基于DBSCAN-IHHO-SVM模型的煤与瓦斯突出预测  

Coal and Gas Outburst Prediction Based on DBSCAN-IHHO-SVM Model

作　　者：郑晓亮[1,2,3] 王琦[1] 来文豪 张贺 张玉婷 ZHENG Xiaoliang;WANG Qi;LAI Wenhao;ZHANG He;ZHANG Yuting(School of Safety Science and Engineering,Anhui University of Science and Technology,Huainan 232001,China;School of Public Safety and Emergency Management,Anhui University of Science and Technology,Hefei 231131,China;School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学安全科学与工程学院,安徽淮南232001 [2]安徽理工大学公共安全与应急管理学院,合肥231131 [3]安徽理工大学电气与信息工程学院,安徽淮南232001

出　　处：《湖北民族大学学报(自然科学版)》2025年第1期53-59,共7页Journal of Hubei Minzu University:Natural Science Edition

基　　金：“十四五”重点研发计划资助项目(2023YFB321103);煤炭安全精准开采国家地方联合工程研究中心开放基金资助项目(EC2021003)。

摘　　要：针对煤与瓦斯突出事故的复杂性以及数据获取困难导致预测准确率低的问题,提出基于密度的噪声应用空间聚类-改进哈里斯鹰优化-支持向量机(density based spatial clustering of applications with noise-improved Harris hawks optimization-support vector machine, DBSCAN-IHHO-SVM)预测模型。首先,选取瓦斯含量、瓦斯压力、煤层孔隙率、煤层坚固性系数作为预测指标,对数据中的缺失值采用均值填补处理,利用生成式对抗网络(generative adversarial network, GAN)扩充突出数据量。接着,采用DBSCAN从非突出数据中识别潜在危险数据,并将其作为新的突出数据。最后,引入IHHO调整SVM模型参数,将处理后的数据输入IHHO-SVM模型进行预测分析。结果表明,相比于原始SVM模型,DBSCAN-IHHO-SVM模型的整体预测准确率、危险数据识别率分别提升了5.87%、38.46%。在突出数据样本有限的情况下,DBSCAN-IHHO-SVM模型能有效挖掘非突出数据潜在信息,实现精准预警,为该领域研究提供了新思路。The complexity of coal and gas outburst accidents and the low prediction accuracy caused by the difficulty of data acquisition were addressed by proposing the density-based spatial clustering of applications with noise-improved Harris hawks optimization-support vector machine(DBSCAN-IHHO-SVM)warning model.Firstly,gas content,gas pressure,coal seam porosity,and the coal seam robustness coefficient were selected as predictors,and missing values in the data were processed by mean filling.The amount of outburst data was expanded using a generative adversarial network(GAN).Secondly,DBSCAN was employed to identify potentially hazardous data from non-outburst data,which were then treated as new outburst data.Finally,the parameters of the SVM model adjusted by IHHO were introduced,and the processed data were fed into the IHHO-SVM model for predictive analysis.Compared with the original SVM model,the results showed that the overall prediction accuracy and hazardous data identification rate of DBSCAN-IHHO-SVM model were improved by 5.87%and 38.46%,respectively.When faced with limited outburst data samples,DBSCAN-IHHO-SVM model effectively mined the potential information of non-outburst data,achieving accurate early warning and offering new insights for research in this field.

关 键 词：煤与瓦斯突出 预测 危险数据识别 数据扩充 IHHO SVM 

分 类 号：TD713[矿业工程—矿井通风与安全]