目前，冲击地压灾害对矿井安全构成严重威胁。为此，本文提出了一种PSO-RF模型预测方法，旨在预测未来时刻的冲击地压危险性级别，为煤矿安全监管提供决策依据。

    为了预测未来时刻冲击地压的危险性等级，首先，通过相关系数热力图结合LightGBM特征选择算法，从多个冲击地压监测系统的指标中筛选出重要性较高的特征量，解决了特征选择过程中的主观性问题，确定冲击地压危险性分级预测的特征集合。其次，建立了基于RNN、LSTM、GRU和TCN的四种冲击地压危险性特征量预测算法，并通过对比MAE和RMSE评价指标选择出最优算法TCN；此外，利用相空间重构法对算法进行优化，预测t+1时刻冲击地压危险性特征量的值；最后，建立了RF、SVM和KNN冲击地压危险性分级预测算法，以准确率、精确率、召回率、F1值为评价指标，选择出最优算法RF；利用PSO优化算法对RF参数进行调优，基于t+1时刻冲击地压危险性特征量的值，实现t+1时刻冲击地压的危险性分级预测。

    实验结果表明：经过LightGBM特征选择后，只保留了与冲击危险关联较大的特征；通过对比四种危险性特征量预测算法的误差指标，发现冲击地压危险性特征量TCN预测算法具有 最小的RMSE均值和MAE均值，优于RNN、LSTM和GRU约0.1~2.5%；结合相空间重构法对TCN冲击地压危险性特征量预测算法进行优化后，经过优化后，PSR-TCN的预测效果进一步提高，其RMSE降低了30%，MAE降低了25%，整体预测效果提高了27.5%；此外，RF冲击地压危险性分级预测算法accuracy为0.962，高于KNN、SVM约20%，并经PSO算法进行参数调优后 ，算法准确度进一步提升了1.62%。

    综上所述，通过本文提出的PSO-RF模型在冲击地压危险性分级预测方面具有较高的准确性；为冲击地压监测预警提供重要参考，对煤矿安全监管具有实际应用价值。

    Currently, rockburst disasters pose a serious threat to mine safety. To address this issue, this paper proposes a PSO-RF model prediction method aimed at forecasting the level of rockburst hazard in future time intervals, providing decision-making support for coal mine safety regulation..

    To predict the hazard level of rockburst in future time intervals, firstly, a combination of correlation coefficient heatmaps and the LightGBM feature selection algorithm is employed to identify the most important features from multiple indicators of rockburst monitoring systems. This approach resolves the subjectivity issue in the feature selection process and determines the feature set for rockburst hazard classification prediction. Secondly, four prediction algorithms for rockburst hazard features, namely RNN, LSTM, GRU, and TCN, are established. By comparing the evaluation metrics of MAE and RMSE, the optimal algorithm TCN is selected. Furthermore, the algorithm is optimized using phase space reconstruction to predict the value of rockburst hazard features at time t+1. Finally, RF, SVM, and KNN algorithms are developed for rockburst hazard classification prediction. The selection of the optimal algorithm RF is based on accuracy, precision, recall, and F1 score. The parameters of RF are fine-tuned using the PSO optimization algorithm, enabling the prediction of rockburst hazard classification at time t+1 based on the value of rockburst hazard features at that time.

    Experimental results demonstrate that after the LightGBM feature selection, only features highly correlated with rockburst risk are retained. By comparing the error metrics of the four hazard feature prediction algorithms, it is observed that the TCN algorithm for rockburst hazard features exhibits the lowest mean RMSE and MAE, surpassing RNN, LSTM, and GRU by approximately 0.1% to 2.5%. After optimizing the TCN algorithm using phase space reconstruction, the prediction performance of PSR-TCN is further improved, with a 30% reduction in RMSE and a 25% decrease in MAE, resulting in an overall prediction improvement of 27.5%. Additionally, the accuracy of the RF algorithm for rockburst hazard classification prediction is 0.962, which is 20% higher than that of KNN and SVM. After parameter tuning using the PSO algorithm, the accuracy of the algorithm is further enhanced by 1.62%.

    In conclusion, the proposed PSO-RF model demonstrates high accuracy in rockburst hazard classification prediction. It provides important references for rockburst monitoring and early warning, and holds practical value for coal mine safety regulation.

[1].张建民,李全生,张勇,曹志国,王新中.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].陈结,高靖宽,蒲源源,姜德义,齐庆新,文志杰,孙启龙,陈林林.冲击地压预测预警的机器学习方法[J].采矿与岩层控制工程学报,2021,3(01):57-68．

[3].何学秋,陈建强,宋大钊,何生全,李振雷,钟涛平,薛雅荣,周超.典型近直立煤层群冲击地压机理及监测预警研究[J].煤炭科学技术,2021,49(06):13-22.

[4].李鹏,蔡美峰,郭奇峰,鲁细根,颜丙乾.煤矿断层错动型冲击地压研究现状与发展趋势[J].哈尔滨工业大学学报,2018,50(03):1-17.

[5].施天威,潘一山,王爱文,代连朋.基于能量贮存及释放主体的煤矿冲击地压分类[J].煤炭学报,2020,45(02):524-532.

[6].夏永学,冯美华,李浩荡.冲击地压地球物理监测方法研究[J].煤炭科学技术,2018,46(12):54-60.

[7].徐景果,朱东旭,王桂峰,陈星,曹安业,刘阳.冲击地压智能应力感知和精准卸压调控技术[J].煤炭科技,2022,43(05):137-141.

[8].李斌.煤矿冲击地压危险监测预警技术及应用[J].矿业装备,2022(04):24-26.

[9].Cao Jinrong,Dou Linming,He Jiang,Zhu Guangan,Wang Zhengyi,Bai Jinzheng,Han Zepeng. Failure mechanism and control of coal bursts triggered by mining induced seismicity in steeply inclined and extra thick coal seam[J]. Frontiers in Earth Science,2023,82(7):1042539.

[10].Gao Hou,Zhao Wusheng,Chen Weizhong,Xie Peiyao,Zhong Kun,Qin Changkun. Continuous three-dimensional stress monitoring in roof of coal mines for investigating the rockburst control effect with hydraulic fracturing[J]. Environmental Earth Sciences,2022,81(17):1-11.

[11].Li Xuelong,Chen Shaojie,Wang Enyuan,Li Zhonghui. Rockburst mechanism in coal rock with structural surface and the microseismic (MS) and electromagnetic radiation (EMR) response[J]. Engineering Failure Analysis,2021,124(3):105396.

[12].Di Yangyang, Wang Enyuan, Rock Burst Precursor Electromagnetic Radiation Signal Recognition Method and Early Warning Application Based on Recurrent Neural Networks, Rock Mechanics and Rock Engineering, 2021,54:1449-1461.

[13].Nosov Viktor V.,Borovkov Alexey I.,Artyushchenko Artem P.. Predicting Rock Bursts in Rock Mass Blocks Using Acoustic Emission[J]. Resources,2022,11(10):87.

[14].Tian-Hui Ma,Chun-An Tang,Shi-Bin Tang,Liang Kuang,Qun Yu,De-Qing Kong,Xu Zhu.Rockburst mechanism and prediction based on microseismic monitoring[J].International Journal of Rock Mechanics and Mining Sciences,2018,110(110):177-188.

[15].G.K. Ghosh,C. Sivakumar.Application of underground microseismic monitoring for ground failure and secure longwall coal mining operation: A case study in an Indian mine[J].Journal of Applied Geophysics,2018,150:21-39.

[16].赵德明,庞锐,王海波.微震震源定位算法优化研究及应用[J].计算机技术与发展,2021,31(07):209-214.

[17].夏永学,蓝航,魏向志.基于微震和地音监测的冲击危险性综合评价技术研究[J].煤炭学报,2011,36(S2):358-364.

[18].谭云亮,张明,徐强,郭伟耀,于凤海,顾士坦.坚硬顶板型冲击地压发生机理及监测预警研究[J].煤炭科学技术,2019,47(01):166-172.

[19].Liu, S., Li, S., Wang, J., & Zhang, Y. (2019). Study on the prediction model of rockburst based on the chaotic theory of mine pressure signals. Journal of Mining and Safety Engineering, 36(5):787-794.

[20].崔峰,何仕凤,来兴平等.基于相空间重构与深度学习的冲击地压矿井时间序列b值趋势研究[J/OL].煤炭学报:1-13[2023-03-12].

[21].秦忠诚,曹斌,吴清法.基于盲数理论改进未确知测度的冲击地压评价模型应用研究[J].煤矿安全,2019,50(07):224-227.

[22].张凯,顾士坦.不同采掘扰动下的冲击地压危险ANN评价方法[J].煤矿安全,2017,48(03):182-185.

[23].赵永芳,张凌云,于丽雅.SPA-ITFN模型在冲击地压评价中的应用[J].矿业研究与开发,2019,39(11):37-42.

[24].袁亮.煤矿典型动力灾害风险判识及监控预警技术研究进展[J].煤炭学报,2020,45(05):1557-1566.

[25].张广辉,邓志刚,蒋军军,安雨佳.KJ768微震监测系统在高瓦斯矿井强矿压灾害预警中的应用[J].煤矿安全,2020,51(09):144-147.

[26].杨建坤.郭屯煤矿冲击地压综合防治措施研究与应用[J].矿山测量,2020,48(04):25-30.

[27].张修峰,曲效成,魏全德.冲击地压多维度多参量监控预警平台开发与应用[J].采矿与岩层控制工程学报,2021,3(01):69-78.

[28].左凌云.基于多因素分析的孤岛工作面冲击地压危险性预测[J].煤炭工程,2019,51(12):96-100.

[29].曾繁慧,张晶,汪北方等.煤矿冲击地压危险性模糊综合评价[J].辽宁工程技术大学学报(自然科学版),2018,37(01):205-209.

[30].李乃录,屈振祖,张恒.基于多因素耦合分析的冲击地压卸压技术研究[J].山东煤炭科技,2021,39(03):156-158+162.

[31].徐隽松,潘鹏志,陈建强等.基于地应力反演的褶曲区煤层冲击危险性评价研究[J/OL].煤炭科学技术:1-11[2023-03-12].

[32].金佩剑,王恩元,刘晓斐,黄宁,王嗣衡.冲击地压危险性综合评价的突变级数法研究[J].采矿与安全工程学报,2013,30(02):256-261.

[33].朱峰,张宏伟.基于“AHP +熵权法”的CW-TOPSIS冲击地压评判模型[J].中国安全科学学报,2017,27 (1):128-133.

[34].Hui Liu,Xiaojun Zhang. Predictive analysis of impact hazard level of coal rock mass based on fuzzy inference network[J]. Journal of Intelligent & Fuzzy Systems,2020,38(2):1509-1518.

[35].张凯.基于神经网络分析的煤柱型冲击地压多参量综合预警研究[D].山东科技大学,2018.

[36].温廷新,陈晓宇,杨红玉,窦融,田煜晨.基于优化Bagging-LSSVM模型的冲击地压预测[J].中国安全科学学报,2017,27(06):121-126.

[37].刘辉.基于机器学习的煤矿冲击危险性综合预测方法研究[D].山东科技大学,2020.

[38].刘慧敏,徐方远,刘宝举,邓敏.基于CNN-LSTM的岩爆危险等级时序预测方法[J].中南大学学报(自然科学版),2021,52(03):659-670.

[39].张凯,顾士坦.不同采掘扰动下的冲击地压危险ANN评价方法[J].煤矿安全,2017,48(03):182-185.

[40].赵会茹,赵一航,郭森.基于互补集合经验模态分解和长短期记忆神经网络的短期电力负荷预测[J].中国电力,2020,53(06):48-55.

[41].彭月,蒙祖强,杨丽娜.一种基于GRU神经网络的语音增强方法[J].广西大学学报(自然科学版),2021,46(06):1533-1548.

[42].王万良,胡明志,张仁贡,董建杭,金雅文.改进TCN和LSTM的泸水河流域月径流量预测模型[J].计算机集成制造系统:1-29.

[43].Bai S, Kolter J Z, Koltun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J].arXiv preprint arXiv:1803.01271,2018.

[44].丁建立,孙玥.基于LightGBM的航班延误多分类预测[J].南京航空航天大学学报,2021,53(06):847-854.

[45].李占山,姚鑫,刘兆赓,张家晨.基于LightGBM的特征选择算法[J].东北大学学报(自然科学版),2021,42(12):1688-1695.

[46].Yaqing Zhao, Howard Bondell. Solution paths for the generalized lasso with applications to spatially varying coefficients regression[J]. Computational Statistics and Data Analysis,2020,12(02):142-156.

[47].Guyon l.Elisseff A.An introduction to variable and feature selection[J.Journal of Machine Learning Research,2003,3(6):1157-1182.

[48].Al-Abed Mohammad A, Al-Bashir Areen K, Al-Rawashdeh Ayman, Alex Raichel M,Zhang Rong, Watenpaugh Donald E,Behbehani Khosrow. Estimation of cerebral blood flow velocity during breath-hold challenge using artificial neural networks.[J]. Computers in biology and medicine,2019,3(08):115-128.

[49].Flesia Ana Georgina,Nieto Paula Sofia,Aon Miguel A,Kembro Jackelyn Melissa. Computational Approaches and Tools as Applied to the Study of Rhythms and Chaos in Biology.[J]. Methods in molecular biology (Clifton, N.J.),2022: 277-341.

[50].Chen Xi,Jin Xiaoling,Huang Zhilong. Data-driven method for identifying the expression of the Lyapunov exponent from random data[J]. International Journal of Non-Linear Mechanics,2023,148:104268.

[51].Li Jianlin,Wang Luyang,Wang Xinyi,Gao Peiqiang. Chaos-generalized regression neural network prediction model of mine water inflow[J]. SN Applied Sciences,2021,3(12):1-11.

[52].冯馨玉,李凯,任天锋,李汉辉,杨旸,周明拓.基于相空间重构方法的三维MIMO无线信道预测[J].中国科学院大学学报,2023,40(01):135-143.

[53].Feng Peng,Peng Feng,Wen Jie,Zhang Yao,Jin Junying. Monthly Streamflow Prediction Based on Random Forest Algorithm and Phase Space Reconstruction Theory[J]. Journal of physics. Conference series,2020,1637(1):012091.

[54].R Vignesh,V Jothiprakash,B Sivakumar. Streamflow variability and classification using false nearest neighbor method[J]. Journal of Hydrology, 2015,531:706-715.

[55].张叶,任鸿翔,王德龙.基于KNN算法的船舶操纵智能评估系统[J].上海海事大学学报,2021,42(04):33-38.

[56].张曼,陈建宏,周智勇.基于SVM的岩爆分级预测模型及R语言实现[J].中国地质灾害与防治学报,2018,29(04):64-69.

[57].Yari Mojtaba,Armaghani Danial Jahed,Maraveas Chrysanthos,Ejlali Alireza Nouri,Mohamad Edy Tonnizam,Asteris Panagiotis G.. Several Tree-Based Solutions for Predicting Flyrock Distance Due to Mine Blasting[J]. Applied Sciences,2023,13(3): 1345.

[58].梁燕华,毛诗允,李金峰.基于SSA优化BP神经网络的冲击地压数据预测研究[J].电子测试,2022,36(07):59-61.DOI:10.16520/j.cnki.1000-8519.2022.07.028.

[59].刘威军,范俊奇,李天斌等.深埋高地应力隧道勘察期岩爆烈度概率分级预测[J].水文地质工程地质,2022,49(06):114-123.DOI:10.16030/j.cnki.issn.1000-3665.202111027.

Liu Xiaojiang,Xu Zhao-Dong,Sun Bin,Liu Xuanya,Xu Dajun. Smart prediction for tunnel fire state evolution based on an improved fire simulation curve through particle swarm optimization algorithm[J]. Fire Safety Journal,2023,136:103763.