【作者】 聂茹；

【导师】 岳建华；

【作者基本信息】 中国矿业大学 ， 地球信息科学， 2009， 博士

【摘要】 我国煤炭资源西多东少,煤炭生产的重点正逐步西移。相对于东部而言,西部地区煤田虽然构造简单,但是煤层结构较为复杂,对比解释困难。因此,在我国西部地震勘探中由构造勘探向岩性勘探转变势在必行,而波阻抗反演技术是实现岩性勘探和解释的重要手段。波阻抗反演属于典型的非线性优化问题,常规反演算法如广义线性反演、最大似然反演等均采用非线性问题的线性化或拟线性化处理方法,导致反演结果严重依赖于初始模型,并且易陷入局部最优以及具有多解性等不可避免的缺陷。而随着前沿学科的交叉渗透,新技术、新方法的引入以及计算机软硬件技术的飞速发展,将智能计算方法应用于非线性波阻抗反演已逐渐成为地球物理领域研究的热点问题之一。本文针对单一模拟退火、遗传算法等智能算法在进行波阻抗反演时存在的早熟收敛、计算量大以及计算结果不够稳定等困难,将不同智能算法相结合并将一些新颖的智能算法引入到波阻抗反演问题中,开展了混合智能算法的改进及其在波阻抗反演中的应用研究。本文的主要研究成果及创新点如下:(1)研究了提高反演精度的小波分频和声波测井曲线重构技术。提出了用于煤田声波测井曲线重构的GA-BP混合优化算法。GA-BP混合优化算法综合了密度、视电阻率曲线信息,通过GA训练神经网络的权值和阈值和引入能够增加收敛速度的BP算子,提高了声波测井曲线重构的准确性,为波阻抗反演提供了可靠的数据保证。(2)建立了约束条件下智能计算反演方法的目标函数。为提高无约束条件目标函数的反演效果,在目标函数中加入测井曲线所提供的先验波阻抗信息和边界波阻抗差值与振幅响应的关系作为约束条件,从而限定解的取值范围,以提高收敛速度和减少运算时间。同时根据波阻抗反演问题的具体特点,对智能计算反演方法中参数的编码和适应度函数的定义等关键问题进行了分析和研究。(3)提出了改进的遗传退火算法。针对波阻抗反演问题的特点,采用十进制对反演参数进行编码,避免了十进制和二进制进行相互转换的环节,提高了算法收敛速度。定义了波阻抗反演问题的个体适应度函数,并引入模拟退火机制进行适应度拉伸,进一步改善了算法的收敛性能。计算结果表明,在无噪声情况下的反演结果与真实模型完全吻合;随着地层数的增加以及噪声的增强,反演误差有所增大;但其反演精度明显优于遗传算法和模拟退火算法。(4)提出了改进的混合粒子群优化算法。为克服基本粒子群算法在进化后期收敛速度慢和易陷于局部极值的缺陷,引入浓度选择和免疫记忆机制,实现免疫算法和粒子群算法的有机结合;综合考虑全局搜索能力和局部搜索能力,提出两阶段搜索策略;提出克隆选择和Logistic序列相结合的最优粒子变异机制。通过理论模型反演实例表明,改进的混合粒子群优化算法反演精度高,收敛速度快,具有良好的抗噪性能;通过间距40m的2m和3m两个薄煤层理论模型反演实例表明,改进算法能够识别2m以下薄煤层。(5)设计了能够更好地模拟现实世界免疫行为的免疫选择、免疫变异、种群重组、混沌增殖等免疫优化算子,在此基础上提出了免疫遗传优化算法。针对波阻抗反演问题的实际特点,为解决全局搜索能力和种群多样性之间的矛盾,在进化过程中动态调整抗体群规模。通过理论模型反演实例表明,改进算法反演精度高,抗噪性能强;通过层间距小于波长的2m、3m和5m三个薄煤层理论模型反演实例表明,改进算法能够很好的反演出2m、3m和5m三个薄煤层。最后,通过改进的混合粒子群和免疫遗传算法在阳泉二矿和成庄煤矿实际资料反演中的应用,证明两种新的混合智能优化方法所得反演剖面的纵向分辨率与实际地震剖面相比有较大提高,弱反射波的连续性和可检测性得到增强,为研究下组煤分布提供了有效方法,能够应用于煤炭领域岩性勘探研究。更多还原

【Abstract】 The focus of coal production in our country is shifting from east to west, for the west occupies most of coal resources. Compared to the east, in the west the formation of the coal field is simple, but the fabric of coal seam is much complicated leading to difficult comparison interpretation. So it is necessary to realize lithology interpretation from structure interpretation. Wave impedance inversion is a key method to achieve lithology interpretation and it plays a key role in geophysical prospecting. Wave impedance inversion belongs to nonlinear optimization. However, conventional inversion methods such as generalized linear inversion and so on are based on linear theory or quasi-linear theory, which results in sensitive to initial model and local extreme and multi-extreme. With the development of computer technology and the introduction of new leading subjects, new methods based on intelligent computation are becoming the focus in geophysical domain. In order to improve premature convergence and reduce calculation cost of the current commonly used intelligent inversion methods,in this dissertation hybrid intelligent algorithms and novel intelligent computation methods are introduced and to be applied to wave impedance inversion. In this paper, the main research works and the achievement obtained include the following contents.(1) GA-BP optimization algorithm was proposed to reconstruct sonic logging curve in coal field. Combining density and resistance, GA-BP algorithm improved the accuracy of reconstruction of the sonic logging curve, which provided reliable data for inversion.(2) The object function with restriction in wave impedance inversion was constructed. To improve the effect of inversion under the object function without restriction, sonic logging information and the relation between the wave impedance difference in the boundary and the amplitude were added to the object function. So the inverted impedance was limited in a certain range and the convergence speed was greatly increased. Considered the characteristic of wave impedance inversion the encoding and fitness function was also discussed.(3) Improved GASA optimization method was proposed by combing modified genetic operators and simulated annealing. According to the characteristic of wave impedance inversion, decimal encoding system was adopted to avoid the transformation from binary system to decimal system and speed the convergence. Fitness function was defined and drawn by simulated annealing mechanism. Computation results showed that without noise the inverted results were consistent to the real model and with the increase of noise and the inverted parameters the precision of inversion were decreased, but the precision were better than GA and SA.(4) Improved hybrid particle swarm optimization (PSO) algorithm was proposed. To avoid the premature convergence of particles and slow convergence in the late process, the immune memory idea and the selection strategy based on antibody density were introduced into PSO and the proposed two-stage search strategy took the global search ability and the local search ability into account. Furthermore, the proposed cloning selection operator accelerated the best particle away from the local extreme and Logistic sequence was adopted to extend the search scope and further the best particle mutation. The simulation results indicated that the improved hybrid PSO had better efficiency and higher accuracy. The theoretical model of two thin coal seam with 40 meters interval was constructed and the inverted results showed that the thickness of less 2m coal seam was identified.(5) Based on immune optimization operators such as immune selection, immune mutation, population recombination and chaos multiplication, an improved immune genetic algorithm was proposed. According to the wave impedance inversion, the size of population was dynamic adjusted to improve population diversity and global convergence. The simulation results indicated that the improved immune genetic algorithm had a better efficiency and a little higher accuracy than immune hybrid PSO algorithm. The theoretical model of three thin coal seam with intervals less than wave length was constructed and the inverted results showed that the thickness of 2m, 3m and 5m coal seams were identified.Finally, the proposed immune hybrid PSO and immune genetic algorithm were applied to Yangquan and Chengzhuang coalfield. The inversion results showed that the inversion resolution were obviously higher than the seismic resolution, the continuity and the detection capability of weak reflection were improved greatly, which provided an effective method for study lower group coal and lithology interpretation.更多还原