【作者】 张学庆；

【导师】 肖慈珣； 刘争平；

【作者基本信息】 成都理工大学 ， 地球探测与信息技术， 2002， 博士

【摘要】 本文对常规地震测井联合反演作了深入剖析，并针对其中存在的问题，为达到认识和描述地下地层的目的，结合智能算法的特点，提出并实现了一种突破传统思想的直接反演测井属性的地震测井联合反演方法——测井属性的地震测井智能联合反演方法，并在实际应用中取得了较好的效果。本文的研究具有较强的开拓性，同时也具有较高的学术价值和实际应用价值。 本文主要在以下几个方面作了深入的分析和研究： (1)通过对演化算法、模糊系统和神经网络以及小波分析算法的深入分析和研究，认为这些技术不是竞争的，而是互补的。作者将它们有效地结合起来设计出了新的混合智能算法—演化神经网络、神经模糊系统以及小波神经网络。这些新的算法在非线性处理、函数逼近、容错性以及自组织和自学习等方面的性能有了很大的提高，其中尤其以小波神经网络更为突出。 (2)测井属性的地震测井智能联合反演方法的提出。对常规地震测井联合反演中作了深入的剖析，认为其在非线性问题、子波处理、模型建立以及对实际资料的容错性等方面仍存在诸多问题。本文在前人研究的基础上，针对这些问题并基于将地震道数据转化为更容易为地质学家所理解的测井属性参数的思想；结合智能算法的特有优势以及现代信息论中数据驱动和信息融合的思想，提出了测井属性的地震测井智能联合反演方法。该方法则突破了常规地震测井联合反演只能直接反演波阻抗这一思路，可以直接反演任意测井属性。避免了对先验模型要求的假设前提，解决了反演的非线性求解问题，同时也避免了复杂的变子波问题。 (3)测井属性的地震测井联合反演的实现和实际应用。基于测井属性的地震测井智能联合反演的基本思路，本文实现了基于原始数据驱动和特征参数驱动的地震测井智能联合反演方法的设计，并以塔河地区一剖面为例做了实际应用。使每一地震道数据都转换为测井属性，最终实现了二维(或三维)地震数据向二维(或三维)测井属性的转换，达到了更清楚和精细地描述井间地层信息的目的。反演预测结果与实测数据之间的具较强的相关性，表明测井属性的地震测井智能联合反演是成功地。反演结果同时还表明了混合智能算法优于常规智能算法；提取的分形和小波特征参数是合理地。另外，为提高反演结果的分辨率，尝试先地震数据插值再进行反演，结果反演效果变差，说明了数据本身信息的不足无法用数学技巧来弥补，不能单纯地追求高分辨率的反演结果。 (4)有效地利用先验知识对反演进行有效的约束是提高复杂地层条件下联合反演效果的一个新的思路。本文对测井属性的地震测井联合反演方法中低频输入信息作了进一步的探讨，认为在复杂地区简单的低频井间插值处理是不合理的。针对这一问题，以塔河地区为例建立了该地区的地质模型并对其进行了数学描述，进而对低频信息进行约束，最终实现了基于先验知识约束的地震测井智能联合反演，反演效果得到了很大改善。更多还原

【Abstract】 This paper puts forward an unconventional method of joint inversion of seismic data and well-logging, the joint inversion of well-logging properties from seismic and log data based on intelligent algorithms for realizing and describing underground strata after studying conventional joint inversion of seismic data and well-logging, which can inverse well-logging properties directly. Fine results have achieved in practice application. The study of this paper has considerable innovation, academic and theory value.In this paper. the main research works and achievement includes following contents :(1) Evolutional algorithm , neural network . fuzzy system and wavelet analysis work and wavelet analysis are compensatory with each other but not competitive. So many new mixed intelligent algorithms have been designed which include evolutionary neural network, neural fuzzy system and wavelet neural network. These new algorithms improved the capability of nonlinear disposing, function approaching, fault-tolerated and self-organizing but also self-learning, wavelet neural network is the best one among them .(2) Putting forward the join inversion of well-logging properties from seismic and log data based on intelligent algorithms. Conventional joint inversion of seismic data and well-logging have many problems such as nonlinear disposing, wavelet problem, model of inversion and reduction and error toleration of data source after studying it in depth. The method of joint inversion of seismic data and well-logging based on intelligent algorithm is built which includes the special merits of intelligent algorithms, thought of data-driven and information fusion. This method broke the basic thought of conventional joint inversion, can inverse well-logging properties directly and has overcome the difficulties of conventional joint inversion, such as can avoid the precondition of requested previous model . settle the problem of nonlinear solution-resolved and avoid the problem of complex variational wavelet.(3 ) Realization and application of the well-logging properties from seismic and log data based on intelligent algorithms . The intelligently joint inversion of seismic data and well-logging based on source data driven and character parameter-driven are realized based on the basic thought of intelligent joint inversion, and are applied in Tahe oil field with one profile successfully. Every path of seismic data can be transformed into well-logging properties, so the profile of seismic data can be transformed into profile of well-logging properties for the aim of realizing and describing the underground strata. The inversed result has high precision and indicates that mixed intelligent algorithm is better than conventional intelligent algorithms. The result become bad when improving the resolution of seismic data for high resolution inversion result, which showed that scarce of information of data can’t be compensated with mathematic algorithm and it was no use to aspire after high resolution result of inversion artificially.(4) Using the previous knowledge to restrict the import information of inversion system is a new method to improve inversion effect especially under the complex conditions. This paper discussed the problem of the low-frequency information, and noticed that it is not reasonable under the condition of complex stratum , and the example also approved this . Aiming this, thegeology model of Tahe oil field is built and described in mathematics as an example for restricting the low-frequency information. The result of inversion is much more better.更多还原