序列Kriging模型的黑箱约束全局优化方法

A Black-box Constrained Global Optimization Method for Sequential Kriging Model

  • 摘要: 针对黑箱约束优化中可行采样点难以获取、优化效率低、收敛速度慢、难以有效平衡全局和局部搜索行为等缺点, 提出一种序列Kriging模型的黑箱约束全局优化方法。在没有初始可行采样点的情况下, 所提出的方法能够快速有效地探索出富有前景的可行采样点, 并在满足所有约束条件下通过高效稳定且可靠的加点采样准则获取全局最优可行解。基准测试函数和燃料电池汽车能量控制策略的仿真实例验证了所提出方法的有效性和实用性。

     

    Abstract: Aiming at the shortcomings that it is difficult to obtain feasible points, low optimization efficiency, slow convergence speed, and difficult to effectively balance the global and local search behaviors, a black- box constrained global optimization method of sequential Kriging (BCGO-SK) is proposed. In the absence of initial feasible sampling points, the proposed method can quickly and efficiently explore promising feasible points, and obtain the global optimal feasible solution by satisfying all constraints under the efficient, stable and reliable sampling point. The test results on benchmark functions and a simulation of fuel cell vehicle energy control strategies verify the effectiveness and practicability of the proposed method.

     

/

返回文章
返回