摘要
针对常规随机反演方法计算效率低的问题,提出一种基于混合遗传算法的叠前随机反演方法。该方法充分利用测井资料中的高频信息,并以地震数据作为约束,首先通过快速傅里叶滑动平均(fast Fourier transform-moving average,FFT-MA)谱模拟算法进行随机模拟得到基于地质统计学的初始模型信息,随后结合提出的混合遗传算法对模拟结果进行快速优化,得到符合地下地质结构的反演剖面,实现对叠前弹性参数的反演。混合遗传算法避免了一般遗传算法常见问题,如收敛速度慢以及产生"早熟"现象,与模拟退火相结合能够快速收敛达到全局最优,保证了反演精度。数值试验结果表明,融入混合遗传算法的叠前随机反演方法,在充分利用叠前信息的同时可以保证反演结果有效收敛,并且与模型数据吻合较好,与传统的叠前反演方法相比具有较高的分辨率,在储层识别和油藏描述中起到了重要作用。
This paper proposed a pre-stack stochastic inversion method based on a hybrid genetic algorithm aiming to resolve the problem of low computation efficiency. It makes full use of the high frequency information of well logging data and is constrained by seismic data at the same time. Firstly, it obtains the geostatistical simulated results through the fast Fourier trans-formmoving average （FFT-MA） spectrum simulation,and then optimizes the initial simulated results using the hybrid genetic algorithm （ HGA） proposed by this paper to obtain the inversion results that correlate with the geological structure. HGA can overcome the drawbacks of conventional genetic algorithm（GA）,such as slow convergence and ＂premature”. It can obtain the optimal results quickly when combined with simulated annealing （ SA）. The numerical testing shows that the pre-stack stochastic inversion based on hybrid genetic algorithm can ensure the convergence of the inversion and also satisfy well data. In addition, this method has high vertical resolution compared with the conventional pre-stack inversion, and may play an important role in reservoir identification and reservoir description.
出处
《中国石油大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2017年第4期65-70,共6页
Journal of China University of Petroleum(Edition of Natural Science)
基金
国家自然科学基金-石油化工基金联合重点项目(U1562215)
国家自然科学基金项目(41204085)
关键词
混合遗传算法
叠前随机反演
分辨率
收敛性
hybrid genetic algorithm
pre-stack stochastic inversion
resolution
convergence
