基于MOSCEM-UA的水文模型多目标参数优化研究

Research on Multi-objective Parameter Optimization of Hydrological Model Based on MOSCEM-UA

水文模型校准经验表明,单一目标函数不足以准确预报观测数据的所有特征。为解决这一问题定义多个目标函数用于度量系统模拟的不同方面,并根据多标准优化确定pareto非支配最优解。提出高效的MOSCEM-UA算法能够解决水文模型的多目标优化问题,MOSCEM是对SCE-UA全局优化算法的一种改进,它使用pareto非劣解概念而不是直接对单目标函数评估,pareto解集将最初的点群发展为一组解决源方案的稳定分布。通过对流域场次洪水的模拟,进行单目标与多目标模拟方案对比,从而选出较优的多目标优化的模型参数。结果表明,采用多目标最优非劣解准则选取最优解使筛选结果更具准确性和可靠性,为水文模型参数优化率定提供了较为综合的方法。

Hydrological model calibration experience shows that a single objective function is not sufficient to accurately predict all features of the observed data. To solve this problem,multiple objective functions are defined to measure different aspects of the system simulation,and the pareto non-dominated optimal solution is determined based on multi-criteria optimization. This paper proposes that the efficient MOSCEM-UA algorithm can solve the multi-objective optimization problem of the hydrological model. MOSCEM is an improvement to the SCE-UA global optimization algorithm. It uses the pareto non-inferior solution concept instead of the direct entry single objective function evaluation,pareto. The solution set develops the initial point group as a set of stable distribution solutions. Through simulation of watershed floods,single-target and multi-target simulation schemes are compared to select the best multi-objective optimization model parameters. The results show that using the multi-objective optimal non-inferior solution criterion to select the optimal solution makes the screening result more accurate and reliable,and provides a more comprehensive method for parameter optimization of hydrological model.