我国矿区水文地质条件复杂，矿井突水一直是威胁煤炭安全开采的重要问题。因此研究有效的矿井突水预警方法对煤炭资源的高效开采有着重要意义。目前基于机器学习的突水预警方法大多用离散的监测指标作为模型特征，不能对连续的水文数据进行趋势分析，无法有效挖掘监测指标数据变化中隐含的突水前兆信息。针对该问题，本文提出了一种基于相似度分析的矿井突水预警方法，依据两相邻工作面回采时监测指标数据曲线的相似性来进行突水预警。主要研究内容如下：

针对传统的曲线相似性度量方法仅能考虑水文数据曲线间单一的距离相似性或趋势相似性无法准确反映突水风险的问题，提出一种基于趋势与距离特征的曲线相似性分析模型(CSAM-TDF)。该模型首先使用基于普氏分析的离散Fréchet距离对两曲线进行趋势相似性度量，然后采用融合了高斯函数的导数动态时间规整算法得出两曲线的距离相似度，最后通过加权得到两曲线间的综合相似度。将CSAM-TDF模型应用在Y矿区进行实验，分析两相邻工作面监测数据曲线的相似性，得到了不同突水风险等级下各监测指标数据曲线的相似度区间划分。最后通过对比灰色关联分析、余弦相似性、相关系数等方法验证了CSAM-TDF在矿井水文数据曲线相似性度量上的适用性和准确性。

在计算各指标数据曲线相似性的基础上，提出一种基于ILSO-PNN的突水风险综合预警模型。首先，采用基于博弈论的组合赋权法求出各监测指标的权重，对各指标相似性进行加权得到多指标综合相似性，并依据综合相似性对突水风险进行划分。然后，采用能实现任意的非线性逼近的概率神经网络(PNN)对突水风险进行预警。为解决PNN预测准确率会受到平滑因子影响的问题，提出改进的狮群优化算法(ILSO)对平滑因子进行寻优。通过引入混沌映射与精英反向学习策略进行种群初始化，并加入Levy飞行随机游走策略帮助算法跳出局部最优。最后，将各类监测指标的相似性及对应的突水风险评价结果输入ILSO-PNN模型中进行训练，完成矿井突水风险预警。实验表明基于ILSO-PNN的矿井突水风险综合预警模型相较于基于PNN和基于PSO-PNN的突水预警模型都具有更高的准确率，能够为煤矿井下突水预测预报提供依据。

最后，结合上述研究内容设计并实现了基于相似度分析的矿井采区突水预警系统。该系统集成了对两相邻工作面各类监测指标曲线进行相似性度量的方法以及突水风险综合预警方法，能够对所选工作面的监测指标数据曲线完成相似性分析，并依据相似性进行突水预警。

Water inrush from mine is always an important problem that threatens the safety of coal mining because of the complicated hydrogeological conditions in mining area. Therefore, the study of effective mine water inrush early warning methods for the efficient exploitation of coal resources is of great significance. At present, most of the early warning methods of water inrush based on machine learning use discrete monitoring indexes as model features, and can not analyze the trend of continuous hydrological data, it can not effectively mine the precursory information of water inrush in the change of monitoring index data. Aiming at this problem, this paper puts forward a mine water inrush early warning method based on similarity analysis, and carries out water inrush early warning according to the similarity of monitoring data curve of two adjacent working faces. The main research contents are as follows:

(1) Aiming at the problem that the traditional curve similarity measure methods can’t reflect the risk of water inrush accurately by only considering the single distance or trend similarity between curves of hydrological data, a curve similarity analysis model(CSAM-TDF) based on trend and distance features is proposed. The model first uses the discrete Fréchet distance based on the Procrustes analysis to measure the trend similarity of the two curves, and then uses the derivative dynamic time warping algorithm with gauss function to obtain the distance similarity of the two curves. Finally, the comprehensive similarity between the two curves is obtained by weighting. The CSAM-TDF model was applied to Y mining area to analyze the similarity of monitoring data curves of two adjacent working faces, and the interval division of the similarity of monitoring data curves under different risk grades of water inrush was obtained. Finally, the applicability and accuracy of CSAM-TDF in similarity measurement of mine hydrologic data curves are verified by comparing grey correlation analysis, cosine similarity and correlation coefficient.

A comprehensive early warning model of water inrush risk based on ILSO-PNN is proposed on the basis of calculating the similarity of data curves of each index. Firstly, the weight of each monitoring index is obtained by the combination weighting method based on the game theory, and the similarity of each index is weighted to obtain the comprehensive similarity of multi-index. Then, the probabilistic neural network(PNN), which can achieve arbitrary nonlinear approximation, is used to predict the risk of water inrush. In order to solve the problem that the prediction accuracy of PNN is affected by the smoothing factor, an improved lion swarm optimization algorithm(ILSO) is proposed to optimize the smoothing factor. Chaos Mapping and elite reverse learning strategy are introduced to initialize the population, and Levy flight random walk strategy is used to help the algorithm jump out of local optimization. Finally, the similarity of various monitoring indicators and the corresponding risk assessment results of water inrush are input into the ILSO-PNN model for training to complete the mine water inrush risk early warning. The experimental results show that the integrated early warning model based on ILSO-PNN has higher accuracy than that based on PNN and PSO-PNN, it can provide the basis for the prediction and forecast of underground water inrush.

Finally, according to the above research contents, the early warning system of water inrush in mining area based on similarity analysis is designed and implemented. The system integrates the method of measuring the similarity of various monitoring index curves of two adjacent working faces and the comprehensive early warning method of water inrush risk, and can finish the similarity analysis of the monitoring index curves of the selected working faces, and carries on the water inrush early warning according to the similarity.

[1] 康红普, 王国法, 王双明, 等. 煤炭行业高质量发展研究 [J]. 中国工程科学, 2021, 23(5): 130-138.

[2] 李振华.煤炭经济发展中煤炭资源整合的重要性研究[J].内蒙古煤炭经济, 2020(18):18-19.

[3] 中矿煤炭产业景气指数研究课题组, 岳福斌. 2020-2021年中国煤炭产业经济形势研究报告 [J]. 中国煤炭, 2021, 47(3): 9-16.

[4] 崔芳鹏, 武强, 刘德民, 等. 煤矿突水灾害综合预防与治理技术 [J]. 煤矿安全, 2015, v.46;No.486(03): 175-177.

[5] 靳德武, 刘英锋, 刘再斌, 等. 煤矿重大突水灾害防治技术研究新进展 [J]. 煤炭科学技术, 2013, 41(1): 5-11.

[6] 李俊杰. 煤矿防治水工作中常见问题及策略 [J]. 矿业装备, 2020, (3): 2-6.

[7] 乔伟, 靳德武, 王皓, 等. 基于云服务的煤矿水害监测大数据智能预警平台构建 [J]. 煤炭学报, 2020, 45(7): 9-14.

[8] Ribicic B M, Hoblaj M K R. Hydrofracturing of Rocks as a Method of Evaluation ofWater, Mud, and Gas Inrush Hazards in Underground Coal Mining [J]. 1991.

[9] 杨志斌. 斯列萨列夫公式在矿井水害防治中的应用分析 [J]. 煤矿安全, 2016, 47(9): 4-7.

[10] J., MotykaA., Pulido-Bosch. Karstic phenomena in calcareous-dolomitic rocks and their influence over the inrushes of water in lead-zinc mines in Olkusz region (South of Poland) [J]. Mine Water&the Environment, 1985, 4(2): 1-11.

[11] Sammarco O. Inrush prevention in an underground mine [J]. International Journal of Mine Water, 1988, 7(4): 43-52.

[12] LawsonHeather, E., TesarikDouglas, et al. Effects of overburden characteristics on dynamic failure in underground coal mining - ScienceDirect [J]. International Journal of Mining Science and Technology, 2017, 27(1): 121-129.

[13] 刘德民. 华北型煤田矿井突水机理及预警技术 [D]; 中国矿业大学(北京), 2015.

[14] 张聪. 基于机器学习的矿井突水安全保障系统及应用 [D]; 北京科技大学, 2019.

[15] 王国际, 杨腾飞, 王公忠,等. 基于"下三带"理论的底板隔水层破坏机理研究 [J]. 中州煤炭, 2011, (11): 4-5.

[16] 周马宁. 中峪井田底板奥灰带压开采突水系数法评价 [J]. 煤炭与化工, 2021, 44(4): 14-19.

[17] 张晓俊. 基于阻水系数法的底板带压开采治理研究 [J]. 煤炭技术, 2021, 40(04): 84-87.

[18] 施龙青. 底板突水机理研究综述 [J]. 山东科技大学学报：自然科学版, 2009, 28(3): 1-12.

[19] 孙建, 王连国, 鲁海峰. 基于隔水关键层理论的倾斜煤层底板突水危险区域分析 [J]. 采矿与安全工程学报, 2017, 34(4):1- 8.

[20] 童柔, 谢天保. 基于机器学习的煤矿突水预测方法 [J]. 计算机系统应用, 2019, 28(12): 243-247.

[21] 师煜, 朱希安, 王占刚, 等. 基于GAPSO-RFR的矿井底板突水预测模型与应用 [J]. 中国矿业, 2020, 29(8): 1-8.

[22] 刘梦杰, 朱希安, 王占刚. 基于FOA-SVR模型的矿井底板突水量预测应用研究 [J]. 中国矿业, 2019, 28(05): 87-91,130.

[23] 尹会永, 周鑫龙, 郎宁, 等. 基于SSA优化的GA-BP神经网络煤层底板突水预测模型与应用 [J]. 煤田地质与勘探, 2021, 49(6): 11-17.

[24] 邓强, 张召千, 王震. 基于门循环单元神经网络模型的煤层底板突水动态预测 [J]. 太原理工大学学报, 2021, 52(5): 7-14.

[25] 余国锋, 袁亮, 任波, 等. 底板突水灾害大数据预测预警平台 [J]. 煤炭学报, 2021, 46(11): 3502-3514.

[26] 孙志伟, 冯海波, 马永军, 等. 基于曲线相似度的供热过程评价方法 [J]. 天津科技大学学报, 2017, 32(05): 68-73.

[27] Mosig A, Clausen M. Approximately matching polygonal curves with respect to the Fréchet distance [J]. Computational Geometry: Theory and Applications, 2005, 30(2): 112-127.

[28] 高兴. 基于特征信息的测井曲线相似度算法研究与应用 [D]; 东北石油大学, 2013.

[29] Wylie T, Zhu B. Intermittent Map Matching with the Discrete Frechet Distance [J]. Eprint Arxiv, 2014.

[30] Zhao J L, Wu Z K, Pan Z K, et al. 3D Face Similarity Measure by Fréchet Distances of Geodesics [J]. 计算机科学技术学报:英文版, 2018, 033(001): 207-222.

[31] 黄世泽, 陈威, 张帆, 等. 基于弗雷歇距离的道岔故障诊断方法 [J]. 同济大学学报:自然科学版, 2018, 46(12): 6-11.

[32] 唐冬来, 倪平波, 张捷, 等. 基于离散弗雷歇距离的户变关系识别方法 [J]. 电力系统自动化, 2021, 45(6): 1-8.

[33] 赵子豪. 测井曲线相似度算法的研究 [D]. 东北石油大学, 2015.

[34] Silva D F, Giusti R, Keogh E, et al. Speeding up similarity search under dynamic time warping by pruning unpromising alignments [J]. Data Mining & Knowledge Discovery, 2018(32):988-1016.

[35] 李正欣, 郭建胜, 王瑛, 等. DTW距离的过滤搜索方法 [J]. 控制与决策, 2018, 33(7): 5-12.

[36] 陶洋, 邓行, 杨飞跃, 等. 基于DTW距离度量的层次聚类算法 [J]. 计算机工程与设计, 2019, 40(1): 6-12.

[37] Deng S, Xiang Y, Nan B, et al. A hybrid model of dynamic time wrapping and hidden Markov model for forecasting and trading in crude oil market [J]. Soft Computing, 2020, 24(3): 6655-6672.

[38] Keogh E J, Pazzani M J. Derivative Dynamic Time Warping [J]. 2002.

[39] 刘苏, 黄纯, 侯帅帅, 等. 基于DDTW距离与DBSCAN算法的户变关系识别方法 [J]. 电力系统自动化, 2021, 45(18): 71-77.

[40] 戴子翔. 狮群优化算法及改进算法的研究与应用实现 [D]; 山东大学, 2020.

[41] Wang B, Jin X P, Cheng B. Lion pride optimizer: An optimization algorithm inspired by lion pride behavior [J]. Science China Information Sciences, 2012, 55(10): 2369-2389.

[42] Rajakumar B R. Lion algorithm for standard and large scale bilinear system identification: A global optimization based on Lion's social behavior; proceedings of the 2014 IEEE Congress on Evolutionary Computation (CEC), July, IEEE, 2014, pp. 2116-2123.

[43] Yazdani M, Jolai F. Lion Optimization Algorithm (LOA): A Nature-Inspired Metaheuristic Algorithm [J]. Journal of Computational Design & Engineering, 2016, 3(1): 24-36.

[44] 刘生建, 杨艳, 周永权. 一种群体智能算法——狮群算法 [J]. 模式识别与人工智能, 2018, 31(5): 11-15.

[45] 郭颖. 基于改进局部搜索机制的狮群算法及其应用 [D]; 山东大学, 2020.

[46] 吴忠强, 谢宗奎, 刘重阳, 等. 基于混沌搜索的改进狮群算法及其在光伏电池参数辨识中的应用 [J]. 计量学报, 2021, 42(4): 415-423.

[47] 张晓通, 韩跃平, 杨洋. 基于遗传-狮群算法优化的独立分量分析算法 [J]. 计算机应用与软件, 2021, 38(2): 244-249.

[48] 汪婵婵. 基于改进狮群算法的汽轮机热耗率模型预测 [J]. 计量学报, 2021, 42(7): 8-12.

[49] 刘德民, 尹尚先, 连会青. 煤矿工作面底板突水灾害预警重点监测区域评价技术 [J]. 煤田地质与勘探, 2019, 47(5): 7-14.

[50] 袁文华, 桂和荣. 任楼煤矿地温特征及在水源判别中的应用 [J]. 安徽理工大学学报:自然科学版, 2005, 25(4): 3-12.

[51] 匡韬, 郑建彬. 一种基于演化计算的在线手写签名验证算法实现 [J]. 计算机应用, 2006, 26(11): 2583-2585.

[52] Hu, Zheng, Zhan, et al. Online Signature Verification Based on a Single Template via Elastic Curve Matching [J]. Sensors, 2019, 19(22): 48-58.

[53] 陈涛. 基于曲线形态的曲线相似性度量 [J]. 现代计算机, 2020, (36): 15-22.

[54] 李妍. 基于博弈论的基础设施PPP模式风险分担研究[D].中国矿业大学,2017.

[55] 胡霞, 钟文杰, 程静静. 基于AHP和熵权法的煤矿安全态势评价模型 [J]. 煤矿安全, 2021, 52(2): 5-8.