基于代理模型的碾压混凝土坝坝体渗控结构多目标优化

Multi-Objective Optimization of Seepage Control Structures of RCC Dams Based on the Surrogate Model

针对现有碾压混凝土坝坝体渗控结构优化研究中常用的单目标、单参数的优化方法难以获得同时满足多个目标要求的渗控结构设计方案,无法充分考虑不同设计参数之间的关联性以及计算效率低的不足,提出了基于代理模型的碾压混凝土坝坝体渗控结构多目标优化方法.其主要包括:①建立以渗控结构设计参数为设计变量,以防渗层最大水力梯度、溢出点高程以及渗控结构造价为优化目标的碾压混凝土坝坝体渗控结构多目标优化数学模型;②提出基于多目标粒子群算法与代理模型的模型求解方法,其中多目标粒子群算法应用于快速求解Pareto解集,而代理模型应用于建立设计参数与优化目标之间的近似关系式,用以替代优化过程中求解耗时的渗流数值模型,从而提高求解效率,同时考虑不同设计参数之间的关联性.利用本文方法对某碾压混凝土坝坝体渗控结构进行多目标优化,并通过与传统方法进行对比验证了该方法的有效性与优越性.

The single-objective and single-parameter optimization method is commonly used in the optimization of seepage control structures of roller-compacted concrete(RCC)dams,for which a seepage control structure design scheme that simultaneously satisfies multiple target requirements is difficult to obtain. Moreover,the method cannot fully consider the correlation between different design parameters,and its computational efficiency is low. In this study,a multi-objective optimization method based on the surrogate model was proposed for the seepage control structures of RCC dams. In the proposed method,①a mathematical model,which utilizes the design parameters of the seepage control structure as the design variables and the maximum hydraulic gradient of the impervious layers, overflow elevation,and seepage control structure cost as the optimization targets,was established for the multiobjective optimization of the seepage control structures of RCC dams and ② a model solution method based on the multi-objective particle swarm optimization(MOPSO)algorithm and the surrogate model,in which the MOPSO algorithm was applied to rapidly solve the Pareto solution set,was proposed. Moreover,the surrogate model was used to establish an approximate relationship between the design parameters and the optimization objectives to replace the time-consuming seepage numerical model in the optimization process. Thus,the solution efficiency was improved and the correlation between different design parameters was fully considered. The proposed method was applied to optimize the seepage control structure of a practical RCC dam,and its effectiveness and advantages were validated by comparing with traditional method.