摘要
基于分形高频初始模型和低频先验信息的储层物性参数随机反演方法综合利用混合先验信息和统计岩石物理模型,在贝叶斯理论框架下直接反演储层物性参数。首先通过分形高斯模型算法和克里金插值得到物性参数的先验信息;然后根据统计岩石物理模型建立弹性参数与储层物性参数之间的关系,构建似然函数;最终利用基于模拟退火法(Simulated Annealing,SA)改进的粒子群优化(Particle Swarm Optimization,PSO)算法(SAPSO)实现后验概率密度的抽样。与确定性反演结果相比,该方法能够有效地融合测井资料中的高频信息,提高反演结果的分辨率,并且先验信息中融合了低频成分,可以得到宽频带的反演结果。一维和二维实际资料反演得到的孔隙度、泥质含量和含水饱和度与井资料吻合很好,分辨率较高,验证了该方法的可行性。
Stochastic inversion of reservoir physical property parameters based on high-frequency initial model from fractal and low frequency prior information is direct inversion of reservoir parameters utilizing mixed prior information and statistical petrophysical model in a Bayesian framework. Firstly,we can get the prior information of reservoir physical parameters through fractal Gaussian model algorithm and Kriging interpolation Then accord ing to the statistical petrophysical model, we get the relationship between elastic parameters and reservoir physical parameters and build a likelihood function Finally we apply SA-PSO optimization algorithm in order to obtain the sampling of the posterior probability density. Compared with deterministic inversion, the method we proposed can effectively integrate high frequency information in logging data to improve the resolution of inversion results. And prior information contains low frequency components, so the inversion results are of broadband. Porosity, clay con tent and water saturation from inversion of 1D and 2D actural data have a good agreement with the well data and higher resolution, verify the feasibility of this method.
出处
《石油物探》
EI
CSCD
北大核心
2014年第5期537-544,共8页
Geophysical Prospecting For Petroleum
基金
国家重点基础研究发展计划(973)项目(2013CB228604)
国家科技重大专项(2011ZX05009)
国家自然科学基金(41204085)
山东省自然科学基金(ZR2011DQ013)
中国石油化工股份有限公司地球物理重点实验室开放基金(WTYJY-WX2013-04-07)
中央高校基本科研业务费专项资金(11CX04011A)联合资助
关键词
分形高斯模型算法
统计岩石物理
SA-PSO优化算法
贝叶斯理论
物性参数反演
fractal Gaussian moclel algorithm
statistical petrophysics
SA-PSO optimization algorithm
Bayesian theory
inversion o{ physical property parameters
