煤矿涌水量是煤矿安全监测的关键指标，准确预测煤矿涌水量对保障煤矿安全生产、防排水系统设防以及预防水害事故发生具有重要意义。机器学习方法因其具有强大的非线性处理能力而被应用于煤矿涌水量预测中，论文结合不同地质构造下的煤矿涌水量特性，分别提出了基于MSGWO-SVR的煤矿涌水量预测模型和基于CNN-GCN-BiLSTM的煤矿涌水量预测模型，并通过算例分析，验证了所提模型用于煤矿涌水量预测的可行性。论文主要研究内容如下：
（1）对于地质构造简单的煤矿，煤矿涌水量主要受含水层水压影响，水压与涌水量之间具有强烈的非线性关系，传统方法难以准确表征水压与涌水量之间的潜在变化规律。针对以上问题，提出了一种基于混合策略改进灰狼算法优化支持向量回归的煤矿涌水量预测模型（MSGWO-SVR）。首先，为了增强传统灰狼优化算法种群质量和跳出局部极值的能力，将精英反向学习机制、非线性收敛因子和高斯变异策略引入传统灰狼优化算法，得到MSGWO算法；其次，利用MSGWO对SVR的惩罚因子和核参数进行寻优，以提高SVR模型的自学习能力和非线性拟合能力；最后，根据最优参数组合建立煤矿涌水量预测模型。实验结果显示，与传统预测模型以及SVR和GWO-SVR模型相比，论文所提出的MSGWO-SVR模型预测准确性和稳定性均有明显提高，表明MSGWO-SVR模型可以实现超参数的自动寻优，从而准确预测处于简单地质构造下的煤矿涌水量。
（2）对于地质构造复杂的煤矿，煤矿涌水量受断层落差、煤层倾角等诸多因素影响，各因素之间存在明显的协同作用，浅层模型对数据间的关联特征信息挖掘不够充分，难以实现精准预测。针对以上问题，提出了一种基于图卷积神经网络和双向长短期记忆神经网络的煤矿涌水量预测模型（CNN-GCN-BiLSTM）。首先，为了获得原始数据的有效表征，采用二维卷积（2D-CNN）神经网络层提取数据深层特征；其次，为了深入挖掘输入变量间的潜在信息，构建以涌水量影响因素为节点、皮尔逊相关系数为边的无向图，利用图卷积（GCN）神经网络提取数据空间特征；最后，使用双向长短期记忆（BiLSTM）神经网络对提取的特征信息所构成的序列数据进行双向序列特征学习，并通过全连接层输出涌水量预测值。实验结果显示，与7种经典模型和基线模型相比，论文所提出的CNN-GCN-BiLSTM模型预测准确率最高，表明CNN-GCN-BiLSTM模型更适合于解决复杂地质构造下的煤矿涌水量预测问题。

The quantity lor='red'>of water inflow in coal mine is the key index lor='red'>of Coal Mine Safety Monitoring, and it is lor='red'>of great significance to ensure safe coal mine production, prevent drainage systems from being set up and prevent water damage accidents. This paper combines the characteristics lor='red'>of coal mine water surges under different geological formations and proposes a prediction model based on MSGWO-SVR and a prediction model based on CNN-GCN-BiLSTM for prediction lor='red'>of quantity lor='red'>of water inflow in coal mine, and through case analysis verified the feasibility lor='red'>of the proposed model used for prediction lor='red'>of the quantity lor='red'>of coal mine water inflow. The main research elements lor='red'>of the thesis are as follows:
(1)For coal mines with simple geological formations, the quantity lor='red'>of coal mine water inflow is mainly affected by the water pressure in the aquifer, and there is a strong non-linear relationship between water pressure and water influx, which makes it difficult for traditional methods to accurately characterise the internal change law between water pressure and quantity lor='red'>of mine water influx. In response to above issues, a model for predicting coal mine water inflow based on a hybrid strategy to improve the grey wolf algorithm to optimise support vector regression is proposed (MSGWO-SVR). Firstly, in order to enhance the population quality and the ability to jump out lor='red'>of local extremum lor='red'>of the traditional grey wolf optimisation algorithm, the elite opposition based learning mechanism, nonlinear convergence factor and Gaussian mutation strategy are introduced into the traditional gray wolf optimization algorithm and obtain MSGWO algorithm; secondly, the penalty factor and kernel parameters lor='red'>of the SVR are optimised using MSGWO in order to improve the self-learning ability and non-linear fitting ability lor='red'>of the SVR model; finally, according to the optimal combination lor='red'>of parameters to establisha the prediction model lor='red'>of coal mine water inflow. The experimental results show that compared with traditional forecasting models and the SVR and GWO-SVR models, the MSGWO-SVR model proposed in this paper has significantly improved the prediction accuracy and stability, which indicates that the MSGWO-SVR model can realize the automatic optimization lor='red'>of the hyperparameters, so as to accurately predict the value lor='red'>of coal mine water inflow under simple geological formations.
(2)For coal mines with complex geological formations, the quantity lor='red'>of coal mine water inflow is affected by many factors, such as fault drop, coal seam inclination and so on, there are obvious synergistic effects between various factors, the shallow model is difficult to accurately mine the correlation among data, which leads to large errors lor='red'>of prediction results. In response to above issues, a prediction model lor='red'>of coal mine water inflow based on graphical convolutional neural network and bi-directional long short-term memory neural network is proposed (CNN-GCN-BiLSTM). Firstly, in order to obtain the effective representation lor='red'>of the original data, the two-dimensional convolutional (2D-CNN) neural network is used to extract the deep features lor='red'>of the data; secondly, in order to mine the underlying information between the input variables, an undirected graph was constructed with the influencing factors lor='red'>of coal mine water inflow as nodes and Pearson correlation coefficients as edges, and the spatial features lor='red'>of the data were extracted using the graph convolution (GCN) neural network; finally, bi-directional long short-term memory (BiLSTM) neural network is used to perform bi-directional sequence feature learning on the sequence data composed lor='red'>of the extracted feature information, and output the predicted value lor='red'>of water inflow through the fully connected layers. The experimental results show that the CNN-GCN-BiLSTM model proposed in this paper has the highest prediction accuracy compared to the seven classical models and the baseline model, which indicates that the CNN-GCN-BiLSTM model is more suitable for solving the problem lor='red'>of predicting coal mine water influx under complex geological structures.

[1]汤家轩,刘具,梁跃强,程坤,张锟.“十四五”时期我国煤炭工业发展思考[J].中国煤炭,2021,47(10):6-10.

[2]Chen S L, Xu K, Yu S B, Li B, Cao B, Shen J J.Comparsion Study on Two Methods of Water Inflow Prediction of Coal Mining [J].Advanced Materials Research, 2015,1093: 1379-1382.

[3]胡雅琴,任红蕾.利用数值法和解析法对祁南煤矿矿坑涌水量进行预测分析[J].地下水,2021,43(5):9-11.

[4]戴岩柯,崔世新,张坤.水均衡法和数值模拟法在矿坑深部涌水量预测中的比较——以西石门铁矿为例[J].地下水,2010,32(1):24-26.

[5]李孝朋,谢道雷,徐万鹏,张延飞,张宪峰,吴霞.多元回归分析在矿井涌水量预测中的应用[J].煤炭技术,2016,35(10):189-190.

[6]段俭君,徐会军,王子河.相关分析法在矿井涌水量预测中的应用[J].煤炭科学技术,2013,41(6):114-116.

[7]王家乐,王施智,黄倩.基于相关分析法的煤矿涌水量预测[J].陕西煤炭,2021,40(2):99-102.

[8]姚辉,张润畦,蒋知廷,刘鑫.矿井涌水量回归模型预测方法研究[J].华北科技学院学报,2021,18(3):24-31.

[9]Deng Y,Cao C,Chen S.Research on correlation analysis and prediction model of agricultural climate factors based on machine learning [J].MATEC Web of Conferences,2021,336(7): 1-7.

[10]刘国辉,侯征,王天意,宋洪伟.一种新型的地下含水层涌水量预测模型[J]. 地球物理学进展, 2012, 27(3):1268-1273.

[11]凌成鹏,孙亚军,杨兰和,姜素,邵飞燕.基于BP神经网络的孔隙充水矿井涌水量预测[J].水文地质工程地质, 2007,34(5):55-58.

[12]张宪峰,魏久传,张延飞,吴霞,李孝朋.基于主成分分析与BP神经网络的矿井涌水量预测研究[J].煤炭技术, 2018, 37(6):201-203.

[13]孟璐.基于人工神经网络的矿井涌水量预测[J].煤炭与化工,2020,43(4):65-68.

[14]窦贤明,杨永国,徐伟伟,靖凤伟.基于遗传算法和BP神经网络的矿井涌水量预测[J]. 中国煤炭地质, 2009, 21(10):37-38.

[15]Shuaihan T. An Improved genetic algorithm for solving TSP[J]. Journal of Physics: Conference Series,2021,1952(4): 1-5.

[16]潘正君,康立山.演化算法[M].北京:北京清华大学出版社,1998.

[17]Carlo M D, Vasile M, Minisci E.Adaptive multi-population inflationary differential evolution [J]. Soft Computing, 2020,24(5): 3861-3891.

[18]Simon D.Biogeography-based optimization[J].IEEE Trans on Evolutionary Computation, 2008,12(6): 702-713.

[19]Rashedi E, Nezamabadi-Pour H, Saryazdi S. GSA: a gravitational search algorithm[J]. Information sciences,2009,179(13): 2232-2248.

[20]Kaveh A, Dadras A. A novel meta-heuristic optimization algorithm: Thermal exchange optimization[J].Advances in Engineering Software, 2017,110: 69-84.

[21]Mirjalili S. SCA: a sine cosine algorithm for solving optimization problems[J]. Knowledge-based systems, 2016,96:120-133.

[22]Qais M H, Hasanien H M, Alghuwainem S. Transient search optimization: A new meta-heuristic optimization algorithm[J].Applied Intelligence, 2020, 50(11): 3926-3941.

[23]赵乃刚,邓景顺.粒子群优化算法综述[J].科技创新导报,2015,12(26):216-217.

[24]李晓磊.一种新型的智能优化方法-人工鱼群算法[D],浙江:浙江大学,2003.

[25]Dorigo M, V Maniezzo, Colorni A. Ant sytem: Optimization by a colony of cooperating agents [J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2002, 26(1):29-41.

[26]Karaboga D, Basturk B. A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm[J]. Journal of global optimization, 2007, 39(3): 459-471.

[27]Guerrero-Luis M, Valdez F, Castillo O. A Review on the Cuckoo Search Algorithm[J]. 2021, 940: 113-124.

[28]Mirjalili, Seyedali, Lewis, Andrew.The Whale Optimization Algorithm[J].Advances in engineering software, 2016,95: 51-67.

[29]Xing B, Gao W J. Fruit Fly Optimization Algorithm [J].Springer International Publishing,2014,62: 167-170.

[30]李雅丽,王淑琴,陈倩茹,王小钢.若干新型群智能优化算法的对比研究[J].计算机工程与应用,2020,56(22):1-12.

[31]Sm A,Smm B,Al A.Grey Wolf Optimizer[J]. Advances in Engineering Software, 2014, 69(7): 46–61.

[32]Tikhamarine Y, Souag-Gamane D, Kisi O.A new intelligent method for monthly streamflow prediction: hybrid wavelet support vector regression based on grey wolf optimizer (WSVR–GWO)[J]. Arabian Journal of Geosciences,2019,12(17): 1-20.

[33]Nayak B, Mohapatra A, Mohanty K B. Parameter estimation of single diode PV module based on GWO algorithm[J]. Renewable Energy Focus, 2019, 30: 1-12.

[34]Qin W,Wang L H,Liu Y H,Xu C.Energy Consumption Estimation of the Electric Bus Based on Grey Wolf Optimization Algorithm and Support Vector Machine Regression[J]. Sustainability,2021,13(9) :4689-4689.

[35]黄波,杨正,王超.基于灰狼算法优化DBN的医院网络异常流量识别[J].微型电脑应用,2022,38(1):34-36.

[36]Chandar S K.Grey Wolf optimization-Elman neural network model for stock price prediction[J].Soft Computing,2021,25(1): 649–658.

[37]陈长倩,慕晓冬,牛犇,王立志.结合高斯分布的改进二进制灰狼优化算法[J].计算机工程与应用,2019,55(13):145-150.

[38]徐辰华,李成县,喻昕,黄清宝.基于Cat混沌与高斯变异的改进灰狼优化算法[J].计算机工程与应用,2017,53(4):1-9.

[39]Kohli M,Arora S.Chaotic grey wolf optimization algorithm for constrained optimization problems-ScienceDirect[J].Journal of Computational Design and Engineering,2018,5(4): 458-472.

[40]谈发明,赵俊杰,王琪.一种改进非线性收敛方式的灰狼优化算法研究[J].微电子学与计算机,2019,36(5):89-95.

[41]李阳,李维刚,赵云涛,刘翱.基于莱维飞行和随机游动策略的灰狼算法[J].计算机科学,2020,47(8):291-296.

[42]徐辰华,骆珠光,吴冠宏,刘斌.基于正弦因子和量子局部搜索的灰狼优化算法[J].计算机工程与应用,2021,57(24):83-89.

[43]龙文,蔡绍洪,焦建军,伍铁斌.一种改进的灰狼优化算法[J].电子学报,2019,47(1):169-175.

[44]莫艳红,聂慧,刘振丙,杨辉华.基于莱维飞行的灰狼优化算法[J].微电子学与计算机,2019,36(4):78-83.

[45]黄晨晨,魏霞,黄德启,叶家豪.求解高维复杂函数的混合蛙跳–灰狼优化算法[J].控制理论与应用,2020,37(7):1655-1666.

[46]Jia S Y,Sun S B,Wang J, Xu F. Research Based on The Improved Grey Wolf Optimization Algorithm[J]. Journal of Physics: Conference Series, 2021,1883(1):1-6.

[47]Gao Z M, Zhao J.An Improved Grey Wolf Optimization Algorithm with Variable Weights[J]. Computational Intelligence and Neuroscience,2019,2019:1-13.

[48]Scarselli F,Gori M,Tsoi A C, Hagenbuchner M, Monfardini G.The Graph Neural Network Model[J]. IEEE Transactions on Neural Networks, 2009,20(1):61-80.

[49]Bruna J, Zaremba W, Szlam A, et al. Spectral Networks and Locally Connected Networks on Graphs[J]. Computer Science, 2013.

[50]Gonzalez G,Gong S,Laponogov I,Bronstein M,Veselkov K.Predicting anticancer hyperfoods with graph convolutional networks[J]. Human Genomics, 2021, 15(1):33-33.

[51]刘臣,李秋,郝宇辰.基于图卷积神经网络的在线社区行为预测[J].计算机技术与发展,2021,31(4):28-33.

[52]Yu B, Lee Y, Sohn K.Forecasting road traffic speeds by considering area-wide spatio-temporal dependencies based on a graph convolutional neural network (GCN)[J].Transportation Research Part C,2020,114(C): 189-204.

[53]张彦青.“下三带”理论对承压水上采煤底板突水的预测研究[J].山西煤炭,2014,34(4):50-52.

[54]张金才,刘天泉,张玉卓.裂隙岩体渗透特征的研究[J].煤炭学报,1997(5):35-39.

[55]于辉.基于薄板理论的首采工作面基本顶来压步距数值研究[J].科学技术与工程,2018,18(21):195-199.

[56]刘鸿泉,王作宇.采场底板承压水运动[J].河北煤炭,1992(4):196-201.

[57]王作宇,刘鸿泉,王培彝,于树春.承压水上采煤学科理论与实践[J].煤炭学报,1994(1):40-48.

[58]王成绪.底板突水的数值计算方法研究[J].煤田地质与勘探,1997(S1):45-47.

[59]钱鸣高,缪协兴.岩层控制中的关键层理论研究[J].煤炭学报, 1996, 21(3):2-7.

[60]赵伟冬.含水层水压对底板断层突水危险性的影响[J].内蒙古煤炭经济, 2018(4):138-144.

[61]严由吉.深埋煤层采动覆岩导水裂隙带发育高度研究[D].西安:西安科技大学,2021.

[62]孙少华,苏礼,苏志毅.煤层倾角对底板破坏带深度的影响[J].煤,2017,26(8):29-32.

[63]刘亮东. 综采工作面覆岩断层采动活化致灾机理研究[D].西安:西安科技大学,2020.

[64]刘方园,王水花,张煜东.支持向量机模型与应用综述[J].计算机系统应用,2018,27(4):1-9.

[65]Choi R Y , Coyner A S , Kalpathy-Cramer J ,Chiang M F，Campbell J P. Introduction to Machine Learning, Neural Networks, and Deep Learning[J]. Translational Vision Science & Technology, 2020, 9(2): 14-14.

[66]Zhao L, Song Y, Zhang C, Liu Y,Li H. T-gcn: A temporal graph convolutional network for traffic prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21(9): 3848-3858.

[67]Al-Smadi M,Qawasmeh O,Al-Ayyoub M,Jararweh Y,Gupta B.Deep Recurrent Neural Network vs. Support Vector Machine for Aspect-Based Sentiment Analysis of Arabic Hotels' Reviews[J]. Journal of Computational ence, 2018, 27: 386-393.

[68]Hochreiter S, Schmidhuber J.Long Short-Term Memory[J]. Neural Computation,1997, 9(8): 1735-1780.

[69]Wang N, Zhao L, Wang L, Hei X H,Zhao L. A review of opposition-based learning from 2005 to 2012[J]. Engineering Applications of Artificial Intelligence: The International Journal of Intelligent Real-Time Automation, 2014, 29(3): 1-12.

[70]李彦苍,吴悦.基于高斯变异的蚁狮算法及其在组合优化中的应用[J].中国科技论文,2022,17(3):295-304.

[71]唐思豪,滕召胜,孙彪,胡清,潘喜福. ADAM改进BP神经网络与动态称重应用[J]. 电子测量与仪器学报, 2021,35(4):127-135.