论文部分内容阅读
基于Kriging代理模型的EGO(Efficient Global Optimization)算法的优化精度不高,并且当构建模型的样本数量较大时,算法优化变得耗时。对此提出了结合全局代理模型与局部代理模型的改进优化算法。使用Kriging作为全局代理模型,RBF作为局部代理模型,通过构建、优化多个局部代理模型来获取多个局部较优点,并在线更新Kriging模型,提高模型精度。针对优化耗时问题,提出了样本点遗忘法以及样本点渐进式增加法,使优化时间较EGO大大缩短。通过4个典型测试函数验证,并在收敛精度、稳定性两方面与EGO,PSO算法比较,结果显示两者都优于EGO与PSO,说明该算法具有强寻优性能、强鲁棒性。
The optimization accuracy of the Efficient Global Optimization (EGO) algorithm based on the Kriging agent model is not high, and the optimization of the algorithm becomes time-consuming when the sample size of the constructed model is large. In this paper, an improved and optimized algorithm combining global agent model and local agent model is proposed. Using Kriging as a global proxy model and RBF as a local proxy model, we obtain multiple local advantages by constructing and optimizing multiple local proxy models, and update the Kriging model online to improve the accuracy of the model. Aiming at the problem of optimization and time-consuming, a sample point forgetting method and a sample point incremental increasing method are proposed, which make the optimization time greatly shortened than that of EGO. It is verified by four typical test functions and compared with EGO and PSO algorithms in terms of convergence precision and stability. The results show that both of them are superior to EGO and PSO, which shows that the algorithm has strong optimization performance and strong robustness.