由于矿井环境的复杂多变性，无线信道往往会出现多径衰落的现象。正交频分复用（Orthogonal Frequency Division Multiplexing, OFDM）技术由于其抗频率选择性衰落和克服信号符号间干扰的优势，在矿井无线传输系统中得到了应用。在矿井OFDM传输系统中，井下环境恶劣导致信道特性的精确估计变得困难，这极大地限制了矿井无线通信的传输质量。面对矿井通信系统存在的问题，深度学习技术提供了一种新的解决方案。基于矿井OFDM系统，本文对基于深度学习的信道估计算法展开研究，主要工作有以下几个方面：

（1）针对煤矿井下环境恶劣，传统信道估计算法存在准确度低的问题，本文提出一种改进图像超分辨率卷积网络（Super-Resolution Convolutional Neural Network, SRCNN）进行信道估计。在改进SRCNN模型中，将导频处的估计值作为输入，改进SRCNN取代了传统信道估计算法中的插值过程，加入高效通道注意力提高通道特征的学习，实现了信道的准确估计。通过对训练数据的自主学习，提取数据规律，本文提出的改进SRCNN信道估计算法精度达到10^-4。在此基础上，在不同信道估计算法、不同数目导频、不同调制方式、不同注意力机制以及不同训练信噪比多个条件下进行了仿真。仿真结果表明，相较于传统的信道估计算法，该算法在信道估计上性能表现更为优异。

（2）考虑到实际矿井通信场景复杂多变，真实的矿井信号数据难以采集，本文提出了一种基于模型迁移的方法来实现矿井的信道估计。该算法首先利用仿真平台生成贴近实际矿井信号的数据集，对神经网络模型进行预先训练以获得基准模型。随后，通过引入少量真实数据进行微调训练，实现对矿井信道的估计。为了验证算法的性能，本文对不同的特征提取网络、导频数量、调制信号、循环前缀以及信道多径多个条件进行了仿真，信道估计精度最优可达10^-5。结果表明，该算法不仅提高了信道估计精度和二次训练效率，同时增强了模型泛化能力。

Due to the complex and changeable environment of mines, multipath fading often occurs in wireless channels. Orthogonal Frequency Division Multiplexing technology has been applied in mine wireless transmission systems due to its advantages in resisting frequency-selective fading and overcoming inter-symbol interference. In mine OFDM transmission systems, the harsh underground environment makes it difficult to accurately estimate channel characteristics, which greatly limits the transmission quality of mine wireless communication. Facing the problems existing in mine communication systems, deep learning technology provides a new solution. Based on mine OFDM systems, this thesis conducts research on channel estimation algorithms based on deep learning. The main work includes the following aspects:

(1) Given the harsh conditions in underground coal mines and the low accuracy of traditional channel estimation algorithms, this thesis proposes an improved Super Resolution Convolutional Network for channel estimation. In the improved SRCNN model, the estimates at the pilot locations are used as input, with the improved SRCNN replacing the interpolation process in traditional channel estimation algorithms. Additionally, an efficient channel attention mechanism is incorporated to enhance the learning of channel features, thereby achieving accurate channel estimation. Through autonomous learning of training data to extract data patterns, the proposed improved SRCNN channel estimation algorithm can achieve an accuracy of 10^-4. Based on this, simulations were conducted under various conditions, including different channel estimation algorithms, different numbers of pilots, different modulation schemes, different attention mechanisms, and varying training signal-to-noise ratios. The simulation results demonstrate that, compared to traditional channel estimation algorithms, the proposed algorithm exhibits superior performance in channel estimation.

(2) Considering the complex and variable communication scenarios in actual mines, where real mine signal data is difficult to collect, this thesis proposes a model transfer-based approach for channel estimation in mines. This algorithm initially utilizes a simulation platform to generate a dataset that closely resembles real mine signals, and pre-trains a neural network model to obtain a baseline model. Subsequently, a small amount of real data is introduced for fine-tuning, achieving accurate channel estimation for the mine environment. To validate the performance of the algorithm, simulations were conducted under various conditions, including different feature extraction networks, numbers of pilots, modulation schemes, cyclic prefixes, and channel multipath settings. The optimal channel estimation accuracy achieved was as high as 10^-5. The results indicate that this algorithm not only improves channel estimation accuracy and secondary training efficiency but also enhances the generalization capability of the model.

[1]Li M, Li D, Xu C, et al. Band-notched dual-polarized antennas for 2G/3G/4G/5G base station using the GCPW DGS-DMS Hybrid Filter[J]. IEEE Access, 2024. 12: 37069-37079.

[2]张华, 隋敬麒, 武鹏飞, 等. 5G无线通信场景需求与技术演进分析[J]. 长江信息通信, 2021, 34(07): 197-199.

[3]赵冠一. 常态化保供形势下我国煤矿安全生产主要问题及重点省份措施建议[J]. 煤炭经济研究, 2024, 44(02): 175-180.

[4]郭明, 姜迁迁. 煤矿安全生产标准化管理模式研究[J]. 内蒙古煤炭经济, 2023, (21): 92-94.

[5]《煤矿安全生产条例》公布[J]. 中国煤炭工业, 2024, (03): 34.

[6]王浩, 成玮. “十三五”期间中国煤矿安全生产状况研究[J]. 能源与节能, 2023, (03): 53-55+79.

[7]崔丽珍, 李丹阳, 王巧利, 等.煤矿井下基于射线跟踪法的信道建模研究[J]. 中国矿业, 2019, 28 (08): 94-98.

[8]Zhang X, Sarris C D. Statistical modeling of electromagnetic wave propagation in tunnels with rough walls using the vector parabolic equation method[J]. IEEE Transactions on Antennas and Propagation, 2019, 67(4): 2645-2654.

[9]杜洋. 无线电波传播特性与信道建模研究现状及发展趋势[J]. 大众科技, 2021, 23(10): 11-14.

[10]姚述福, 余伟健. 矿井空巷道无线信道特性分析与仿真[J]. 矿业工程研究, 2020, 35(02): 73-78.

[11]Akyildiz I F, Sun Z, Vuran M C. Signal propagation techniques for wireless underground communication networks[J]. Physical Communication, 2009, 2(3): 167-183.

[12]Liu G. Data collection in MI-assisted wireless powered underground sensor networks: Directions, recent advances, and challenges[J]. IEEE Communications Magazine, 2021, 59(4): 132-138.

[13]Lin S C, Akyildiz I F, Wang P, et al. Distributed cross-layer protocol design for magnetic induction communication in wireless underground sensor networks[J]. IEEE Transactions on Wireless Communications, 2015, 14(7): 4006-4019.

[14]Liu G. A Q-learning-based distributed routing protocol for frequency-switchable magnetic induction-based wireless underground sensor networks[J]. Future Generation Computer Systems, 2023, 139: 253-266.

[15]Abboud M K, Sabbar B M. Performance evaluation of high mobility OFDM channel estimation techniques[J]. International Journal of Electrical and Computer Engineering, 2020, 10(3): 2562.

[16]王义元, 常俊, 卢中奎, 等. 深度学习辅助的5G OFDM系统的信道估计[J]. 电讯技术, 2024, 64(01): 36-42.

[17]Mishra H B, Singh P, Prasad A K, et al. OTFS channel estimation and data detection designs with superimposed pilots[J]. IEEE transactions on wireless communications, 2021, 21(4): 2258-2274.

[18]He X, Zhou R, Chen N, et al. Segment training based individual channel estimation for multi-pair two-way relay network with power allocation[J]. KSII Transactions on Internet and Information Systems (TIIS), 2018, 12(2): 566-578.

[19]何世彪, 李雪, 杨植景, 等. 面向高速移动场景的OFDM系统信道估计综述[J]. 重庆邮电大学学报(自然科学版), 2023, 35(04): 584-595.

[20]Chang R W. Orthogonal Frequency Multiplex Data Transmission System, US, 1970.

[21]高尚蕾, 张治中, 段浴, 等. 5G系统中基于解调参考信号的信道估计方法[J]. 电讯技术, 2021, 61(02): 191-196.

[22]Nejah M ,Ines K ,Mohamed S .Blind and semi-blind channel estimation and collision resolution for the uplink of MU-MIMO and massive MIMO systems for B5G networks[J]. Wireless Personal Communications, 2024, 134(2): 881-899.

[23]史清林, 刘丽哲, 李行健. 基于大规模MIMO的散射信道估计技术[J]. 计算机与现代化, 2022, (12): 18-25.

[24]尚斐, 鄢化彪. 5G系统PDSCH几种典型信道估计算法对比[J]. 山西电子技术, 2023, (01): 76-78.

[25]Luengo D, Martino L, Bugallo M, et al. A survey of Monte Carlo methods for parameter estimation[J]. EURASIP Journal on Advances in Signal Processing, 2020, 2020: 1-62.

[26]王晓君, 李笑添. 一种基于子空间分析的快速信源个数估计方法[J]. 太赫兹科学与电子信息学报, 2022, 20(04): 366-371.

[27]Abdallah S, Darya A M. Semi-blind channel estimation for diffusive molecular communication[J]. IEEE Communications Letters, 2020, 24(11): 2503-2507.

[28]Zakharov O V, Laptev A G, Lysenko V G, et al. Discrete algorithm for a disk morphological filter based on piecewise linear interpolation[J]. Measurement Techniques, 2022, 65(8): 577-583.

[29]Sun Y, Yang T, Liu Z. A whale optimization algorithm based on quadratic interpolation for high-dimensional global optimization problems[J]. Applied Soft Computing, 2019, 85: 105744.

[30]卫鑫, 熊威, 范伟伦. 三次样条插值在惯导数据处理中的应用[J]. 舰船电子工程, 2023, 43(08): 94-97.

[31]Yang J, Wen C K, Jin S, et al. Beamspace channel estimation in mmWave systems via cosparse image reconstruction technique[J]. IEEE Transactions on Communications, 2018, 66(10): 4767-4782.

[32]Soltani M, Pourahmadi V, Mirzaei A, et al. Deep learning-based channel estimation[J]. IEEE Communications Letters, 2019, 23(4): 652-655.

[33]邵凯, 陈连成, 刘胤. 高移动性Jakes信道的学习与估计[J]. 系统工程与电子技术, 2021, 43(04): 1119-1125.

[34]Wang M, Wang A, Liu Z, et al. Deep learning based channel estimation method for mine OFDM system[J]. Scientific Reports, 2023, 13(1): 17105.

[35]Mehrabi M, Mohammadkarimi M, Ardakani M, et al. A deep learning based channel estimation for high mobility vehicular communications[C]//2020 International Conference on Computing, Networking and Communications (ICNC). IEEE, 2020: 338-342.

[36]Ma X, Ye H, Li Y. Learning assisted estimation for time-varying channels[C]//2018 15th international symposium on wireless communication systems (ISWCS). IEEE, 2018: 1-5.

[37]Peng Q, Li J, Shi H. Deep learning based channel estimation for OFDM systems with doubly selective channel[J]. IEEE Communications Letters, 2022, 26(9): 2067-2071.

[38]Bai Q, Wang J, Zhang Y, et al. Deep learning-based channel estimation algorithm over time selective fading channels[J]. IEEE Transactions on Cognitive Communications and Networking, 2019, 6(1): 125-134.

[39]廖勇, 花远肖, 姚海梅, 等. 高速移动环境下基于深度学习的信道估计方法[J]. 电子学报, 2019, 47(08): 1701-1707.

[40]Mehrabi M, Mohammadkarimi M, Ardakani M, et al. A deep learning based channel estimation for high mobility vehicular communications[C]//2020 International Conference on Computing, Networking and Communications (ICNC). IEEE, 2020: 338-342.

[41]Gizzini A K, Chafii M, Nimr A, et al. Deep learning based channel estimation schemes for IEEE 802.11 p standard[J]. IEEE Access, 2020, 8: 113751-113765.

[42]Huang X, Liu S. Massive mimo channel estimation for vehicular communications: A deep learning based approach[C]//2020 IEEE International Conference on Communications Workshops (ICC Workshops). IEEE, 2020: 1-6.

[43]Yang J, Li H. Deep learning based detection and channel tracking for MIMO systems[C]//2020 5th International Conference on Computer and Communication Systems (ICCCS). IEEE, 2020: 635-639.

[44]王安义, 梁艳. 基于改进图像超分辨卷积网络的矿井OFDM信道估计研究[J]. 煤矿安全, 2024, 55(02): 211-217.

[45]金龙, 吴游, 张泳翔. 基于改进SRGAN的OFDM信道估计方法[J]. 计算机与现代化, 2021, (10): 112-118.

[46]杨舒心. 人工智能赋能经济高质量发展研究[J]. 合作经济与科技, 2024, (15): 17-19.

[47]史加荣, 王丹, 尚凡华, 等. 随机梯度下降算法研究进展[J]. 自动化学报, 2021, 47(09): 2103-2119.

[48]姜志侠, 宋佳帅, 刘宇宁. 一种改进的自适应动量梯度下降算法[J]. 华中科技大学学报(自然科学版), 2023, 51(05): 137-143.

[49]陈 爽, 张长伦, 黎铭亮. 深度学习中的优化算法研究[J]. 人工智能与机器人研究, 2022, 11(4): 448-462.

[50]Liu Y, Gao Y, Yin W. An improved analysis of stochastic gradient descent with momentum[J]. Advances in Neural Information Processing Systems, 2020, 33: 18261-18271.

[51]常禧龙, 梁琨, 李文涛. 深度学习优化器进展综述[J]. 计算机工程与应用, 2024, 60(07): 1-12.

[52]韩昌芝, 俞璐, 李林, 等. 无源无监督迁移学习综述[J]. 信息技术与信息化, 2023, (12): 46-51.

[53]赵颖超, 张菀, 岳新宇, 等. 基于AELSTM模型迁移学习的滚动轴承剩余寿命预测[J]. 国外电子测量技术, 2024, 43(02): 43-50.

[54]朱静茹, 张育芝, 王安义, 等. 基于Q-学习算法的矿井自适应OFDM调制研究[J]. 工矿自动化, 2021, 47(06): 109-115.