1-9.

[37] 黄庆享,黄克军,赵萌烨.浅埋煤层群大采高采场初次来压顶板结构及支架载荷研究[J].采矿与安全工程学报,2018,35(05):940-944.

[38] 鞠金峰,许家林,朱卫兵.浅埋特大采高综采工作面关键层“悬臂梁”结构运动对端面漏冒的影响[J].煤炭学报,2014,39(07):1197-1204.

[39] 许家林,鞠金峰.特大采高综采面关键层结构形态及其对矿压显现的影响[J].岩石力学与工程学报,2011,30(08):1547-1556.

[40] 张杰,何义峰,王旭,冯冲,杨涛,康小杰,李宏儒.浅埋近距离煤层群重复采动覆岩破坏规律分析研究[J].矿业研究与开发,2022,42(02):60-64.

[41] 侯忠杰,吴文湘,肖民.厚土层薄基岩浅埋煤层“支架-围岩”关系实验研究[J].湖南科技大学学报(自然科学版),2007(01):9-12.

[42] 赵毅鑫,刘文超,张村,等.近距离煤层蹬空开采围岩应力及裂隙演化规律[J].煤炭学报,2022,47(01):259-273.

[43] 赵毅鑫,刘文超,张村,等.近距离煤层蹬空开采围岩应力及裂隙演化规律[J].煤炭学报,2022,47(01):259-273.

[44] 王家臣.基于采动岩层控制的煤炭科学开采[J].采矿与岩层控制工程学报,2019,1(02):40-47.

[45] 张杰,王斌,白文勇,等.浅埋近距间隔式采空区顶板“双拱桥”结构稳定性研究[J].中国矿业大学学报,2021,50(03):598-605.

[46] 常峰.基于GA-BP神经网络的工作面顶板矿压预测模型应用研究[D].中国矿业大学,2019.

[47] 钱鸣高.采场围岩控制理论与实践[J].矿山压力与顶板管理,1999(Z1):12-15.

[48] 谭学瑞,邓聚龙.灰色关联分析:多因素统计分析新方法[J].统计研究,1995(03):46-48.

[49] 邓聚龙,孙顺庚.灰色系统理论与科技进步[J].科技进步与对策,1986(05):40-41

[50] Gang Pan,Pengcheng Li,Lianjun Chen, et al. A study of the effect of rheological properties of fresh concrete on shotcrete-rebound based on different additive components[J]. Construction and Building Materials,2019,224.

[51] Moawad Gaby,Tyan Paul,Marfori , et al. Effect of Postoperative Partial Bladder Filling After Minimally Invasive Hysterectomy on Post Anesthesia Care Unit Discharge and Cost: A Single-Blinded Randomized Controlled Trial.[J]. American journal of obstetrics and gynecology,2019 72: 73-80.

[52] Yang J, Zhang D, Frangi A F, et al. Two-dimensional PCA: a new approach to appearance-based face representation and recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2004, 26(1): 131-137.

[53] Granato D, Santos J S, Escher G B, et al. Use of principal component analysis (PCA) and hierarchical cluster analysis (HCA) for multivariate association between bioactive compounds and functional properties in foods: A critical perspective[J]. Trends in Food Science & Technology, 2018, 72: 83-90.

[54] 吴明圣.径向基神经网络和支持向量机的参数优化方法研究及应用[D].中南大学,2007.

[55] Cherkassky V, Ma Y. Practical selection of SVM parameters and noise estimation for SVM regression[J]. Neural networks, 2004, 17(1): 113-126.

[56] 李宝胜,秦传东.基于粒子群优化的SVM多分类的电动车价格预测研究[J].计算机科学,2020,47(S2):421-424.

[57] Marini F, Walczak B. Particle swarm optimization (PSO). A tutorial[J]. Chemometrics and Intelligent Laboratory Systems, 2015, 149: 153-165.

[58] Cement Research; Studies in the Area of Cement Research Reported from SouthwestPetroleum University (Characterising the hydration process of cement slurry system based on low-field NMR)[J]. Energy Weekly News,2020.

[59] 何瑞江.基于GRU-SVM神经网络的大数据入侵检测方法研究[J].微型电脑应用, 2022,38(02):127-129.

[60] 郭继坤,赵清,徐峰.基于SVM的煤矿井下超宽带穿透成像算法研究[J].煤炭学报, 2018,43(02):584-590.

[61] 王家臣,李良晖,杨胜利.不同照度下煤矸图像灰度及纹理特征提取的实验研究[J].煤炭学报,2018,43(11):3051-3061.

[62] R. Price, BP Gaber, Y. Lvov R. In-vitro release characteristics of tetracycline HCl, khellin and nicotinamide adenine dineculeotide from halloysite; a cylindrical mineral[J]. Journal of microencapsulation, 2001, 18(6): 713-722.

[63] Stuiver M, Reimer P J, Bard E, et al. INTCAL98 radiocarbon age calibration, 24,000–0 cal BP[J]. radiocarbon, 1998, 40(3): 1041-1083.

[64] 陈帅,黄腾,高大龙.改进BP神经网络模型在索塔变形预测中的应用[J].地理空间信息,2022,20(02):27-32.

[65] 林涛,王松,刘英舜,等.基于最小二乘和BP神经网络算法的转辙机测力方法探究[J].测控技术,2022,41(01):46-50.