碎屑岩粒度分析中的子体参数拟合新方法

New Fitting Method of Daughter Parameter in the Size Analysis of Clastic Rock

粒度分析方法在石油地质研究中,特别是储层沉积相研究中,有很广泛的用途。其中,粒度子体分离问题需要求解一个约束最小值问题,但是拟合目标函数具有函数值和导数矩阵不易计算、函数值多峰的特点,常规数值优化算法不易奏效。使用序贯数论网格优化算法(RSNTO)研究了粒度混合正态分布子体参数求解问题,考察了不同范数定义的拟合函数形式对子体参数求解的影响程度。以某油田铁5井深度储层S1+2样品粒度实验数据为算例,进行了数值试验。数值试验结果表明,RSNTO算法可以很好地解决子体参数拟合问题;并且,使用一致范数代替欧氏范数定义拟合目标函数,数值试验结果显示,前者的拟合效果在粒度中间值范围内更好一些,计算出来的累计百分含量曲线与实测点更贴近。

Method of size analysis is widely used in the study of petroleum geology, especially in the study of sedimentary facies of reservoirs.A constrained minimum value should be solved for the problem of daughter separation of the size.But fitting-objective function has the characteristics of being difficult on the calculation of the function value and derivative matrix and multi-peak of the function value.Conventional numerical opti- mization algorithm is not available.Optimization algorithm of sequential number theoretic grid （RSNTO） is ap- plied to the study of solving the problem of size parameter of mixed daughter with normal distribution and eter.Numerical experiment is done taking in certain oilfield as computed case.The results show that RSNTO can solve the problem of daughter parame- ter fitting.Compared with the fitting-objective function defined by Euclidean norm,the fitting-objective function defined by uniform norm may get satisfied fitting result for the medium value of the size and more accurate result for the calculated cumulative centigrade content with the actual value.