节点文献
预测砂岩孔隙度的地震多属性优化模式对比
Multi-attribute optimization analysis for sandstone porosity prediction
【摘要】 利用地震多属性分析技术预测地下储层参数时,常常面临着对多个地震属性进行选择优化的问题。为了比较众多优化方法的异同,本文从数学上的相容性与冗余性、独立性与非相关性的角度出发,建立了一套地震多属性优化分析的数学表达式。在此基础上,将突出独立性的主成分分析优化、突出相容性的粗糙集属性优化、突出相关性的有效性—离散度—相关性三参数属性优化,从流程上串联起来,形成了8种地震多属性的优化模式,将优化后的8组属性输入径向基函数神经网络,外推砂岩储层的孔隙度参数。应用实例表明,粗糙集属性优化—主成分分析—神经网络模式识别的综合预测效果最好,全属性直接神经网络模式识别的综合预测效果要差一些,8种流程计算出的砂岩孔隙度的相对误差都在可接受范围内。
【Abstract】 When we deal with a large number of seismic attributes,we have to choose some of them for optimization to improve reservoir prediction.In this work,we introduce a set of mathematical expressions for multi-attribute optimization analysis based on compatibility and redundancy,independence and non-related of multi-attribute.We combine KL transform highlighting independence,RS optimization highlighting compatibility,and SDC optimization highlighting relevance into 8 series optimization model for comparison.Then we input these 8 groups of optimized attributes in RBFNN to extrapolate porosity of sandstones.Finally,we apply our workflow on the field data from Bohai Bay,China.This field example shows that RS-KL-RBFNN is the most effective workflow for reliable porosity measurements and the relative error of 8 porosity estimated by 8 processes are acceptable for reservoir characterization.
【Key words】 seismic multi-attribute; significance-dispersion-correlation(SDC) optimization; rough set(RS) optimization; KL transformation; radial base function neural network(RBFNN); porosity;
- 【文献出处】 石油地球物理勘探 ,Oil Geophysical Prospecting , 编辑部邮箱 ,2011年03期
- 【分类号】P631.4
- 【被引频次】10
- 【下载频次】321