模拟退火算法反演水文地质参数算例研究

Application of simulated annealing algorithm to hydrogeological parameter inversion

基于模拟退火算法(SA)这一全局优化技术,耦合地下水模拟的有限元模型,给出水文地质参数的反演方法。并用水文地质参数反演算例来验证反演模型的可用性和适用性。结果表明,用"变尺度"产生新的渗透系数反演结果明显优于"随机法",当初始温度在500～650、降温系数在0.92～0.95时,模拟退火法在正常情况获得了很好的反演结果,在考虑减少观测孔个数和观测水头值有误差的情况下,也取得了较为满意的反演结果。为了提高运算效率,可考虑与传统反演方法或其它人工智能算法相结合。

The authors present an approach for inverse modeling of groundwater systems based upon a simulated annealing algorithm (SA) global optimization technique which is coupled with a finite element (FEM) flow simulation model. SA is applied to inverse hydrogeological parameter in three zones of unconfined aquifer including permeability coefficient. Applicability, adequacy of SA is tested. 'Variable scale' accelerates convergence of parameter and satisfactory result is obtained, and is obviously better than random method. Commonly, SA can calculate satisfactory result, and similar solution to 'standard value' considering measuring error while initial temperature is 500～650 and temperature-decreasing coefficient is 0.92～0.95. Due to the simulated SA is slow in convergence and time-consuming in compute, conditional methods in generation function, or combination of simulated annealing and genetic algorithms can be used to get more reasonable parameters.