代理模型辅助的初始可行解产生方法

Surrogate Model-Based Initial Feasible Solution Generation Method

采用约束保持法求解单目标约束优化问题时,初始化产生可行解的过程存在计算费时问题。因此提出了代理模型辅助的初始可行解产生方法,采用径向基函数构建代理模型,在初始解的产生过程中,预先使用代理模型估计试验粒子的约束冲突值,若满足约束才进行实际计算,从而减少粒子的评价次数以提高算法效率。采用该方法对多个标准函数进行测试,结果表明,与现有算法相比,所提算法生成相同数量的可行解评价次数会大大减少。该算法可以有效解决利用约束保持法求解单目标约束优化问题时初始解产生耗时的问题。

When the constraint-maintaining method is used to solve the single-objective constrained optimization problem,the process of initializing the feasible solution is time consuming・So the surrogate model-assisted initial feasible solution generation method is proposed.The radial basis function is used to construct the surrogate model.In the initial solution generation process,the surrogate model is firstly used to estimate the constraint conflict value of the test particle.And if the constraint is satisfied,the actual calculation is performed.Therefore,the algorithm can improve the efficiency by reducing the number of evaluations of the particles.Tests on standard functions show that compared with the existing algorithms,when the proposed algorithm produces the same number of feasible solutions,the number of evaluations is greatly reduced.The proposed algorithm can be used to effectively solve the time-consuming problem of the initial solution when solving the single-objective constrained optimization problem by the constraint preservation method.