基于构造复杂程度定量评价的瓦斯含量预测BP模型  

BP model of gas content prediction based on quantitative assessment of geological structure complexity

作　　者：魏建平[1] 郝天轩[1] 刘明举[1] 

机构地区：[1]河南理工大学安全科学与工程学院,河南焦作454000

出　　处：《煤炭学报》2009年第8期1090-1094,共5页Journal of China Coal Society

基　　金：国家自然科学基金资助项目(50504008);国家重点基础研究发展计划"973"课题资助项目(2005CB221501);教育部新世纪优秀人才支持计划资助项目(NECT-07-0257);教育部长江学者和创新团队发展计划资助项目(IRT0618)

摘　　要：首先分析总结了构造复杂程度的3个定量评价指标——Kd(断层复杂程度系数)、Kz(褶皱复杂程度系数)和Kq(倾角复杂程度系数);然后,在分析潘三煤矿瓦斯地质特征的基础上,把Kd,Kz,Kq,煤层埋深和基岩厚度作为影响该矿煤层瓦斯含量的因素,建立了潘三煤矿的瓦斯含量预测BP神经网络模型;对建立的模型进行学习训练,经5 470次迭代,模型收敛,其精度高于多元回归模型,说明利用构造复杂程度定量评价系数来预测瓦斯瓦斯含量是可行的.Three kinds of quantitative assessment indexes were analyzed and summarized, which were Kd （repre- senting the fault structure complexity）, Kz （representing the fold structure complexity）, Kq （representing the inclined angle structure complexity）. Then, the BP neural model of Pansan Mine＇ gas content prediction model was found based on analysis its gas-geological characteristic by selecting Kd, Kz, Kq, buried depth and bedrock depth as affecting factors. The BP neural model was convergent by learning and training of 5 470 repetitions and the mod- el precision was greatly higher than muti-variable regression model, which shows that using quantitative assessment coefficient of the geological structure complexity to prediction gas content is feasible.

关 键 词：构造复杂程度 定量评价 瓦斯含量预测 BP神经网络 

分 类 号：TD712.3[矿业工程—矿井通风与安全]