多波叠前AVA非线性反演方法研究
摘要
油气检测和油藏描述中,介质的弹性参数十分重要,这些参数与地层岩性和流体性质有关。描述地层岩性的最主要的3个参数是纵波速度、横波速度和密度,叠前反演是定量求取这三个参数的有效方法。
本文以地震波传播理论为基础,以模拟退火和遗传算法为工具,对多波叠前非线性反演进行研究,反演所用数据为带有入射角的叠前多波动校道集。
首先对单层参数做反演,利用已知的上层参数和反射界面的AVA曲线递推反演下层参数。在初始模型远离真值的情况下,仍可以得到精度较高的下层参数。但这种递推的反演方式对上层参数有强依赖性,反演效率低,不利于多层的推广。
出于适用性和效率的考虑,进一步建立了多层模型,以角道集的误差函数作为目标函数,实现基于波形的多层参数同步反演,反演结果仍具有不依赖于初始模型的优势。通过计算参数扰动所对应的扰动响应,分析了参数的敏感性特征,论证了目标函数的可行性。
为了适应多参数的同步优化,对算法和模型做了改进,包括:①对模拟退火算法增加了快速热浴、记忆保优、回温退火,并将经典退火改为非常快速模拟退火;②将遗传算法的实数编码改进为更适用于地球物理反演问题的五进制编码机制。③改进方案采用层和样点的递进模型,样点模型主要针对层位,以确定反射界面为目的,层模型则针对层参数值的精确反演,根据两种模型的特点进行模型的递进反演;④将纵横波进行联合反演,与单一波型反演相比,能有效提高反演精度,尤其是横波速度的反演精度。
改进后的方案对单界面、多层等多种理论模型均有较高的反演精度,反演对初始模型依赖性低、抗噪能力较强,对实际资料有一定的反演能力。反演所得参数能为线性和广义线性反演提供高质量的初始模型,也可以为油藏描述提供参考。
The elastic parameters of formation are related with lithology and fluid properties, so they are very important for hydrocarbon detection and reservoir description. P-wave velocity, S-wave velocity and density are three main parameters which can descript the formation. In addition, Pre-inversion is an effective method to obtain the value of these three parameters.
On the background of seismic wave propagation theory and adopting the simulated annealing and genetic algorithm, this thesis researched the multi-wave pre-stack nonlinear inversion. The data used for the inversion is multi-wave pre-stack NMO gathers.
First, this thesis inversed the parameters of single layer-model by using the parameters of upper-layer and the AVA curve of the reflector to inverse the parameters of under-layer. Even the initial model is far from the true one, high precision parameters can still be obtained. But the inversion method has a strong dependence on the upper parameters, and its efficiency is so poor that can’t apply to the inversion of multi-layer.
Considering the applicability and efficiency factors, multi-layer model has been established. This model used the error function of angle gather as the objective function to achieve simultaneous inversion. The result of inversion shows it has the advantage of not relying on initial model. According to computing response of every parameter’s disturbance, the sensitivity of parameters has been analyzed. Furthermore the feasibility of the objective function has been demonstrated.
In order to adapting the simultaneous optimization of multi-parameter, this thesis made some improvement for algorithms and model which are as follows: (1) The improvements on SA including fast heat-bath, returning temperature strategy, memory function and changing the classical SA to very fast SA. (2) Switching float encoding to quinary encoding mechanism which is more suitable for geophysical inversion problem to improve GA. (3) The improved method adopted sample points model and layer model, the sample point model focused on layer and aimed at determining the reflection interface, layer model focused on inversing high accuracy parameters of layers. Combining the advantages of these two models can realize progressive inversion. (4) This thesis also jointed PP-wave and PSV-wave to realized inversion. Compared to the sole-wave inversion, joint inversion can improve the accuracy, especially for the S-wave velocity.
The improved inversion method has high accuracy for a variety of theoretic models. It has low dependence on the initial model and strong resistance to noise. It can be applied to the real seismic data and get good inversion result. The inversion results can provide a high-quality original model for linear and generalized linear inversion; also it can give a reference for the reservoir description.
本文以地震波传播理论为基础,以模拟退火和遗传算法为工具,对多波叠前非线性反演进行研究,反演所用数据为带有入射角的叠前多波动校道集。
首先对单层参数做反演,利用已知的上层参数和反射界面的AVA曲线递推反演下层参数。在初始模型远离真值的情况下,仍可以得到精度较高的下层参数。但这种递推的反演方式对上层参数有强依赖性,反演效率低,不利于多层的推广。
出于适用性和效率的考虑,进一步建立了多层模型,以角道集的误差函数作为目标函数,实现基于波形的多层参数同步反演,反演结果仍具有不依赖于初始模型的优势。通过计算参数扰动所对应的扰动响应,分析了参数的敏感性特征,论证了目标函数的可行性。
为了适应多参数的同步优化,对算法和模型做了改进,包括:①对模拟退火算法增加了快速热浴、记忆保优、回温退火,并将经典退火改为非常快速模拟退火;②将遗传算法的实数编码改进为更适用于地球物理反演问题的五进制编码机制。③改进方案采用层和样点的递进模型,样点模型主要针对层位,以确定反射界面为目的,层模型则针对层参数值的精确反演,根据两种模型的特点进行模型的递进反演;④将纵横波进行联合反演,与单一波型反演相比,能有效提高反演精度,尤其是横波速度的反演精度。
改进后的方案对单界面、多层等多种理论模型均有较高的反演精度,反演对初始模型依赖性低、抗噪能力较强,对实际资料有一定的反演能力。反演所得参数能为线性和广义线性反演提供高质量的初始模型,也可以为油藏描述提供参考。
The elastic parameters of formation are related with lithology and fluid properties, so they are very important for hydrocarbon detection and reservoir description. P-wave velocity, S-wave velocity and density are three main parameters which can descript the formation. In addition, Pre-inversion is an effective method to obtain the value of these three parameters.
On the background of seismic wave propagation theory and adopting the simulated annealing and genetic algorithm, this thesis researched the multi-wave pre-stack nonlinear inversion. The data used for the inversion is multi-wave pre-stack NMO gathers.
First, this thesis inversed the parameters of single layer-model by using the parameters of upper-layer and the AVA curve of the reflector to inverse the parameters of under-layer. Even the initial model is far from the true one, high precision parameters can still be obtained. But the inversion method has a strong dependence on the upper parameters, and its efficiency is so poor that can’t apply to the inversion of multi-layer.
Considering the applicability and efficiency factors, multi-layer model has been established. This model used the error function of angle gather as the objective function to achieve simultaneous inversion. The result of inversion shows it has the advantage of not relying on initial model. According to computing response of every parameter’s disturbance, the sensitivity of parameters has been analyzed. Furthermore the feasibility of the objective function has been demonstrated.
In order to adapting the simultaneous optimization of multi-parameter, this thesis made some improvement for algorithms and model which are as follows: (1) The improvements on SA including fast heat-bath, returning temperature strategy, memory function and changing the classical SA to very fast SA. (2) Switching float encoding to quinary encoding mechanism which is more suitable for geophysical inversion problem to improve GA. (3) The improved method adopted sample points model and layer model, the sample point model focused on layer and aimed at determining the reflection interface, layer model focused on inversing high accuracy parameters of layers. Combining the advantages of these two models can realize progressive inversion. (4) This thesis also jointed PP-wave and PSV-wave to realized inversion. Compared to the sole-wave inversion, joint inversion can improve the accuracy, especially for the S-wave velocity.
The improved inversion method has high accuracy for a variety of theoretic models. It has low dependence on the initial model and strong resistance to noise. It can be applied to the real seismic data and get good inversion result. The inversion results can provide a high-quality original model for linear and generalized linear inversion; also it can give a reference for the reservoir description.
引文
[1]李录明,罗省贤.多波多分量地震勘探原理及数据处理方法[M].成都:成都科技大学出版社,1997:251
[2]李录明,李正文.地震勘探原理、方法和解释[M].北京:地质出版社,2007.3
[3]黄绪德.反褶积与地震道反演[M].北京:石油工业出版社,1992
[4]牟永光,刘洋等.地震数据处理方法[M].北京:石油工业出版社,2007.8
[5]殷八斤.AVO技术的理论与实践[M].北京:石油工业出版社,1995:3371
[6]舒梦珵等. P-P波和P-SV波联合AVO反演研究[J].地球物理学进展,2008.10.23:1556~1559
[7]陈天胜,刘洋,魏修成.几种P—SV转换波反射系数近似公式的比较[J].石油物探,2005.1:1000-1441
[8]杨新菊,马建德.AVO正演模型研究及应用[J].海洋石油,2006.9
[9]王学军,马劲风.考虑振幅随炮检距变化的人工合成地震记录制作方法[J].石油地球物理勘探,2001.2(36-1)
[10]周怀来.多波波场正演模拟与波场特征分析研究[D] .成都理工大学博士学位论文,2009
[11]隋容亮.多波多分量地震勘探技术[D].中国地质大学博士论文,2006
[12]雍阳.横向各向同性介质多波振幅特征及参数反演方法研究[D].成都理工大学硕士学位论文,2003
[13] Muskal M, M. W. Meres, Reflection and Transmission Coefficients for Plane Waves in Elastic Media [J].Geophysics. 1940, 5 :149~155.
[14] Ostrander W J. Plane—wave reflection coefficients for gas sands at non-normal angles of incidence[J].Geophysics, 1984, 49(6):1637～1648.
[15]姚姚.地球物理反演基本理论与应用方法[M].武汉:中国地质大学出版社2002.8
[16]邢文训,谢金星.现代优化计算方法[M].北京:清华大学出版社,2005.9
[17]蒋金山,何春雄,潘少华.最优化计算方法[M].广州:华南理工大学出版社2007.10
[18] Metropolis N, Rosenbluth A, Rosenbluth M et al. Equation of state calculation by fast computing machines[J]. Journal of Chemical Physics, 1953, 21: 1087-1092
[19] Kirkpatrick S, Gelatt Jr C D, Vecchi M P. Optimization by simulated annealing [J]. Science,1983, 220:671-680
[20]康立山,谢云,尤矢勇,罗祖华.非数值并行算法——模拟退火算法[M].北京:科学出版社,1994.4
[21] Kirkpatrik S, Celatt C D, Vecchi M P. Optimization by simulated annealing [J]. Chem. Phys. ,1983,21:1087~1092
[22] Sen, M. K, Stoffa, P. L, Nonlinear one-dimensional seismic waveform inversion usingsimulated annealing[J], Geophysics, 56, 1624-1638, 1991
[23] Jacob. M. P., Peter. D. V., Theo. Z.,模拟退火法地震反演[C].第61届SEG年会论文集.北京:石油工业出版社,1993.
[24] Stoffa P L and Sen M K. Nonlinear multi-Parameter optimization using genetic algorithms: inversion of plane wave seismograms[J]. Geophysics,1991,56(11):1794-1810
[25] Rothman H.著,田希泰译.大近地表异常、地震反射数据和模拟退火[C].外协论文集,第一集No. 5,石油部物探局研究院,1987
[26] Rothman D. H. Automatic estimation of large residual statics correction, Geophysics, 1986, 51(2):323-346
[27]师学明,王家映.模拟退火法[J].工程地球物理学报,2007.6
[28]陈国良,王熙法,庄镇全,王东生.遗传算法及其应用[M].北京:人民邮电出版社,1996.1
[29]李敏强,寇纪淞,林丹,李书全.遗传算法的基本理论与应用[M].北京:科学出版社,2002.3
[30]刘勇,康立山,陈毓屏.非数值并行算法——遗传算法[M].北京:科学出版社,1995
[31] Goldberg, D. E. Genetic algorithms in search, optimization and machine learning [M]. MA: Addison- Wesley Publishing Company, 1989
[32] Holland J H. Adaption in Natural and Artificial Systems [C]. Ann Arbor: University of Michigan, 1975
[33]石琳珂,孙铭心,王广国,赵盛华.地球物理遗传反演方法[M].北京:地震出版社,2000.5
[34]黄河,任韶萱.遗传算法及其在速度结构与震源联合反演中的应用[J].东北地震研究,1998 ,14 (4) :1～14.
[35] Sen M. K, Stoffa, P L. Rapid sampling of model space using genetic algorithms: examples from seismic waveform inversion [J]. Geophys. J. Internat. 1992,108:281~292
[36]万永革,李鸿吉.京津唐张地区速度结构和震源位置联合反演的遗传算法[J].地震地磁观测与研究,1996 ,17 (3) : 57～66.
[37]王文萍,王庆良.利用遗传算法和最小二乘联合反演共和地震位错参数[J].地震学报,1999 , 21 (3) : 285～290.
[38]王玉兰,沈越江,杨绍国.GA解决多参数超大解空间优化问题的启发式收索方法[J].物探化探计算技术,2005,27(2):115.118.
[39]袁慰平,孙志忠等.计算方法与实习[M].南京:东南大学出版社,2000
[40]谭浩强.C程序设计(第二版)[M].北京:清华大学出版社,1999
[41]陈永琴,苏三买,牟锦辉.随机数对遗传算法的影响分析及改进研究[J] .航空计算技术,2008.3(38-2)
[42]李珂景.浅析C语言中的随机数问题[J].长春大学学报2008.6(18-3)
[43]郭继展,郭勇,苏辉.新编C语言程序设计[M].北京:机械工业出版社,2007.7
[2]李录明,李正文.地震勘探原理、方法和解释[M].北京:地质出版社,2007.3
[3]黄绪德.反褶积与地震道反演[M].北京:石油工业出版社,1992
[4]牟永光,刘洋等.地震数据处理方法[M].北京:石油工业出版社,2007.8
[5]殷八斤.AVO技术的理论与实践[M].北京:石油工业出版社,1995:3371
[6]舒梦珵等. P-P波和P-SV波联合AVO反演研究[J].地球物理学进展,2008.10.23:1556~1559
[7]陈天胜,刘洋,魏修成.几种P—SV转换波反射系数近似公式的比较[J].石油物探,2005.1:1000-1441
[8]杨新菊,马建德.AVO正演模型研究及应用[J].海洋石油,2006.9
[9]王学军,马劲风.考虑振幅随炮检距变化的人工合成地震记录制作方法[J].石油地球物理勘探,2001.2(36-1)
[10]周怀来.多波波场正演模拟与波场特征分析研究[D] .成都理工大学博士学位论文,2009
[11]隋容亮.多波多分量地震勘探技术[D].中国地质大学博士论文,2006
[12]雍阳.横向各向同性介质多波振幅特征及参数反演方法研究[D].成都理工大学硕士学位论文,2003
[13] Muskal M, M. W. Meres, Reflection and Transmission Coefficients for Plane Waves in Elastic Media [J].Geophysics. 1940, 5 :149~155.
[14] Ostrander W J. Plane—wave reflection coefficients for gas sands at non-normal angles of incidence[J].Geophysics, 1984, 49(6):1637～1648.
[15]姚姚.地球物理反演基本理论与应用方法[M].武汉:中国地质大学出版社2002.8
[16]邢文训,谢金星.现代优化计算方法[M].北京:清华大学出版社,2005.9
[17]蒋金山,何春雄,潘少华.最优化计算方法[M].广州:华南理工大学出版社2007.10
[18] Metropolis N, Rosenbluth A, Rosenbluth M et al. Equation of state calculation by fast computing machines[J]. Journal of Chemical Physics, 1953, 21: 1087-1092
[19] Kirkpatrick S, Gelatt Jr C D, Vecchi M P. Optimization by simulated annealing [J]. Science,1983, 220:671-680
[20]康立山,谢云,尤矢勇,罗祖华.非数值并行算法——模拟退火算法[M].北京:科学出版社,1994.4
[21] Kirkpatrik S, Celatt C D, Vecchi M P. Optimization by simulated annealing [J]. Chem. Phys. ,1983,21:1087~1092
[22] Sen, M. K, Stoffa, P. L, Nonlinear one-dimensional seismic waveform inversion usingsimulated annealing[J], Geophysics, 56, 1624-1638, 1991
[23] Jacob. M. P., Peter. D. V., Theo. Z.,模拟退火法地震反演[C].第61届SEG年会论文集.北京:石油工业出版社,1993.
[24] Stoffa P L and Sen M K. Nonlinear multi-Parameter optimization using genetic algorithms: inversion of plane wave seismograms[J]. Geophysics,1991,56(11):1794-1810
[25] Rothman H.著,田希泰译.大近地表异常、地震反射数据和模拟退火[C].外协论文集,第一集No. 5,石油部物探局研究院,1987
[26] Rothman D. H. Automatic estimation of large residual statics correction, Geophysics, 1986, 51(2):323-346
[27]师学明,王家映.模拟退火法[J].工程地球物理学报,2007.6
[28]陈国良,王熙法,庄镇全,王东生.遗传算法及其应用[M].北京:人民邮电出版社,1996.1
[29]李敏强,寇纪淞,林丹,李书全.遗传算法的基本理论与应用[M].北京:科学出版社,2002.3
[30]刘勇,康立山,陈毓屏.非数值并行算法——遗传算法[M].北京:科学出版社,1995
[31] Goldberg, D. E. Genetic algorithms in search, optimization and machine learning [M]. MA: Addison- Wesley Publishing Company, 1989
[32] Holland J H. Adaption in Natural and Artificial Systems [C]. Ann Arbor: University of Michigan, 1975
[33]石琳珂,孙铭心,王广国,赵盛华.地球物理遗传反演方法[M].北京:地震出版社,2000.5
[34]黄河,任韶萱.遗传算法及其在速度结构与震源联合反演中的应用[J].东北地震研究,1998 ,14 (4) :1～14.
[35] Sen M. K, Stoffa, P L. Rapid sampling of model space using genetic algorithms: examples from seismic waveform inversion [J]. Geophys. J. Internat. 1992,108:281~292
[36]万永革,李鸿吉.京津唐张地区速度结构和震源位置联合反演的遗传算法[J].地震地磁观测与研究,1996 ,17 (3) : 57～66.
[37]王文萍,王庆良.利用遗传算法和最小二乘联合反演共和地震位错参数[J].地震学报,1999 , 21 (3) : 285～290.
[38]王玉兰,沈越江,杨绍国.GA解决多参数超大解空间优化问题的启发式收索方法[J].物探化探计算技术,2005,27(2):115.118.
[39]袁慰平,孙志忠等.计算方法与实习[M].南京:东南大学出版社,2000
[40]谭浩强.C程序设计(第二版)[M].北京:清华大学出版社,1999
[41]陈永琴,苏三买,牟锦辉.随机数对遗传算法的影响分析及改进研究[J] .航空计算技术,2008.3(38-2)
[42]李珂景.浅析C语言中的随机数问题[J].长春大学学报2008.6(18-3)
[43]郭继展,郭勇,苏辉.新编C语言程序设计[M].北京:机械工业出版社,2007.7