最优化测井解释的遗传算法实现

REALIZING GENETIC ALGORITHM OF OPTIMAL LOG INTERPRETATION

遗传算法是具有全局寻优特点的一种新型最优化方法 ,它适合于求解非线性优化问题 ,近年来已经在复杂函数的优化求解、结构优化设计、自适应控制、系统控制、模式识别等领域取得成功应用。文章将遗传算法引入测井最优化解释中 ,实现了用遗传算法进行最优化测井解释 ,充分利用遗传算法的全局寻优特点 ,研究多参数、非线性最优化测井数据处理问题 ;同时在充分讨论常规遗传算法原理的基础上 ,针对常规时齐遗传算法的强不收敛性 ,引入了非时齐遗传算法 ,从结合整体模拟退火算法、引入动态记忆窗解决成熟前收敛问题两方面对常规遗传算法作了改进 ,经正演模型模拟运算和实际测井资料的解释结果表明 ,经改进后的遗传算法更易于收敛 ,可以较好地研究多参数、非线性最优化测井资料解释。

Genetic algorithm is a novel optimizing method with the character of global optimization and it is suitable for solving the problems of non-linear optimization.In recent years,it has been successfully applied to many fields,such as optimal solution of complex function,structure optimal design,adaptive control,system control and pattern recognition,etc.The genetic algorithm was introduced into optimal log interpretation by the authors,in other words,the optimal log interpretation was realized by use of the genetic algorithm.The questions about multi-parameter and nonlinear optimum log data processing were studied by fully applying the character of global optimization of the genetic algorithm.On the basis of fully discussing conventional genetic algorithm theory,it is pointed out in the paper that the conventional genetic algorithm was improved from two aspects:① introducing non-time-homogeneous genetic algorithm in light of the severe non-convergence of the conventional time-homogeneous genetic algorithm;and ② in combination with integral simulation annealing algorithm,introducing dynamic record window technique in order to solve the problem of pre-mature convergence.Through forward model simulation operation and practical log data interpretation,it is shown that the genetic algorithm improved is much easier to be converged,by which the multi-parameter and nonlinear optimum log data interpretation may be well studied.