随着煤矿生产智能化的发展，建立可靠的井下通信系统是保障煤矿安全开采的关键。由于矿井巷道狭窄细长，空间受限且环境复杂，无线电波在传输过程中会受到多种衰落因子的影响，导致接收点的场强波动和信号衰减剧烈。本文利用深度学习方法建立复杂矿井巷道无线电波场强预测模型，对接收端信号场强的分布趋势进行预测，为高效建设矿井无线通信系统提供重要依据。

本文以矩形巷道为研究对象，通过波导模式理论及公式推导，总结了巷道中影响无线电波传输的主要因素，如天线工作频率、巷道宽度和高度、巷道壁倾斜度和粗糙度、煤岩电参数等，将其作为场强预测模型的输入变量。为验证预测模型的有效性，利用ZigBee无线数据传输技术，设计并实现了一套以CC2530芯片为核心的矿井巷道无线信道测量系统，并在矩形巷道和半圆拱形巷道中完成实地测量，获得真实巷道场强数据集。

针对传统场强预测建模计算复杂度高、预测精度低及普适性差等问题，结合矿井巷道大尺度和小尺度衰落特性，建立了基于深度学习的矿井无线电波场强预测模型。设计改进卷积神经网络（Convolutional Neural Network, CNN）预测模型，在每个卷积层后加入批量归一化层，加快网络收敛速度。为提取时间序列的隐含特征，设计了双向长短期记忆（Bidirectional Long Short Term Memory Network, BiLSTM）网络预测模型。将改进CNN和BiLSTM模型应用于矿井巷道场强预测中，实验结果表明，两种模型在预测场强方面均表现出良好的性能和预测精度。

为进一步提高模型的拟合效果和鲁棒性，增强模型对数据特征的挖掘能力，在单一网络的基础上，将改进CNN和BiLSTM网络串联组合，同时对CNN-BiLSTM网络中的参数进行调优，并引入注意力机制，提出了CNN-BiLSTM-AM场强预测模型。利用实测数据集对模型性能进行验证，实验结果表明，所提模型的相关系数为0.9989，预测效果最优。说明注意力机制与深度学习的方法相结合，可以有效提高矿井巷道场强分布预测的准确度。

With the development of intellectualization of coal mine production, establishing a reliable underground communication system is the key to ensure the safe mining of coal mines. Because of the narrow and long mine roadway, limited space and complex environment, radio waves are affected by many fading factors in the transmission process, resulting in strong field fluctuation and signal attenuation at the receiving point. In this thesis, a deep learning method is used to establish a radio wave field strength prediction model for complex mine roadways, and to predict the distribution trend of the signal field strength at the receiving end, which provides an important basis for the efficient construction of a mine wireless communication system.

In this thesis, the rectangular roadway is taken as the research object, and the main factors affecting the transmission of radio waves in the roadway are summarized through the wave guide mode theory and formula derivation, such as the working frequency of the antenna, the width and height of the roadway, the inclination and roughness of the roadway wall, and the electrical parameters of coal and rock, and uses them as the input variable of the field strength prediction model. In order to verify the validity of the prediction model, a set of wireless channel measurement system for mine roadway with CC2530 chip as the core is designed and implemented by using ZigBee wireless data transmission technology, and the field measurement is completed in the rectangular roadway and semi-circular arch roadway, and real roadway field strength dataset are obtained.

Aiming at the problems of high computational complexity, low prediction accuracy and poor universality of traditional field strength prediction modeling, a mine radio field strength prediction model based on deep learning is established in combination with the large-scale and small-scale decay characteristics of mine roadways. An improved convolution neural network (Convolutional Neural Network, CNN) prediction model is designed. A batch normalization layer is added after each convolution layer to speed up network convergence. To extract the implicit characteristics of time series, a bidirectional long-term and short-term memory (Bidirectional Long Short Term Memory Network, BiLSTM) network prediction model is designed. The improved CNN and BiLSTM models are applied to the prediction of mine roadway field strength. The experimental results show that both models have good performance and prediction accuracy in predicting field strength.

To further improve the fitting effect and robustness of the model, and to enhance the mining ability of the model for data characteristics, on the basis of a single network, the improved CNN and BiLSTM network will be combined in series, the parameters of CNN-BiLSTM network are optimized, and the attention mechanism is introduced, and a CNN-BiLSTM-AM field strength prediction model is proposed. The model performance is validated by the measured data set. The results show that the correlation coefficient of the proposed model is 0.9989, and the prediction effect is the best. It shows that the combination of attention mechanism and deep learning method can effectively improve the accuracy of field intensity distribution prediction in mine roadway.

[1] 王国法, 任世华, 庞义辉, 等. 煤炭工业“十三五”发展成效与“双碳”目标实施路径[J]. 煤炭科学技术, 2021, 49(09): 1-8.

[2] 张博, 孙旭东, 刘颖, 彭苏萍. 能源新技术新兴产业发展动态与2035战略对策[J]. 中国工程科学, 2020, 22(02): 38-46.

[3] 王国法, 赵国瑞, 胡亚辉. 5G技术在煤矿智能化中的应用展望[J]. 煤炭学报, 2020, 45(01): 16-23.

[4] 孙继平. 煤矿智能化与矿用5G[J]. 工矿自动化, 2020, 46(08): 1-7.

[5] 王国法. 煤矿智能化最新技术进展与问题探讨[J]. 煤炭科学技术, 2022, 50(01): 1-27.

[6] 范京道, 闫振国, 李川. 基于5G技术的煤矿智能化开采关键技术探索[J]. 煤炭科学技术, 2020, 48(07): 92-97.

[7] 姚善化. 基于镜像法的矿井隧道电磁波多径信道模型[J]. 工矿自动化,2017, 43(04): 46-49.

[8] 樊荣, 宋文, 黄强. 矿井无线通信系统研究与发展[J]. 西安科技大学学报, 2010, 30(04): 471-474.

[9] 胡青松, 杨维, 丁恩杰, 等. 煤矿应急救援通信技术的现状与趋势[J]. 通信学报, 2019, 40(05): 163-179.

[10] 张长森, 张艳芳. 矿井移动通信中Nakagami衰落信道模型的研究[J]. 计算机工程与应用, 2014, 50(07): 238-241．

[11] 张晓光, 李蕾蕾, 郑红党, 等. 煤矿井下MIMO系统建模与信道容量分析[J]. 西安科技大学学报, 2011, 31(03): 366-370．

[12] Sun Z, Akyildiz I F. Channel modeling and analysis for wireless networks in underground mines and road tunnels[J]. IEEE Transactions on communications, 2010, 58(6): 1758-1768.

[13] Forooshani A E, Bashir S, Michelson D G, et al. A survey of wireless communications and propagation modeling in underground mines[J]. IEEE Communications surveys & tutorials, 2013, 15(4): 1524-1545.

[14] Sun J, Cheng L. Equivalent analysis method of electromagnetic waves propagation in arched tunnel[J]. Applied mathematics and computation, 2006, 178(2): 332-337.

[15] 姚善化. 基于镜像法的矿井隧道电磁波多径信道模型[J]. 工矿自动化, 2017, 43(04): 46-49.

[16] Degli-Esposti V, LU J S, WU J N, et al. A semi-deterministic model for outdoor-to-indoor prediction in urban areas[J]. IEEE Antennas and Wireless Propagation Letters, 2017, 16: 2412-2415.

[17] 陶成, 由明昕, 谈振辉, 等. 基于射线跟踪法的隧道无线信道特性[J]. 北京交通大学学报, 2016, 40(02): 1-7.

[18] 米广双, 杨维. 基于镜像原理的半圆拱形矿井巷道电磁波传播场强预测[J]. 煤炭学报, 2020, 45(S2): 1089-1099.

[19] Zhou C. Ray tracing and modal methods for modeling radio propagation in tunnels with rough walls[J]. IEEE transactions on antennas and propagation, 2017, 65(5): 2624-2634.

[20] Ghaddar M, Mabrouk I B, Nedl M, et al. Deterministic modeling of 5G millimeter-wave communication in an underground mine tunnel[J]. IEEE Acess, 2019, 7: 116519-116528.

[21] 刘留, 张建华, 樊圆圆, 等. 机器学习在信道建模中的应用综述[J]. 通信学报, 2021, 42(2): 134-153.

[22] Zheng K, Ou S, Yin X. Massive MIMO channel models: A survey[J]. International Journal of Antennas and Propagation, 2014, 2014(11): 1-10.

[23] Liu J, Jin X, Dong F, et al. Fading channel modelling using single-hidden layer feedforward neural networks[J]. Multidimensional Systems and Signal Processing, 2017, 28(3): 885-903.

[24] 蒋育康, 郭爱煌, 艾勃, 等. 城市轨道交通隧道环境下大规模MIMO信道建模[J]. 铁道学报, 2018, 40(11): 84-90.

[25] 李翠然, 季一凡, 吕安琪, 等. 几何随机分布模型的高速铁路山地信道建模[J]. 重庆邮电大学学报(自然科学版), 2021, 33(03): 403-411.

[26] Zhou C, Plass T, Jacksha R, et al. RF propagation in mines and tunnels: Extensive measurements for vertically, horizontally, and cross-polarized signals in mines and tunnels[J]. IEEE Antennas and Propagation Magazine, 2015, 57(4): 88-102.

[27] Yarkan S, Arslan H. Statistical wireless channel propagation characteristics in underground mines at 900MHz[C]//MILCOM 2007-IEEE Military Communications Conference. IEEE, 2007: 1-7.

[28] Hakem N, Delisle G, Coulibaly Y. Radio-wave propagation into an underground mine environment at 2.4 GHz, 5.8 GHz and 60 GHz[C]//The 8th European Conference on Antennas and Propagation (EuCAP 2014). IEEE, 2014: 3592-3595.

[29] Tariq S A M, Despins C, Affes S, et al. Statistical modeling of 60 GHz wireless channel in an underground mine gallery[C]//2015 IEEE International Conference on Ubiquitous Wireless Broadband (ICUWB). IEEE, 2015: 1-5.

[30] Rissafi Y, Talbi L, Ghaddar M. Experimental characterization of an UWB propagation channel in underground mines[J]. IEEE Transactions on Antennas and Propagation, 2011, 60(1): 240-246.

[31] El Khaled M, Fortier P, Ammari M L. A performance study of line-of-sight millimeter-wave underground mine channel[J]. IEEE Antennas and Wireless Propagation Letters, 2014, 13: 1148-1151.

[32] 王安义, 郭世坤. 基于LS-SVM的矿井巷道场强预测[J]. 工矿自动化, 2014, 40(10): 36-40.

[33] 武金莉. 基于PSO优化LS-SVM参数的矿井巷道场强预测[D]. 西安: 西安科技大学, 2017.

[34] 王安义, 郗茜. 基于遗传算法优化的LS-SVM井下场强预测[J]. 工矿自动化, 2016, 42(12): 46-50.

[35] 马晶晶. 煤矿复杂环境下的无线信道特性分析与建模仿真[D]. 西安: 西安科技大学, 2019.

[36] O’shea T, Hoydis J. An introduction to deep learning for the physical layer[J]. IEEE Transactions on Cognitive Communications and Networking, 2017, 3(4): 563-575.

[37] Ye H, Li G Y, Juang B H. Power of deep learning for channel estimation and signal detection in OFDM systems[J]. IEEE Wireless Communications Letters, 2017, 7(1): 114-117.

[38] Erpek T, O’shea T, Sagduyu Y, et al. Deep learning for wireless communications[M]. Berlin: Springer, 2020.

[39] Liu J, Jin X, Dong F, et al. Fading channel modelling using single-hidden layer feedforward neural networks[J]. Multidimensional systems and signal processing, 2017, 28(3): 885-903.

[40] Zhu G, Liu Y, Li S. Study on the propagation characteristics of millimeter wave channel at 39 GHz based on BP neural network[C]//2021 International Conference on Microwave and Millimeter Wave Technology (ICMMT). IEEE, 2021: 1-3.

[41] Huang J, Wang C X, Bai L, et al. A big data enabled channel model for 5G wireless communication systems[J]. IEEE Transactions on Big Data, 2018, 6(2): 211-222.

[42] Bai L, Wang C X, Huang J, et al. Predicting wireless mmWave massive MIMO channel characteristics using machine learning algorithms[J]. Wireless Communications and Mobile Computing, 2018, 2018:1-12.

[43] Zaarour N, Affes S, Kandil N, et al. Comparative study on a 60 GHz path loss channel modeling in a mine environment using neural networks[C]//2015 IEEE International Conference on Ubiquitous Wireless Broadband (ICUWB). IEEE, 2015: 1-4.

[44] Lu T, Sun J, Wu K, et al. High-speed channel modeling with machine learning methods for signal integrity analysis[J]. IEEE Transactions on Electromagnetic Compatibility, 2018, 60(6): 1957-1964.

[45] 孙湘琛. 基于BP神经网络方法的28GHz毫米波信道建模研究[D]. 南京: 南京邮电大学, 2019.

[46] Zhao X, Du F, Geng S, et al. Neural network and GBSM based time-varying and stochastic channel modeling for 5G millimeter wave communications[J]. China Communications, 2019, 16(6): 80-90.

[47] 王艳芬. 矿井特殊环境下的超宽带无线通信信道模型研究[M]. 徐州: 中国矿业大学出版社. 2012: 20.

[48] 李芳, 牛会永, 李石林, 等. 媒介电常数影响因素研究现状及分析[J]. 煤矿安全, 2017, 48(12): 218-220, 224.

[49] 孟庆勇, 李长峰, 温良, 等. 矿井巷道ZigBee无线传输特性研究[J]. 工矿自动化, 2012, 38(01): 40-42.

[50] Gu J, Wang Z, Kuen J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377.

[51] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. PMLR, 2015: 448-456.

[52] 王安义, 周孝铭. 基于改进卷积神经网络的矿井巷道场强预测[J]. 工矿自动化, 2021, 47(10): 49-53+84.

[53] Jiang W, Schotten H D. Deep learning for fading channel prediction[J]. IEEE Open Journal of the Communications Society, 2020, 1: 320-332.

[54] Sak H, Senior A W, Beaufays F. Long short-term memory recurrent neural network architectures for large scale acoustic modeling[J]. Computer Science, 2014:338-342.

[55] Liu H, Mi X, Li Y. Wind speed forecasting method based on deep learning strategy using empirical wavelet transform, long short term memory neural network and Elman neural network[J]. Energy conversion and management, 2018, 156: 498-514.

[56] 金宸, 李维华, 姬晨, 等. 基于双向LSTM神经网络模型的中文分词[J]. 中文信息学报, 2018, 32(02): 29-37.

[57] Zou H, Wu Y, Zhang H, et al. Short-term traffic flow prediction based on PCC-BiLSTM[C]//2020 International Conference on Computer Engineering and Application (ICCEA). IEEE, 2020: 489-493.

[58] Chen Y, Wang Y, Dong Z, et al. 2-D regional short-term wind speed forecast based on CNN-LSTM deep learning model[J]. Energy Conversion and Management, 2021, 244: 114451.

[59] Yan R, Liao J, Yang J, et al. Multi-hour and multi-site air quality index forecasting in Beijing using CNN, LSTM, CNN-LSTM, and spatiotemporal clustering[J]. Expert Systems with Applications, 2021, 169: 114513.

[60] Kim T Y, Cho S B. Predicting residential energy consumption using CNN-LSTM neural networks[J]. Energy, 2019, 182: 72-81.

[61] 朱张莉, 饶元, 吴渊, 等. 注意力机制在深度学习中的研究进展[J]. 中文信息学报, 2019, 33(06): 1-11.