摘要
为解决地震资料波阻抗反演中的不适定性,提高反演精度和收敛速度,基于马尔柯夫随机域的领域系统及势函数建立目标函数,采用完全非线性遗传算法进行数值反演。主要研究遗传算法的改进措施,采用二进制与实数混合编码机制;应用无放回式余数随机选择,实现二进制多次交叉和均匀实数交叉策略,充分利用种群中优秀个体的有效模式,提高收敛速度和反演精度;依据保持种群多样性原则,采用父代与子代竞争及动态调整子代策略,引入更新概率,提高全局搜索能力,避免早熟收敛。结果表明,建立理论速度模型反演,采用改进的遗传算法全局搜索性强,反演精度高,具有一定的抗噪能力,能减少早熟收敛的概率。
In order to solve the ill-posedness in the seismic wave impedance inversion,improve the inversion accuracy and convergence speed,this paper builds the objective function based on the neighborhood system of MARKOV random field,and adopts complete nonlinear genetic algorithm to perform numerical inversion.It studies the improvement measures of genetic algorithm by using binary and real number with mixed encoding mechanism,adopts the remainder without replacement to achieve the strategy of the binary multiple and uniform crossover and makes full use of effective model of excellent individuals in the population to improve the convergence speed and inversion accuracy.To keep population diversity,competition between the parent and offspring,the strategy of adjusting offspring dynamically and updating probability are involved to improve the ability of global search and avoid premature convergence.Building theoretic velocity model and using improved genetic algorithm lead to better global search ability,higher retrieval accuracy,certain noise immunity and reduced probability of premature convergence.
作者
尹德强
Yin Deqiang(Northeast Electric Power Design Institute Limited Company, China Power Engineering Consulting Group, Changchun Jilin 130000,China)
出处
《工程地球物理学报》
2020年第3期327-333,共7页
Chinese Journal of Engineering Geophysics
关键词
波阻抗
遗传算法
正则化
反演
impedance
genetic algorithm
regularization
inversion
