【作者】 颜爱华；

【导师】 何学秋； 聂百胜； 王恩营；

【作者基本信息】 中国矿业大学（北京） ， 安全技术及工程， 2010， 博士

【摘要】 采用瓦斯地质理论和方法研究地质构造、煤层煤质、围岩岩性与结构、煤层埋深等地质因素对煤层瓦斯含量的影响,筛选出瓦斯含量主导控制因素。以告成煤矿为例,确定煤层基岩埋深是控制该矿煤层瓦斯含量的主要因素,其次是煤层顶板50m岩层效应厚度影响系数、含砂率和煤层厚度,而新生界地层厚度和煤质(包括水分、灰份)对瓦斯含量的影响较小。利用多元线性回归、BP神经网络、支持向量机等理论模型建立煤矿煤层瓦斯含量预测模型,并进行了精度比较。研究分析矿井勘探期间和生产期间实测的煤层瓦斯含量,利用瓦斯涌出量反算煤层瓦斯含量和利用瓦斯参数计算煤层瓦斯含量,对以上四种来源的数据进行综合分析校正。依据告成煤矿瓦斯含量实测数据,对多元线性回归法、BP神经网络法以及支持向量机方法进行模型验证对比,结果表明支持向量机预测模型是较为精确科学的瓦斯含量预测方法,可用于煤矿现场瓦斯含量预测。更多还原

【Abstract】 Relationship between methane content and geologic structure, coal properties, lithology and structure of wall rock, coal seam burial depth was studied based on gas geology. Main factors which affect methane content can be filtered. As an example of Gaocheng Mine, burial depth of bedrock is the leading controlling factor of methane content; followed by the depth correlation coefficient of 50m rock of roof, sand factor and coal seam thickness, while the Cenozoic stratum burial depth, moisture, ash of coal are less interrelated. Methane content prediction model was established using multiple linear regression, BP neural network and support vector machine theory. Methane content measured during exploration and coal production, inversed methane content using out-flow of methane, and methane content calculated using gas parameters were analyzed and regulated. Among multiple linear regression, BP neural network and support vector machine prediction model, support vector machine is proved to be the most accurate one and is useful to predict gas content.更多还原