瓦斯事故长期以来制约着我国煤矿安全生产和煤炭行业的可持续发展，其引发的主要原因是甲烷浓度超限，给矿井下生产带来严重的事故隐患。综采工作面作为煤矿开采的第一现场，是瓦斯事故频发的区域。现阶段，煤矿企业对甲烷浓度的防治主要通过各传感器监测甲烷浓度数据来实现，预警能力尚待提升。因此本文对甲烷浓度变化趋势进行研究，实现甲烷浓度的准确预测并对超限甲烷浓度进行分级预警，对煤炭企业的发展和保障井下人员生命安全具有重要的现实意义。
  本文以综采工作面甲烷浓度监测数据为研究对象，首先通过分析工作面甲烷浓度的影响因素，采用灰色关联法计算各因素权重，确定影响甲烷浓度的 6 种主要因素。并通过改进小波阈值方法去除传感器采集过程中存在的噪声干扰，采用主成分分析法(Principal Component Analysis, PCA)将门控循环单元(Gate Recurrent Unit, GRU)模型 6 维输入参数降维到 4 维主成分参数，减小网络模型的数据冗余。其次甲烷浓度预测模型以GRU 模型为基础，为了解决 GRU 网络容易陷入梯度问题，采用改进粒子群优化算法(IPSO)对 GRU 网络隐含层参数进行寻优并对甲烷浓度进行预测，通过对比实验得出IPSO-GRU 甲烷浓度预测模型的均方根误差为 0.017，平均绝对误差为 0.019，可以看出该模型预测准确度高且拟合效果好。最后在.NET 框架上实现甲烷浓度预警系统的设计和开发，将监测数据与 IPSO-GRU 模型的预测数据通过灰色关联度分析法对预警等级和预警区间进行划分，预警系统由客户端表现层、技术封装应用层、数据库集成层三层架构组成，实现了对综采工作面甲烷浓度的预测、分级预警以及数据曲线可视化显示功能。
  本文对综采工作面甲烷浓度数据预警方法的研究，可以实现对综采工作面甲烷浓度的预测以及超限甲烷浓度的分级预警，能够保障矿井人员的安全，为煤矿安全生产提供了支持和决策，具有一定的理论研究价值和工程应用价值。

    Gas accidents have long restricted the safe production of coal mines and the sustainable development of the coal industry in my country. The main reason for their triggering is the over-limit of methane concentration, which brings serious accident hazards to the underground production of mines. As the first site of coal mining, the general mining working face is the area where gas accidents occur frequently. At this stage, the prevention and control of methane concentration in coal mines is mainly achieved by monitoring methane concentration data through various sensors, and the early warning capability needs to be improved. Therefore, it is of great practical significance for the development of coal enterprises and safety of underground personnel to study the change trend of methane concentration, to achieve accurate prediction of methane concentration and carry out graded early warning of over-limit methane concentration.

    This thesis takes themethane concentration monitoring data of the comprehensive mining working face as the research object. Firstly, by analyzing the influencing factors of the methane concentration of the working face, the grey correlation method is used to calculate the weights of each factor and determine the six main factors affecting the methane concentration. And by improving the wavelet threshold method to remove the noise interference in the sensor acquisition process, and using principal component analysis (PCA) to downscale the 6-dimensional input parameters of the gated recirculation unit (GRU) model to 4-dimensional principal component parameters to reduce the data redundancy of the network model. Secondly, the methane concentration prediction model is based on the GRU model. In order to solve the problem that the GRU network is prone to fall into the gradient, the improved particle swarm optimization algorithm (IPSO) is used to find the optimal values of the implied layer weights of the GRU network and to predict the methane concentration. Experiments show that the root mean square error of the IPSO-GRU methane concentration prediction model is 0.017, and the mean absolute error is 0.019, which shows that the model predicts with high accuracy and good fit. Finally, the methane concentration early warning system is design and developed on the .NET framework. The monitoring data and the prediction data of the IPSO-GRU model are divided into warning levels and warning intervals by the gray correlation analysis. The system is composed of three layers: client presentation layer, technical packaging application layer and database integration layer. which realizes the functions of predicting the methane concentration of the fully mechanized mining face, graded early warning and visual display of the data curve.

    In this thesis, the research on the early warning method of methane concentration data in the mechanized mining face, it can realize the prediction of methane concentration in the mechanized mining face and the grading early warning function of over-limit methane concentration, which can guaranteee the safety of mine personnel, and provide support and decision for coal mine safety production, and has certain theoretical research value and engineering application value.

[1] 袁亮. 我国煤矿安全发展战略研究[J]. 中国煤炭. 2021, 47(06): 1-6.

[2] 齐翊希. 煤炭经济发展态势与对策分析[J]. 现代工业经济和信息化, 2019, 9(7): 7-8.

[3] 煤炭是保障国民经济持续较快发展的重要能源支撑 [J]. 中国煤炭, 2019, 45(10): 5-9.

[4] 王国栋, 杨秀铁. 近年来煤矿瓦斯爆炸事故技术原因及应对措施研究[J]. 煤矿安全, 2018, 49(1): 230-232+236.

[5] 黄继广, 马汉鹏, 范春姣, 等. 我国煤矿安全事故统计分析及预测[J]. 陕西煤炭, 2020, 39(03): 34-39+6.

[6] 于海云, 张淑同.近年我国煤矿瓦斯事故发生规律及防治对策探讨[J]. 矿业安全与环保, 2016(1): 108-110+114.

[7] 张镭. 煤矿瓦斯防治技术现状与问题[J]. 中国科技信息, 2020(17): 105-106.

[8] 孙继平. 煤矿信息化自动化新技术与发展[J]. 煤炭科学, 2016, 44(1): 19-23.

[9] 白龙西, 李德仁, 文浩, 等. 煤矿瓦斯防治技术现状与问题[J]. 陕西煤炭, 2021, 40(3): 188-190.

[10]黄强辉. 煤矿安全监控系统的现状与发展趋势分析[J]. 建筑工程技术与设计, 2018(33): 47-53.

[11]郑学召, 童鑫, 郭军,等. 煤矿智能监测与预警技术研究现状与发展趋势[J]. 工矿自动化, 2020, 46(06): 35-40.

[12]胡森荣, 朱宗玖. 基于LoRa与HMM的煤矿安全监测系统设计[J]. 煤矿机械. 2020, 41(07): 7-9.

[13]郭越, 李舸帆. 煤矿安全监测系统研究与开发[J]. 煤炭技术, 2013, 32(7): 84-85.

[14]孙皓月, 赵翊君, 李薇. 基于物联网的煤矿安全监测系统[J]. 煤炭技术, 2013(12): 95-96.

[15]孙霞, 陆玉正. 基于DSP和CPLD的煤矿安全监测系统[J]. 煤矿安全, 2012, 43(7): 89-91.

[16]Boddu R, Balanagu P, et al. Zigbee Based Mine Safety Monitoring System with GSM[J]. International Journal of Computer & Communication Technology, 2012, 3(5): 63-67.

[17]Muduli L, Mishra D P, et al. Application of Wireless Sensor Network for Environmental Monitoring in Underground Coal Mines: A Systematic Review[J]. Journal of Network and Computer Applications, 2018, 106: 48-67.

[18]Wei S, Li L. Multi-Parameter Monitoring System for Coal Mine Based on Wireless Sensor Network Technology[C]. 2009 International Conference on Industrial Mechatronics and Automation. IEEE, 2009: 225-227.

[19]张震, 朱权洁, 李青松, 等. 基于Keras长短时记忆网络的矿井瓦斯浓度预测研究[J]. 安全与环境工程, 2021, 28(1): 61-67+78.

[20]马莉, 潘少波, 代新冠, 等. 基于PSO-Adam-GRU的煤矿瓦斯浓度预测模型[J]. 西安科技大学学报, 2020, 40(2): 363-368.

[21]李树刚, 马莉, 潘少波, 等. 基于循环神经网络的煤矿工作面瓦斯浓度预测模型研究[J]. 煤炭科学技术, 2020, 48(1): 33-38.

[22]耿越. 基于混沌粒子群神经网络的瓦斯浓度预测[J]. 中国煤炭, 2017, 43(3): 124-129.

[23]李欢, 贾佳, 杨秀宇, 等. 煤矿综采工作面瓦斯浓度预测模型[J]. 工矿自动化, 2018, 44(12): 48-53.

[24]王雨虹, 王淑月, 王志中, 等. 基于改进蝗虫算法优化长短时记忆神经网络的多参数瓦斯浓度预测模型研究[J]. 传感技术学报, 2021, 34(9): 1196-1203.

[25]李伟山, 王琳,卫晨. LSTM在煤矿瓦斯预测预警系统中的应用与设计[J]. 西安科技大学学报, 2018, 38(6): 1027-1035.

[26]白永明, 张杰. 矿井瓦斯预测预警的云计算模式研究[J]. 工矿自动化, 2014, 40(5): 76-80.

[27]李栋, 孙振明, 李梅, 等.基于混沌粒子群的AWLSSVM瓦斯预测研究[J].煤矿安全, 2020, 51(08): 193-198.

[28]赵亮. 基于混沌时间序列的煤矿工作面瓦斯浓度 预测研究[J]. 煤矿机械, 2017, 38(7): 178-179.

[29]许满贵, 高帅帅, 曹艳军, 等. 基于灰色理论-多元回归分析的瓦斯含量预测[J]. 煤矿安全, 2018, 49(09): 211-214.

[30]Duda A, Krzemień A. Forecast of Methane Emission from Closed Underground Coal Mines Exploited by Longwall Mining–A Case Study of Anna Coal Mine[J]. Journal of Sustainable Mining, 2018, 17(4): 184-194.

[31]谢素璞, 马鹏. 煤层瓦斯含量与埋深之间的关系研究[J]. 能源技术与管理, 2016, 41(4): 21-23.

[32]杨玉中, 高金龙. 煤层瓦斯含量与煤层埋藏深度之间的定量关系[J]. 煤矿安全, 2014, 45(01): 1-4.

[33]时天. 基于混沌时间序列的瓦斯浓度预测研究[D]. 西安科技大学, 2009.

[34]李琪, 屈峰涛, 何璟彬, 等. 基于PSO-BP的钻井机械钻速预测模型[J]. 科学技术与工程, 2021, 21(19): 7984-7990.

[35]王金华, 黄乐亭, 李首滨, 等. 综采工作面智能化技术与装备的发展[J]. 煤炭学报, 2014, 39(08): 1418-1423.

[36]王鹏, 伍永平, 王栓林, 等. 矿井瓦斯浓度Lagrange-ARIMA实时预测模型研究[J]. 煤炭科学技术, 2019, 47(4): 141-146.

[37]李文顺, 屈世甲. 监控系统瓦斯浓度数据预处理方法的探讨[J]. 能源技术与管理, 2014, 39(3): 23-26 .

[38]陈辉, 周文超, 魏永强, 等. 矿压缺失数据插值算法研究[J]. 工矿自动化, 2011, 37(11): 22-25.

[39]Mishra P, Biancolillo A. New Data Preprocessing Trends Based on Ensemble of Multiple Preprocessing Techniques[J]. TrAC Trends in Analytical Chemistry, 2020, 132: 116045.

[40]窦超, 张生鹏, 赵利江. 小波阈值在GNSS时间序列去噪中的选取研究[J]. 甘肃科学学报, 2021, 33(06): 6-9.

[41]张博. 基于小波去噪的BP神经网络在变形监测中的应用[J]. 北京测绘, 2021, 35(12): 1592-1596.

[42]Guo X, Yang L, et al. Research on Gas Concentration Prediction Based on Wavelet Denoising and ARIMA Model[C]. ICMSP. IEEE, 2021: 387-390.

[43]平力, 赵蓉, 杨斌, 等. 基于PCA-BP神经网络的光谱消光法颗粒粒径反演算法研究[J]. 光谱学与光谱分析, 2021, 41(11): 3639-3647.

[44]Li Z, Wang T, Wang Y, et al. A Wavelet Threshold Denoising-Based Imbalance Fault Detection Method for Marine Current Turbines[J]. IEEE Access, 2020, 8: 29815-29825

[45]Ma J, Yuan Y. Dimension Reduction of Image Deep Feature Using PCA[J]. Journal of Visual Communication and Image Representation, 2019, 63: 102578.

[46]张文娟, 侯媛彬, 张文涛, 等. 基于GA-LSSVR的煤矿瓦斯数据去噪研究[J]. 矿业安全与环保, 2017, 44(1): 45-48+61.

[47]刘剑桥. 基于改进LSTM循环神经网络瓦斯数据时间序列预测研究[D]. 2018.

[48]Leevy J L, Khoshgoftaar T M, et al. Investigating the Relationship between Time And Predictive Model Maintenance[J]. Journal of Big Data, 2020, 7(1): 1-19.

[49]杨茂林, 薛友欣, 姜耀东, 等. 离柳矿区综采工作面瓦斯涌出规律研究[J]. 煤炭学报, 2009, 34(10): 1349-1353.

[50]Sajjad M, Khan Z A, et al. A Novel CNN-GRU-Based Hybrid Approach for Short-Term Residential Load Forecasting[J]. IEEE Access, 2020, 8: 143759-143768.

[51]刘俊娥, 杨晓帆, 郭章林. 基于 FIG-SVM 的煤矿瓦斯浓度预测[J]. 中国安全科学学报, 2013, 23(02): 80-84.

[52]Valtierra-Rodriguez M, de Jesus Romero-Troncoso R, et al. Detection and Classification of Single and Combined Power Quality Disturbances Using Neural Networks[J]. IEEE Transactions on Industrial Electronics, 2013, 61(5): 2473-2482.

[53]王永福. 黄陵二号煤矿综采工作面瓦斯灾害防治技术[J]. 陕西煤炭, 2015, 34(02): 79-80+118.

[54]张洋铭, 万定生. 改进PSO-LSTM的水文时间序列预测[J]. 计算机工程与设计, 2022, 43(1): 203-209.

[55]董丁稳, 屈世甲, 王红刚. 基于监测数据关联分析的瓦斯浓度预警方法[J]. 工矿自动化, 2015, 41(6):1-5.

[56]McBride S K, Bostrom A, Sutton J, et al. Developing Post-Alert Messaging for ShakeAlert, The Earthquake Early Warning System for The West Coast of The United States of America[J]. International Journal of Disaster Risk Reduction, 2020, 50: 101713.

[57]Wang Y, Liu J, He X, et al. Design and Realization of Rock Salt Gas Storage Database Management System Based on SQL Server[J]. Petroleum, 2018, 4(4): 466-472.