基于Condor 的模型参数自动识别实例研究

基于物理机理的分布参数模型广泛用于评价和模拟地下水流和水力响应。传统的确定数值模型参数的试错法(人工方法),过多地依赖于建模者的经验和主观判断,并且是一个非常耗时的过程。引进一种带约束、非线性、全局收敛且无需求导的Condor优化算法,以理想的地下水模型为例,实现了模型参数自动识别的完整过程,并与广泛采用的遗传算法收敛效果进行对比。结果表明,相比于遗传算法,Condor算法受参数初值影响小,寻优效率提升显著。

Case study of parameter auto2calibration of distributed parameter model based on Condor algorithm

Phy sically2based distributed parameter models hav e been widely used to evaluat e and pr edict g roundw ater flow and hy2 draulic respo nse. The traditio na l tr ial2and2erro r approach for ca librat ing the numerical model par ameters depends on the ex per i2 ence and subject ive assessment of the modeler and can be v ery time2co nsuming. In this paper , the Co ndor a lgo rithm, a co nstr ain2 ed, non2linea r, and der ivative2free o ptimizer, is intr oduced into parameter auto2calibration o f a synthet ic g ro undwater mo del. It a2 chieves a complete pro cess of par ameter auto2calibration o f the model. T he co nv erg ence effects ar e compared w ith those using the g enet ic alg or ithm, w hich sugg ests that the Condo r alg or ithm is less affected by the initial par amet ers and impr ov es the o pt i2 mizing efficiency sig nificantly compar ed w ith genetic algo rithm.