Sacramento模型的多步骤参数估计方法及应用

Application of Multi-step Parameter Calibration Method with Sacramento Model

为了解决传统优化算法在Sacramento模型参数估计中存在早熟、收敛速度慢、容易陷入局部最优和传统求解过程出现模型模拟吻合度较差等问题。对于人工生成的理想水文资料,分别采用SCE-UA算法、并行遗传算法(PGA)、改进粒子群算法(SMSE-PSO)和提出的免疫克隆选择算法(ICSA)进行参数率定,比较结果选出最优算法,同时,将最优算法与多步骤参数估计方法结合进行实测资料的洪水预报,并比较单步骤与多步骤方法的预报效果。结果表明:ICSA收敛结果更好,效率和精度更高,将其与多步骤参数估计结合提高了洪水预报精度。ICSA算法和多步骤参数估计方法结合为Sacramento模型参数估计提供了一条新途径。

The study was done in order to overcome the disadvantages of traditional optimal algorithm for estimating parameter of Sacramento： early maturity,poor convergence,local optima and poor simulation precision.For ideal data,SCE-UA,PGA,SMSE-PSO and ICSA were applied to estimate parameters and get comparison result,in the meaning time,combining ICSA and multi-step method was used to forecast flood with measured data and get comparison result of multi-step and single-step method.The comparison results showed that ICSA had higher efficiency and precision in parameter estimation and multi-step greatly improved the forecast precision.Combining ICSA and multi-step will be a new method for Sacramento model＇s parameter estimation.