基于代理模型的水文模型参数多目标优化!

Multi-objective Optimization for Hydrological Models Using Surrogate Modeling

针对传统多目标优化算法存在计算复杂且效率偏低的问题,提出了一种基于代理模型的多目标优化方案。以淮河大坡岭水文站以上流域为例,采用多元自适应回归样条方法构建新安江模型参数与不同目标的响应曲面关系,进而估计参数的近似Pareto解集。采用4种目标函数(总水量误差系数、均方根误差、高水流量误差系数和低水流量误差系数)和4种模型精度评价指标(Nash-Sutcliffe效率系数、洪峰流量相对误差、径流深相对误差和峰现时间误差)评定模型优化结果,选择10场洪水过程和4种不确定性评价指标估计Pareto解集的模型预测区间特征。结果表明,代理模型可有效降低模型评估与优化过程中的计算消耗,为实现多目标优化的高效性奠定了基础。此外,不确定性分析结果也进一步验证了方法的有效性和结果的可靠性,为复杂模型参数优化与不确定性评估提供了参考。

A new multi-objective optimization scheme based on surrogate modeling was proposed. Taking the Dapoling catchment as a case study, the response relationship between the parameter of Xin＇ anjiang model and different objectives was constructed based on multivariate adaptive regression splines,to estimate the Pareto sets or non-dominant solutions. Four objective functions of overall water balance error, root mean square error, relative error of peak flows, and low flows were used to optimize model parameters, and four evalu- ation criteria of Nash-Sutcliffe efficiency coefficient （NSE） ,relative error of peak flow and runoff volume （REPF and RERV） ,and time error of peak flow （TEPF） were selected to quantify the goodness-of-fit of observations against simulation model calculated values. In addition, four uncertainty criteria were applied to assess the hydrological uncertainty ranges with the Pareto solutions for ten flood e- vents. Results demonstrated that the surrogate-modeling based method increases the feasibility of applying parameter optimization to computationally intensive simulation models via reducing the number of simulation runs. Simultaneously, uncertainty analysis results also revealed that the proposed method based On surrogate modeling is high efficiency and easy to operate. Thereby, the method is feasible for practical operations for complex simulation models in model calibration and uncertainty analysis.