摘要
煤储层渗透率是影响煤层气开采的重要参数,为了有效的预测煤层渗透率,提出了多元逐步回归与支持向量回归结合的MSR-SVR预测模型。通过研究沁水盆地柿庄北地区测井数据,采用多元逐步回归法分析因变量渗透率与自变量孔隙度、声波时差、自然伽马、电阻率、自然电位之间的关系,优选出显著性自变量孔隙度和自然伽马,将其作为初始样本输入到支持向量回归中,以此建立模型对煤层进行渗透率预测。利用测试样本和验证样本对模型的有效性进行验证,其相对误差均在25%以内。MSR-SVR预测模型实例表明:沁水盆地柿庄北地区煤层的渗透率普遍偏低;MSR-SVR预测模型在地质条件复杂、渗透率规律性差的煤层中进行渗透率预测具有合理性和可行性。
The coal reservoir permeability is an important parameter impacting CBM exploitation. To effectively predict coal seam per- meability, a combined multiple stepwise regression and support vector regression （MSR-SVR） prediction model put forward. Using well logging data from the Shizhuang north area, Qinshui Basin, through MSR method analyzed relationship between dependent variable per- meability and independent variable porosity, sonic differential time, natural gamma ray, resistivity and natural potential. Then opti- mized significant independent variable porosity and natural gamma ray, input to SVR as initial samples and established a model to pre- dict coal seam permeability. Using testing samples and verifying samples carried out model effectiveness verification, their relative er- rors are all less than 25%. The MSR-SVR prediction model case studies have shown that： the one is coal seam permeability in Shi- zhuang north area generally on the low side; the second is MSR-SVR prediction model has rationality and feasibility in permeability prediction under coal seams with complicated geological condition and poor permeability regularity.
出处
《中国煤炭地质》
2016年第1期23-26,共4页
Coal Geology of China
基金
国家"十二五’重大专项:CO2注入后的运移监测和安全技术研究(2011ZX05042-003-002)
人才培养项目-引领学科(5111524100)
河北省重点实验开放基金(KJZH2014K04)
关键词
渗透率
测井参数
多元逐步回归
支持向量回归
permeability
well logging parameter
multiple stepwise regression
support vector regression
