吸附性土壤溶质运移参数识别的粒子群-差分算法

非饱和土壤水分和溶质运移参数(导水率、扩散率、水动力弥散系数)的识别是进行数值模拟的关键.基于垂直一维非饱和吸附性土壤水分-溶质运移方程,以溴化钾为入渗溶液,分别以土壤含水率和溶质溶度的实测值与计算值标准差最小为优化目标,建立水分-溶质运移参数识别的多目标优化模型.应用权重系数法,将这一多目标优化问题转化为单目标优化问题,采用动态权重、异步时变学习因子的粒子群算法对模型求解.通过土柱实验,测定了220、380和780 min三组时刻的土壤含水率以及钾离子浓度的空间分布值,以前两组数据识别参数,第三组数据进行验证.结果表明入渗780 min权重系数为0.5时含水率以及钾离子浓度的计算值和实测值相关系数分别为0.977和0.952,标准差分别为0.007 4、2.369,实测值和计算值吻合较好,这表明粒子群识别水分和溶质运移参数是可行的.数值模拟表明,权重越大含水率实测和计算值的相关系数越大而浓度实测值和计算值的相关系数越小.

Using particle swarm-finite difference algorithm to identify transport parameters of water-solute in adsorptive soil

Abstract It is the key to numerical modeling of water and solute transportation in unsaturated soil to identify parameters of the transportation. Based on the vertical one-dimensional equation for transportation of water-solute in unsaturated adsorptive soil, a multi-target optimized model for identification of parameters of water-solute transportation was established, using KBr as infiltration solution and the minimum standard deviations between measured and observed values of water content and solute concentration as targets of optimization. The multi-target optimization issue was converted into a single-target one, with the weight coefficient method. The particle swarm optimization (PSO) algorithm of dynamic weight and asynchronous time-varying learning factors was used to solve the model. A soil column experiment was conducted in lab to determine water content and spatial distribution of K at the time of 220,380 and 780 minutes, separately. The data of the first two time groups were used for identification of parameters, and those of the third group for validation. Results show that the correlation coefficient between measured and calculated values of water content and K concentration was 0.977 and 0.952, when the weight coefficient of the treatment of 780 min infiltration was 0.5, demonstrating that they are well dovetailed and that it is feasible to use PSO to identify soil transportation parameters of water and solute in the soil. The numerical simulation indicates that the higher the weight, the higher the correlation coefficient between measured and calculated values of water content and the lower the correlation coefficient between measured and calculated values of K concentration.