基于粗糙集及最小二乘支持向量机的煤层厚度预测  

Forecasting of coal thickness based on rough set and LS-SVM

作　　者：陈优阔[1] 杨永国[1] 张鑫[1] 张辉[1] 

机构地区：[1]中国矿业大学资源与地球科学学院,徐州221008

出　　处：《地球物理学进展》2015年第5期2136-2141,共6页Progress in Geophysics

基　　金：国家自然科学基金(41372342);国家科技重大专项(2011ZX05034-005)联合资助

摘　　要：煤层厚度是煤矿设计与开采必不可少的数据,准确地预测煤层厚度,能够给煤矿生产提供有力的地质保障,煤层在地震勘探中属于薄层,其薄层厚度预测一直是公认的难题之一,传统的预测方法是利用钻孔资料的内插对比获得,精度比较低.本文提出了基于三维地震属性数据的粗糙集(RS)-最小二乘支持向量机(LS-SVM)算法模型,用于预测煤层厚度.利用粗糙集对地震属性数据所包含的大量干扰数据进行简约,减少样本维数,将简约后属性数据作为LS-SVM的输入预测煤层厚度.并运用PSO算法优化获得核函数的核参数及最佳正则化参数.实际钻孔数据试验验证了算法模型的可行性,并对整个研究区进行了煤层厚度预测,取得了较好的效果,最后探讨了VTK支持下的煤层可视化技术,对煤层实现了三维展示,达到了预期效果.The thinkness of coal seam is essentia for the design and exploition of coal mine.Accurate prediction of coal seam thinckness provides a powerful geological guarantee for the mine production.The coal seam in seismic exploration belongs to the thin layer,while the prediction of layer thickness has long been recognized as one of challenges.The traditional forecasting method is to use the data on the comparison of interpolation drilling,but with poor precision. This paper presents a Rough Set（RS）and LS-SVM Algorithm Model based on 3DSeismic Attribute data,for the prediction of coal thickness,using the Rough Set to modify mass disturbed date in seismic attributes data to reduce dimension of samples and to put modified attribute date into LS-SVM to predict the coal thickness as well as using the PSO algorithm to obtain the optimal kernel parameter and the most regularized parameter of the Kernel Function. Tests on Practical-borehole-data can verify the feasibility of the Algorithm Model and predicte the coal thickness around the whole area under study. This Algorithm Model has showed good effect.Finally, the technology on coal seam visualization based on VTK is carefully discussed,with 3Ddisplay of coal seam realized to have reached the desired effect.

关 键 词：粗糙集 最小二乘支持向量机 三维地震属性数据 煤层厚度预测 三维可视化 

分 类 号：P631[天文地球—地质矿产勘探]