利用遗传算法拟合实验变差函数

理论变差函数模型的获取是地质统计学中的基础性工作,它是描述储层非均质性、进行储层随机模拟必不可缺的关键环节。针对现有的拟合实验变差函数方法:人工拟合法、非线性回归最小二乘法、加权多项式拟合法等的不足之处,充分利用遗传算法在求解非线性优化问题时具有的全局寻优的特点,提出实验变差函数的遗传算法拟合方法;在常规遗传算法中引入了模拟退火算法,加速了遗传算法的收敛速度。以川中某气藏的孔隙度变差函数为例进行了计算,结果表明:利用遗传算法可以简单、快速地实现实验变差函数的参数估计。

APPLICATION OF GENENTIC ALGORITHM TO MATCH THEORETICAL VARIOGRAM

Theoreticalmodel of variogram is essential in geostatistics,which is the key parameter in reservoir heterogeneity character ization and stochastic simulation. Aiming at the disadvantages of conventionalusedmethods thatmatch theoretical variogram such as man-made matching method, non- linearmethod of least square, weighted polynomialmatching, an new matching method was presented, which used genetic algorithm to acquire the best fit of theoreticalvariogram. Genetic algorithm is an novelo ptimization method with the characteristic of global optimization.Combined with simulated annealing method, it converged faster than conventional genetic algorithm. Application of the method tomatch theoretical porosity variogram in gas reservoir of Chuan-zhong gas field has shown thismethod was simple and quick for estimating theoretical variogram parameters.