概念性流域水文模型参数多目标优化率定

针对传统基于单一目标的水文模型参数优化率定方法不能充分挖掘水文系统不同动态行为特征的缺陷,提出一种多目标文化混合复形差分进化算法(Multi-objective Culture Shuffled Complex Differential Evolution,MOCSCDE)用于求解水文模型参数多目标优化问题。MOCSCDE算法将混合复形进化算法(Shuffled Complex Evolution,SCE-UA)置于文化算法(Cultural Algorithms,CA)进化的框架中,利用种群进化过程中提取的各种知识指导算法的运行,提高算法的运行效率,同时考虑到SCE-UA中单纯形算子不能充分利用种群个体信息的不足,采用全局搜索能力强的差分进化算法(Differential Evolution,DE)替代单纯形算子,可以更加充分利用种群个体信息进行演化计算,进一步提高算法的计算效率。将MOCSCDE算法应用于概念性水文模型——新安江模型的参数多目标优化率定,并与NSGA-Ⅱ和SPEA2算法进行对比分析,结果表明MOCSCDE算法的收敛性和分布性均优于NSGA-Ⅱ和SPEA2,可为水文预报提供更为全面可靠的参数组合决策依据。

Multi-objective optimization for conceptual hydrological models

Traditionally,conceptual hydrological models are calibrated by optimizing the model parameters using a single objective function,which makes is difficult to properly consider all behavior of a natural hydrological system.To circumvent this problem,we propose a novel Multi-objective Culture Shuffled Complex Differential Evolution(MOCSCDE) algorithm for model calibration based on multi-objective functions in this paper.The MOCSCDE algorithm takes the Cultural Algorithms(CA) as the evolving framework and adopts the Shuffled Complex Evolution(SCE-UA) in the population space.This evolution strategy can make use of the problem-solving knowledge obtained along with the evolution process to guide the algorithm toward the optimization direction.Meanwhile,as the simplex search operator in the SCE-UA algorithm cannot use the whole information of the individuals,a Differential Evolution(DE) algorithm is employed to serve as a substitute of the simplex search operator.The DE algorithm can more thoroughly utilize the information of the population and enhance the search efficacy of the algorithm.The performance of the MOCSCDE algorithm is tested for parameter estimation of a conceptual hydrological model(Xinanjiang model).The MOCSCDE results are compared to those obtained with the NSGA-II and SPEA2 algorithms.It can be found that the MOCSCDE can get better convergence and spread performance,and can provide more reliable and comprehensive solutions in practical applications.