Abstract: When the parameters of groundwater flow model are estimated, in order to see the real movement, optimization algorithm is used to count the approximate solution. Based on the traditional simulated annealing algorithm, an improved simulated annealing genetic algorithm is proposed. The new algorithm assimilates the advantage of genetic algorithm for the capability of global searching and best individual protection. It avoids earliness and strengthens the property of local searching. Taking 2D unsteady state flows in anisotropy confined aquifer for example, it estimates the parameters of groundwater flow model. The result indicates that the new algorithm avoids the disadvantage of low convergence rate and excessive iterations, possesses the advantages of high precision and parallelism.

Key words: simulated anneal, genetic algorithm, groundwater, parameter estimation

摘要： 在估计地下水数值模型参数时，常运用智能优化算法求解数学模型的近似解，以再现现实地下水流的运动。在传统的模拟退火算法基础上，结合遗传算法，提出了一种改进模拟退火遗传算法，它吸收了遗传算法的全局搜索性能和保护最优个体的策略，解决了遗传算法早熟的问题，加强了模拟退火的局部搜索能力。以非均质各向异性承压二维流为例，运用该算法对地下水流数值模型参数进行了反演计算。计算结果表明，该算法克服了传统全局搜索算法收敛速度慢、迭代次数多的缺点，具有计算精度高，可以并行计算等优点。

关键词: 模拟退火, 遗传算法, 地下水, 参数估计