井震联合数据驱动下，多智能技术融合的煤层气储层参数预测与评价

项目名称： 井震联合数据驱动下，多智能技术融合的煤层气储层参数预测与评价

项目编号： No.51504085

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 矿业工程

项目作者： 边莉

作者单位： 黑龙江科技大学

项目金额： 20万元

中文摘要： 煤层气储层的强非均质性导致同一煤层小范围内的储层参数发生明显改变，常规储层参数预测方法难以适用。特别是井震数据的冗余更增加了参数预测难度。在测井-地震数据联合驱动下，以粗糙集作为井震数据约简工具，以混合核支持向量机技术构建非线性预测模型，以交叉熵理论提升支持向量机和粗糙集性能，结合理论分析和模拟评价，建立一种有效的多智能技术融合的煤层气储层参数预测模型。整理岩心、测井曲线和地震属性等数据，建立纵、横向原始决策表（粗糙集决策表）；利用组合型交叉熵算法对粗糙集决策表离散化和条件属性约简，优选井震数据，形成相应的约简决策表；针对煤层气储层非线性特性，结合约简决策表，利用线性加权构建混合核支持向量机的储层参数预测模型，采用连续型交叉熵算法优化支持向量机参数，修正预测模型；利用测试样本评价预测模型的有效性，完成研究工区储层参数的纵、横向预测。预测结果可为煤层气的勘探、开采和利用提供强有力的技术支持。

中文关键词： 煤层气；储层参数；交叉熵算法；数据驱动；预测与评价

英文摘要： The strong heterogeneity of Coal Bed Methane (CBM) reservoir leads to the reservoir parameters changed significantly at a small range within the same coal seam, regular reservoir parameters prediction methods are difficult to apply. Especially, the redundancy of well-seismic data also increases the difficulty of the parameters prediction. ..Under well-seismic date joint driven, taking rough set as the tool of well-seismic data reduction, building non-linear prediction models by the mixed kernel support vector machine technology, enhancing support vector machine and rough set performance by cross-entropy theory, combining with theoretical analysis and simulation evaluation, an effective prediction model of CBM reservoir parameters based on multi-agent technology integration is built...Processing core, well logging curve and seismic data establishes the conventional vertical and horizontal original decision table (rough set decision table); the use of combination cross-entropy algorithm, rough set decision table discretization and condition attribute reduction select well-seismic data to form the corresponding reduction decision table; for coalbed methane reservoir nonlinear characteristics, combined with the reduction decision table uses a linear weighted to build a reservoir parameters prediction model of the mixed kernel support vector machine and use continuous cross-entropy algorithm to optimize the parameters of support vector machine and to revise the prediction model. It utilizes the test samples to evaluate the effectiveness of the prediction model and accomplishes the vertical and horizontal reservoir parameters prediction of the work area...The results of prediction can provide strong technical supports for the exploration, exploitation and utilization of CBM.

英文关键词： coal bed methane;reservoir parameters;cross entropy algorithm;date driven;prediction and evaluation