随着煤炭开采越来越深，井下开采环境也愈发复杂，极易诱发采空区遗煤发生自燃灾害。如何有效地防范煤自燃灾害发生，对促进煤炭安全开采，保障国民经济平稳发展具有重要意义。论文为提高预测煤自燃危险性的精度，进行了以下研究：

(1) 针对高噪音的多特征煤自燃监测数据影响煤自燃危险性预测精度的问题，提出了基于深度学习的煤自燃多特征指标数据动态融合方法(DMCNN)。首先通过自编码器对特征进行降噪处理；其次将降噪数据转成二维特征矩阵，采用滑动窗口将特征矩阵进行切片；最后采用CNN卷积神经网络提取特征矩阵上的有效特征，并进行有效特征数据融合。实验结果表明，论文提出的DMCNN融合算法融合的数据在MAE上误差比未融合的数据降低20.99%，提升了数据的可信度和利用率，增强了模型预测的拟合效果。

(2) 针对传统煤自燃温度预测模型预测精度不高且鲁棒性较差的问题，提出了一种基于改进粒子群算法优化简单循环神经网络的煤自燃温度预测模型(PSO-SRU)。首先构建SRU预测模型学习训练集中煤自燃温度与特征指标间非线性规律，对煤自燃温度进行预测；其次利用PSO算法优化模型参数，为防止寻优陷入局部最小值，改进PSO优化算法平衡全局和局部搜索能力；最后利用搜寻到SRU预测模型的最优超参数，建立最优预测模型。实验结果表明，改进PSO-SRU预测模型在一定程度上提高了预测精度；且RMSE指标降低了60.76%，增强了SRU预测模型的泛化性和鲁棒性。

针对煤矿安全管理系统煤自燃灾害预警功能单一的问题，论文设计并实现了煤自燃智能预警系统，将基于深度神经网络的煤自燃温度预测方法应用到预警系统中，实现了对煤自燃危险性的预测预警功能，经过工程应用实测，预警系统页面布局合理，操作简单流畅，所用功能模块均能正常使用，系统整体运行良好，取得了较好的煤自燃监测预警预期效果，能够及时发现煤自燃灾害趋势并发布告警，达到开发的要求，为减少煤自燃发生的可能性，保障煤矿安全生产的平稳运行，提供了有力的技术支撑。

As the coal mining becomes deeper and deeper, the underground mining environment becomes more and more complex, and it is easy to induce spontaneous combustion disasters in the coal leftover in the goaf. How to effectively prevent the occurrence of coal spontaneous combustion disasters is of great significance to promoting the safe mining of coal and ensuring the stable development of the national economy. In order to improve the accuracy of predicting the risk of spontaneous combustion of coal, the following researches are carried out:

(1) Aiming at the problem that the high-noise multi-feature coal spontaneous combustion monitoring data affects the prediction accuracy of coal spontaneous combustion risk, a deep learning-based multi-feature index data dynamic fusion method for coal spontaneous combustion (Denoising Autoencoder Mobile CNN Fusion, DMCNN) is proposed. First, the features are denoised by the autoencoder; secondly, the denoised data is converted into a two-dimensional feature matrix, and the feature matrix is sliced by a sliding window; finally, the CNN(Convolutional neural network, CNN) convolutional neural network is used to extract the effective features on the feature matrix, and the Effective feature data fusion. The experimental results show that the error of the data fused by the DMCNN fusion algorithm proposed in the paper is 20.99% lower than that of the unfused data in MAE, which improves the utilization of data and enhances the fitting effect of model prediction.

(2) Aiming at the problems of low prediction accuracy and poor robustness of the traditional coal spontaneous combustion temperature prediction model, this thesis proposes a coal spontaneous combustion temperature prediction model (PSO-SRU) based on the improved particle swarm optimization optimization simple recurrent neural network. Firstly, the SRU prediction model is constructed to learn the nonlinear law between coal spontaneous combustion temperature and characteristic indexes in the training set, and then the coal spontaneous combustion temperature is predicted; secondly, the PSO algorithm is used to optimize the model parameters. In order to prevent the optimization from falling into the local minimum, the improved PSO optimization algorithm balances the global and Local search ability; finally, the optimal hyperparameters of the SRU prediction model are found to establish the optimal prediction model. The experimental results show that the improved PSO-SRU prediction model improves the prediction accuracy to a certain extent, and the RMSE index is reduced by 60.76%, which enhances the generalization and robustness of the SRU prediction model.

Aiming at the problem of the single coal spontaneous combustion disaster early warning function in the coal mine safety management system, this thesis designs and implements an intelligent coal spontaneous combustion early warning system, and applies the coal spontaneous combustion temperature prediction method based on the deep neural network to the early warning system. The prediction and early warning function has been tested by engineering application. The page layout of the early warning system is reasonable, the operation is simple and smooth, all functional modules can be used normally, the system is running well as a whole, and the expected effect of coal spontaneous combustion monitoring and early warning has been achieved, and the coal spontaneous combustion disaster can be detected in time. Trends and alarms are issued to meet the development requirements, providing strong technical support for reducing the possibility of spontaneous coal combustion and ensuring the smooth operation of coal mine safety production.

[1] 2020中国统计年鉴[J].统计理论与实践, 2021(1): 2.

[2] 谢和平, 吴立新, 郑德志. 2025年中国能源消费及煤炭需求预测[J]. 煤炭学报, 2019, 44(7): 1949-1960.

[3] 孙旭东, 张蕾欣, 张博. 碳中和背景下我国煤炭行业的发展与转型研究[J]. 中国矿业, 2021, 30(02): 1-6.

[4] 秦波涛, 仲晓星, 王德明, 等. 煤自燃过程特性及防治技术研究进展[J]. 煤炭科学技术, 2021, 49(1): 66-99.

[5] 梁运涛, 侯贤军, 罗海珠, 等. 我国煤矿火灾防治现状及发展对策[J]. 煤炭科学技术, 2016, 44(6):1-6+13. DOI:10.13199/j. cnki.cst. 2016. 06. 001.

[6] 刘泽健, 徐永亮, 吕志广, 等. 热-应力耦合作用下不同粒径煤自燃氧化升温特性[J].工矿自动化, 2021, 47(09):91-95. DOI:10. 13272/j. issn. 1671-251x. 17805.

[7] 邓军, 白祖锦, 肖旸, 等. 煤自燃灾害防治技术现状与挑战[J]. 煤矿安全, 2020, 51(10): 118-125.

[8] 王双明. 对我国煤炭主体能源地位与绿色开采的思考[J]. 中国煤炭, 2020, 46(2): 11-16. DOI:10. 19880/j. cnki. ccm. 2020. 02. 002.

[9] 吴群英, 陈国梁, 冯涛. 煤炭资源开发与利用的立体式生态模式—以陕北矿区“1+4”绿色可持续生态建设为例[J]. 煤炭学报, 2020, 45(12): 4163-4169. DOI:10. 13225/j. cnki. jccs. 2020. 1552.

[10] 张伟霞. 基于SEM的采空区煤自燃危险性预判方法研究与应用[D]. 西安科技大学, 2019.

[11] 寇砾文, 蒋曙光, 王兰云, 等. 煤自燃指标气体产生规律及影响因素分析[J]. 矿业研究与开发, 2012, 32(2):67-70+116. DOI:10. 13827/j. cnki. kyyk. 2012. 02. 025.

[12] 朱红青, 王海燕, 王斐然, 等. 煤堆测温技术研究进展[J]. 煤炭科学技术, 2014, 42(1): 50-54.

[13] 陈欢, 杨永亮. 煤自燃预测技术研究现状[J]. 煤矿安全, 2013, 44(9): 194-197.

[14] 杨志凌, 刘俊华. 基于数据融合和Wiener过程的风电轴承剩余寿命预测[J]. 太阳能学报, 2021, 42(10): 189-194. DOI:10. 19912/j. 0254-0096. tynxb. 2019-0790.

[15] 董真杰, 郑琛瑶, 张国龙. 不同精度数据融合的自适应加权平均法研究[J]. 舰船电子工程, 2014, 34(10): 31-33+126.

[16] 胡旭冉, 韩震, 李静, 等. 基于集合卡尔曼滤波的海表面温度融合研究[J]. 海洋科学进展, 2018, 36(3): 394-401.

[17] 安世奇, 由东媛. 用于胎压监测系统的一种改进贝叶斯估计数据融合的研究[J]. 传感技术学报, 2018, 31(6): 915-919.

[18] 任超, 牛宏侠, 陈光武, 等. D-S证据下的高铁信号系统异构数据融合[J]. 传感器与微系统, 2022, 41(3): 54-57+61. DOI:10. 13873/J. 1000-9787(2022)03-0054-04.

[19] Azcarate Silvana M, Ríos Reina Rocío, Amigo José M , et al. Data Handling in Data Fusion: Methodologies and Applications[J]. TrAC Trends in Analytical Chemistry, 2021(prepublish).

[20] Fawzy Dina, Moussa Sherin, Badr Nagwa. The Spatiotemporal Data Fusion (STDF) Approach: IoT-Based Data Fusion Using Big Data Analytics[J]. Sensors, 2021, 21(21).

[21] 张帅涛, 蒋品群, 宋树祥, 等. 基于注意力机制和CNN-LSTM融合模型的锂电池SOC预测[J/OL]. 电源学报: 1-15. http: //kns. cnki. net/kcms/detail/12. 1420. TM. 20220222. 1322. 005. html

[22] 孙淑娥, 姚柳, 赵怡. 基于卷积神经网络的时间序列数据融合算法[J]. 西安石油大学学报(自然科学版), 2021, 36(5): 136-142.

[23] Chandra B S, Sastry C S, Jana S. Robust Heartbeat Detection From Multimodal Data via CNN-Based Generalizable Information Fusion.[J]. IEEE transactions on bio-medical engineering, 2019, 66(3).

[24] 仲晓星, 王建涛, 周昆. 矿井煤自燃监测预警技术研究现状及智能化发展趋势[J].工矿自动化, 2021, 47(9): 7-17. DOI: 10. 13272/j. issn. 1671-251x. 17841.

[25] 李同锁, 何启林. 集对分析在煤炭自然发火危险性评价上的应用[J]. 中国安全生产科学技术, 2013, 9(08): 54-58.

[26] 邓军, 杨一帆, 翟小伟. 基于层次分析法的矿井煤自燃危险性模糊评价[J]. 中国安全生产科学技术, 2014, 10(4):120-125.

[27] 王明重, 刘泽功, 张箫剑, 等. 基于AHP和扩展集对理论的采空区遗煤自燃危险性评价研究[J]. 中国安全生产科学技术, 2014, 10(08):182-188.

[28] 桂祥友, 郁钟铭, 孟絮屹. 基于神经网络的自然发火危险性评价与预测[J]. 采矿与安全工程学报, 2008, 25(04):453-457.

[29] 左云飞. 基于熵权物元可拓的采空区自燃危险性评价[J]. 煤炭技术, 2018, 37(7):192-195. DOI: 10. 13301/j. cnki. ct. 2018. 07. 070.

[30] 皮子坤, 董子文, 罗陈, 等. 基于IIAHP-Entropy-ssd的采空区遗煤自燃危险性评价模型及应用[J]. 中国安全生产科学技术, 2020, 16(07): 61-67.

[31] 岳宁芳, 金彦, 孙明福, 等. 基于多指标气体的煤自燃进程分级预警研究[J]. 安全与环境学报, 2020, 20(6):2139-2146. DOI:10. 13637/j. issn. 1009-6094. 2020. 1032.

[32] 邓军, 雷昌奎, 曹凯, 等. 煤自燃预测的支持向量回归方法[J]. 西安科技大学学报, 2018, 38(2):175-180. DOI:10. 13800/j. cnki. xakjdxxb. 2018. 0201.

[33] Zhang Qian, Li Hai Gang. An improved least squares SVM with adaptive PSO for the prediction of coal spontaneous combustion.[J]. Mathematical biosciences and engineering : MBE, 2019, 16(4).

[34] 刘宝, 穆坤, 叶飞, 等. 基于相关向量机的煤自燃预测方法[J]. 工矿自动化, 2020, 46(9):104-108. DOI:10. 13272/j. issn. 1671-251x. 17578.

[35] Qifei WANG, Chengwu LI, Yulu SHANG. Gas Indexes Optimization of Coal Spontaneous Combustion Prediction based on the Information Entropy Theory[C]//. Proceedings of the 7th Academic Conference of Geology Resource Management and Sustainable Development, 2019: 264-272.

[36] 昝军才, 魏成才, 蒋可娟, 等. 基于BP神经网络的煤自燃温度预测研究[J]. 煤炭工程, 2019, 51(10):113-117.

[37] 郑学召, 李梦涵, 张嬿妮, 等. 基于随机森林算法的煤自燃温度预测模型研究[J]. 工矿自动化, 2021, 47(5):58-64. DOI:10. 13272/j. issn. 1671-251x. 17700.

[38] 岳宁芳, 金彦, 孙明福, 等. 基于多指标气体的煤自燃进程分级预警研究[J]. 安全与环境学报, 2020, 20(6):2139-2146. DOI:10. 13637/j. issn. 1009-6094. 2020. 1032.

[39] 孟倩. 基于计算智能的煤矿自然发火预警研究[D]. 中国矿业大学, 2015.

[40] 王伟峰, 邓军, 侯媛彬. 煤自燃危险区域高密度网络化监测预警系统设计与应用[J].煤炭技术, 2018, 37(3):218-220. DOI:10. 13301/j. cnki. ct. 2018. 03. 083.

[41] 邓慧芳, 谢锐, 孙传猛, 等. 高可靠性煤矿井下火源无线预警监测系统[J]. 中国测试, 2018, 44(6): 83-86+94.

[42] 张巨峰, 施式亮, 鲁义, 等. 大数据下瓦斯与煤自燃共生灾害智能预警系统:数据特征、应用架构、关键技术[J]. 中国安全科学学报, 2021, 31(9): 60-66. DOI: 10. 16265/j. cnki. issn1003-3033. 2021. 09. 009.

[43] 肖国强, 范新丽, 马砺, 等. 矿井火灾环境信息动态感知与应急隔离系统[J]. 工矿自动化, 2021, 47(9): 65-69. DOI: 10. 13272/j. issn. 1671-251x. 17819.

[44] 邓军, 徐精彩, 陈晓坤. 煤自燃机理及预测理论研究进展[J]. 辽宁工程技术大学学报, 2003(4): 455-459.

[45] Cortes Corinna, Vapnik Vladimir. Support-vector networks[J]. Machine Learning, 1995, 20: 273-297.

[46] Zhang Qian , Li Hai Gang. An improved least squares SVM with adaptive PSO for the prediction of coal spontaneous combustion.[J]. Mathematical biosciences and engineering : MBE, 2019, 16(4): 3169-3182.

[47] Haiqing Yang, Hima Nikafshan Rad, Mahdi Hasanipanah, et al. Prediction of Vibration Velocity Generated in Mine Blasting Using Support Vector Regression Improved by Optimization Algorithms[J]. Natural Resources Research, 2020, 29(2) : 807-830.

[48] Breiman L. Random forests[J]. Machine Learning, 2001, 45(1):5-32.

[49] 王伟同, 范海东, 梁成思, 等. 基于随机森林算法的对冲锅炉出口NO_(x)排放量预测模型研究[J/OL]. 热力发电:1-9. DOI:10. 19666/j. rlfd. 202110220.

[50] Rumelhart, David E, G. E. Hinton, et al. Learning Representations by Back Propagating Errors[J]. Nature, 1986, 323 (6088): 533-536.

[51] Luan Zhangbao. Research on Geological Disaster Forecast Method Based on PCA Structure BP Model[J]. Journal of Physics: Conference Series, 2021, 2083(4).

[52] TAO Lei, YU Zhang, HUI Dai, et al. Simple recurrent units for highly parallelizable recurrence[J]. arXiv preprint arXiv: 1709. 02755, 2017.

[50] Han Junjun, Zhu Xiaoqing. The two-layer SRU Neural Network Based Analysis of Time Series[J]. Journal of Physics: Conference Series, 2021, 1952(4).

[53] Kennedy J, Eberhart R C. Particle swarm optimization //Proceedings of the IEEE International Conference on Neural Networks. Perth, 1995:1942 -1948.

[54] Zhang Xinming,Lin Qiuying. Three-learning strategy particle swarm algorithm for global optimization problems[J]. Information Sciences,2022,593.

[55] 闫群民, 马瑞卿, 马永翔, 等. 一种自适应模拟退火粒子群优化算法[J]. 西安电子科技大学学报, 2021, 48(4):120-127. DOI:10.19665/j. issn1001-2400. 2021. 04. 016.

[56] 季新芳, 张勇, 巩敦卫, 等. 异构集成代理辅助的区间多模态粒子群优化算法[J/OL]. 自动化学报: 1-23. DOI:10. 16383/j. aas. c210223.

[57] Storn R, Price K. Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces[J]. Journal of Global Optimization. 1997(11): 341–359.

[58] 于成龙, 付国霞, 孙超利, 等. 全局与局部模型交替辅助的差分进化算法[J]. 计算机工程, 2022, 48(3):115-123. DOI:10. 19678/j. issn. 1000-3428. 0060693.

[59] 邹东池, 白富生. 基于差分进化算法的神经网络优化设计[J]. 重庆师范大学学报(自然科学版), 2022, 39(1): 79-89.

[60] 邬博文, 朱良宽, 王璟瑀. 基于差分进化鲸鱼优化算法的森林冠层图像分割[J]. 西北林学院学报, 2022, 37(1): 67-73+136.

[61] 陶杰, 王欣, 孙慧贤, 等. 一种窄带高斯白噪声的生成与实现方法[J]. 通讯世界, 2017(23): 350-351.