顶板灾害是煤矿生产的主要灾害之一，矿压预测是顶板灾害预防的重要手段。目前，深度学习因其在灾害建模和预测方面的优势得到了研究者的广泛关注，但现有深度学习预测模型在预测准确性、泛化能力及长周期预测顶板压力变化趋势等方面还存在不足。因此本文研究基于深度学习的矿压预测模型对预测工作面矿压变化趋势具有重要意义。

首先，针对矿压预测的准确性和泛化能力不足的问题，提出一种改进的回声状态网络模型，即SSA-ESN的矿压预测方法。基于回声状态网络（Echo State Network，ESN），在输入层与储备池之间加入参数随机初始化的隐藏层，提高预测精度；同时使用樽海鞘（Salp Swarm Algorithm，SSA)算法优化ESN储备池参数，将优化后的最佳参数注入ESN网络，提高ESN的泛化能力。通过仿真实验，相较于Adam-LSTM、GA-BP和ESN模型，SSA-ESN模型的平均绝对误差分别降低了47.2%，28.5%和23.8%；均方误差分别降低了49.4%，17.1%和59.8%；平均绝对百分比误差分别降低了38.6%，17.1%和11.9%。构建另一组30116工作面数据集验证泛化性，SSA-ESN模型在该数据集上的误差评价指标均表现最优。实验结果表明，SSA-ESN模型预测精度提高，拥有良好的泛化性，能反映出来压最大值变化情况。

其次，针对矿压预测模型在长周期矿压预测中精度低的问题，提出一种基于SSA-ESN的时空矿压预测方法。利用傅里叶函数和灰色关联度方法对矿压序列进行周期分析和空间关联分析，建立时空数据集；使用多元回归算法融合时空预测模型，建立基于SSA-ESN矿压时空预测模型。通过仿真实验，相比于GA-BP、Adam-LSTM和SSA-ESN模型，时空SSA-ESN的平均绝对误差分别降低了28.4%，6.1%和20.2%；均方误差分别降低了47.2%，14.1%和32.7%；平均绝对百分比误差减少46.7%，15.9%和29.5%。实验结果表明，SSA-ESN时空预测模型在预测下一周期矿压数据实验中，提高长周期矿压预测的预测精度，具有良好的预测能力。

最后，结合来压步距和以上两种来压趋势预测模型，将该模型应用到陕西某煤矿30302工作面的矿压预测中，周期来压步距预测相对误差为0.115。测试结果显示，该方法能够较准确地预测工作面来压趋势变化情况，对工作面来压预测具有重要的参考意义。

Roof disaster is one of the main disasters in coal mine production, and mine pressure prediction is an important means of roof disaster prevention. Currently, deep learning has received extensive attention from researchers due to its advantages in disaster modeling and prediction. However, the existing deep learning prediction models still have shortcomings in prediction accuracy, generalization ability and long-period prediction of roof pressure change trends. Therefore, it is of great significance to study the prediction model of mine pressure based on deep learning for predicting the change trend of mine pressure in working face.

Firstly, aiming at the problem of insufficient accuracy and generalization ability of mine pressure prediction, an improved echo state network model, namely SSA-ESN, was proposed. Based on Echo State Network (ESN), a hidden layer with randomly initialized parameters is added between the input layer and the reservoir to improve the prediction accuracy. At the same time, the Salp Swarm Algorithm(SSA) is used to optimize the parameters of the reservoir, and the optimized best parameters are injected into the ESN network to improve the generalization of the network.Through simulation experiments, compared with Adam-LSTM, GA-BP and ESN models, the mean absolute error of the SSA-ESN model is reduced by 47.2%, 28.5% and 23.8% respectively. The mean square error is reduced by 49.4%, 17.1% and 59.8%, respectively. The mean absolute percentage error is reduced by 38.6%, 17.1% and 11.9%, respectively.Another 30116 face data set was constructed to verify the generalization. The error evaluation indexes of the SSA-ESN model on this data set were the best. Experimental results show that the prediction accuracy of the SSA-ESN model is improved, and it has good generalization, and can reflect the change of the maximum pressure value.

Secondly, aiming at the problem of low accuracy of the mine pressure prediction model in long-period mine pressure prediction, a spatio-temporal mine pressure prediction method based on SSA-ESN is proposed. The Fourier function and grey correlation degree method are used to carry out periodic analysis and spatial correlation analysis of the mine pressure sequence, and the spatio-temporal data set is established. The multiple regression algorithm is used to fuse the spatio-temporal prediction model, and the spatio-temporal prediction model of mine pressure based on SSA-ESN is established. Through simulation experiments, compared with GA-BP, Adam-LSTM and SSA-ESN models, the mean absolute error of spatio-temporal SSA-ESN is reduced by 28.4%, 6.1% and 20.2% respectively. The mean square error is reduced by 47.2%, 14.1% and 32.7%, respectively. The mean absolute percentage error decreases by 46.7%, 15.9% and 29.5%. The experimental results show that the SSA-ESN spatio-temporal prediction model improves the prediction accuracy of long-period mine pressure prediction in predicting the next period of mine pressure data experiment, and has good prediction ability.

Finally, combined with the pressure step and the above two pressure trend prediction models, the model was applied to the mine pressure prediction of the 30302 working face of a coal mine in Shaanxi ,and the relative error between the prediction and the actual value of the cycle pressure step distance was 0.115. The test results show that the method can accurately predict the change of the pressure trend of the working face, which has important reference significance for the prediction of the pressure.

[1]张建民, 李全生, 张勇, 等. 煤炭深部开采界定及采动响应分析[J]. 煤炭学报, 2019, 44(05): 1314-1325.

[2]徐刚, 黄志增, 范志忠, 等.工作面顶板灾害类型､监测与防治技术体系[J]. 煤炭科学技术, 2021, 49(02): 1-11.

[3]赵承国.采煤工作面矿压观测技术现状及改进研究[J].能源与节能,2024,(03):122-124.

[4]Molaei F, Rahimi E, Siavoshi H, et al. A comprehensive review on internet of things (IoT) and its implications in the mining industry[J]. American Journal of Engineering and Applied Sciences, 2020, 13(3): 499-515.

[5]Qi C. Big data management in the mining industry[J]. International Journal of Minerals, Metallurgy and Materials, 2020, 27(2): 131-139.

[6]郭奋超,拓宏平.煤矿智能化建设面临的问题与解决思路[J].智能矿山,2024,5(02):8-12.

[7]Fan C, Xu H, Wang G, et al. Determination of roof horizontal long drilling hole layout layer by dynamic porosity evolution law of coal and rock[J]. Powder Technology, 2021, 394: 970-985.

[8]王志红, 杨富强, 武国旺, 等.煤炭大数据研究及应用现状分析[J]. 煤炭技术, 2020, 39(11): 185-188.

[9]刘昆轮, 闫瑞兵. 基于地音监测的近直立煤层冲击地压前兆特征研究[J]. 煤炭工程, 2020, 52(04): 48-51.

[10]Li Y, Shi X. Mine Pressure Prediction Study Based on Fuzzy Cognitive Maps[J]. International Journal of Computational Intelligence and Applications, 2020, 19(03): 2050023.

[11]Zhou H, Huang Q, Liu Y, et al. Study on the Mine Pressure Law and the Pressure Frame Mechanism of an Overlying Goaf in a Shallow Coal Seam[J]. Shock and Vibration, 2021, 2021: 1-15.

[12]Zhang J, Li X, Qin Q, et al. Study on overlying strata movement patterns and mechanisms in super-large mining height stopes[J]. Bulletin of Engineering Geology and the Environment, 2023, 82(4): 142.

[13]Wang S, Wu X, Zhao Y, et al. Mechanical Performances of Pressure Arch in Thick Bedrock during Shallow Coal Mining[J]. Geofluids, 2018, 2018: 1-13.

[14]Han H, Hu Z, Liu Y, et al. Mining stress formation and distribution: Predictive model based on overburden key strata structure[J]. Energy Science & Engineering, 2024,12(4):1551-1568.

[15]Huo B J, Jing X D, Yu B, et al. A classification forecasting method for the weighting intensity of stopes of thick coal seams with hard roofs[J]. Chinese Journal of Rock Mechanics and Engineering, 2019, 38(9): 1828-1835.

[16]陈清. 煤矿联网数据分析模型设计与实现[J]. 煤矿安全, 2021, 52(08): 162-166.

[17]Wang Y, Liu M, Huang Y, et al. Knowledge-based and data-driven underground pressure forecasting based on graph structure learning[J]. International Journal of Machine Learning and Cybernetics, 2024, 15(1): 3-18.

[18]Chen Y, Liu C, Liu J, et al. Intelligent prediction for support resistance in working faces of coal mine: A research based on deep spatiotemporal sequence models[J]. Expert Systems with Applications, 2024, 238: 122020.

[19]崔海尚, 郭松, 郭广礼, 李怀展. 基于降维模糊C均值聚类算法的采动场地稳定性评价[J]. 煤炭科学技术, 2020, 48(10): 142-147.

[20]刘航舵. 基于循环神经网络的瓦斯灾害预测预警研究[D]. 西安科技大学, 2020.

[21]余琼芳, 王联港, 杨艺. 基于LSTM-TCN的综采工作面顶板压力预测[J]. 煤炭技术, 2023, 42(6): 5-9.

[22]Wang K, Zhuang X, Zhao X, et al. Roof pressure prediction in coal mine based on grey neural network[J]. IEEE access, 2020, 8: 117051-117061.

[23]Zhang M. Prediction of rockburst hazard based on particle swarm algorithm and neural network[J]. Neural Computing and Applications, 2022, 34(4): 2649-2659.

[24]Lan T, Guo X, Zhang Z, et al. Prediction of microseismic events in rock burst mines based on MEA-BP neural network[J]. Scientific Reports, 2023, 13(1): 9523.

[25]Bhandari T, Prasad Mantripragada VNS, Deb D, et al. Real-Time Shield Pressure Analysis for the Evaluation of Roof Behavior in Longwall Panels[J]. International Journal of Geomechanics, 2023, 23(8): 04023109.

[26]Gao T, Li Z, Li H. A Prediction Network for Hydraulic Support Pressure Based on Multitimescale Feature Extraction[J]. Mathematical Problems in Engineering, 2022, 2022: 1-9.

[27]Lai X, Tu Y, Yan B, et al. A Method for Predicting Ground Pressure in Meihua**g Coal Mine Based on Improved BP Neural Network by Immune Algorithm-Particle Swarm Optimization[J]. Processes, 2024, 12(1): 147.

[28]张帆, 李闯, 李昊, 等. 面向智能矿山与新工科的数字孪生技术研究[J]. 工矿自动化, 2020, 46(5): 6.

[29]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(6): 35-40.

[30]李化敏, 王伸, 李东印, 等. 煤矿采场智能岩层控制原理及方法[J]. 煤炭学报, 2019, 44(01): 134-147.

[31]高潮,田斌,李治祥,等.工作面矿压大数据动态分析与预测系统[J].煤矿机械,2024,45(02):193-196.

[32]顾小清, 李世瑾, 李睿. 人工智能创新应用的国际视野——美国NSF人工智能研究所的前瞻进展与未来教育展望[J]. 中国远程教育, 2021, 563(12): 1-9+76.

[33]顾清华, 江松, 李学现等. 人工智能背景下采矿系统工程发展现状与展望[J]. 金属矿山, 2022(5): 10-25.

[34]尹春雷, 魏文艳, 何勇华, 等.基于大数据分析与线性回归模型的工作面顶板压力研究[J]. 自动化应用, 2021(05): 46-50.

[35]吕文玉, 王海金, 伍永平, 等.基于PSO-SVR预测模型的综采工作面周期来压研究[J]. 煤炭工程, 2022, 54(04): 86-91.

[36]Dong J, Xie Z, Jiang H, et al. Multiple regression method for working face mining pressure prediction based on hydraulic support monitoring dataset[J]. Frontiers in Earth Science, 2023, 11: 1114033.

[37]Govil H, Chatterjee R S, Bhaumik P, et al. Deformation monitoring of Surakachhar underground coal mines of Korba, India using SAR interferometry[J]. Advances in Space Research, 2022, 70(12): 3905-3916.

[38]Gao X, Hu Y, Liu S, et al. An AGCRN Algorithm for Pressure Prediction in an Ultra-Long Mining Face in a Medium–Thick Coal Seam in the Northern Shaanxi Area, China[J]. Applied Sciences, 2023, 13(20): 11369.

[39]Chen Y, Liu C, Liu J, et al. Case Study on Dynamic Identification of Overburden Fracture and Strong Mine Pressure Mechanism of Isolated Working Face Based on Microseismic Clustering[J]. Sustainability, 2022, 15(1): 436.

[40]吕鹏飞, 邱林. 基于PSO-LSSVM的矿井冲击地压分级预测研究[J]. 矿业安全与环保, 2021, 48(01): 120-125.

[41]赵毅鑫, 杨志良, 马斌杰, 等.基于深度学习的大采高工作面矿压预测分析及模型泛化[J]. 煤炭学报, 2020, 45(01): 54-65.

[42]Lu J,Liu Z,Zhang W,et al. Pressure Prediction Study of Coal Mining Working Face based on Adam-LSTM[J]. IEEE Access, 2023, 11: 83867-83880.

[43]卢振龙, 徐刚, 尹希文, 等.煤矿智能化顶板矿压预警技术研究[J]. 煤炭工程, 2023, 55(12): 22-27.

[44]Tan T, Yang Z, Chang F, et al. Prediction of the first weighting from the working face roof in a coal mine based on a GA-BP neural network[J]. Applied Sciences, 2019, 9(19): 4159.

[45]Zhu B, Shi Y, Hao J, et al. Prediction of Coal Mine Pressure Hazard Based on Logistic Regression and Adagrad Algorithm—A Case Study of C Coal Mine[J]. Applied Sciences, 2023, 13(22): 12227.

[46]Rong H, Cheng T. Research and control of underground pressure law of coal mining face in soft rock strata based on neural network under the background of wireless network communication[J]. Mobile Information Systems, 2021, 2021: 1-14.

[47]Hu H, Li Y, Yao Q, et al. A Prediction Approach Based on Clustering Reconstruction for Abnormal Mining Pressure of Longwall Face under Residual Coal Pillars[J]. Processes, 2024, 12(2): 283.

[48]Zhou M, Shi Y, Hao J, et al. Automatic Acquisition System for Mine Pressure Monitoring in Coal Mine Working-Face Footage[J]. Journal of Sensors, 2024, 2024:1-14

[49]Zhang Y, Yu Q, Tang G, et al. Multi-step prediction of roof pressure based on multi-scale contextual fusion network[J]. Sensors and Actuators A: Physical, 2024: 115130

[50]王慧雯, 薛晔, 薛崇义. 耦合效应下的煤矿顶板事故风险因素评价研究[J]. 矿业安全与环保, 2021, 48(01): 115-119.

[51]Wang X, Zhang L. A multicriteria decision model based on analytic hierarchy process for managing safety in coal mines[J]. Geofluids, 2022, 2022: 1-9.

[52]Liu Q, Zhao P, Zhang Y, et al. Evaluation and empirical research on the operation effect of the risk precontrol management system of safety in coal mines based on OSWM-TOPSIS[J]. Work, 2021, 70(3): 795-804.

[53]赵怀志, 王晓东. 顶板来压作用下液压支架安全阀特性试验研究[J]. 煤炭技术, 2022, 41(05):222-225.

[54]郝秦霞, 汪连连, 张金锁. 基于R2-MOEA/D算法的顶板支护决策模型[J]. 工矿自动化, 2021, 47(10): 77-84.

[55]李金波, 王素玲, 董康兴, 等.基于数值模拟的塔山煤矿近场顶板来压及控制规律研究[J]. 力学与实践, 2020, 42(05): 580-587.

[56]Zen'Kov I V,Izhmulkina E,Maglinets Y A ,et al.Results of the Study of Ecosystem Formation in Coal Mines in the Western Part of the Central Regions of Kuzbass Using Remote Sensing Resources[J]. Ecology and Industry of Russia, 2018, 22: 40-45.

[57]经来旺, 李树文, 彭绍驰, 等.高水平应力下矩形巷道顶板形状优化[J]. 煤炭技术, 2022, 41(07): 42-46.

[58]Zekić-Sušac M, Has A, Knežević M. Predicting energy cost of public buildings by artificial neural networks, CART, and random forest[J]. Neurocomputing, 2021, 439: 223-233.

[59]Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural computation, 1997, 9(8): 1735-1780.

[60]Jaeger H, Haas H. Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication[J]. Science, 2004, 304(5667): 78-80.

Mirjalili S, Gandomi A H, Mirjalili S Z, et al. Salp Swarm Algorithm: A bio-inspired optimizer for engineering design problems[J]. Advances in engineering software, 2017, 114: 163-191.