用于快速仿真优化的改进差分进化算法及其应用

提出一种改进的差分进化算法,采用一种"位变"方法计算收缩因子,该方法首先根据适应值对种群排序,然后根据各个体的排列位置确定收缩因子;采用正态分布函数对算法参数进行随机扰动来维持种群的多样性;该算法还提出一种新的变异算子,并将其与基本的差分变异算子结合使用以提高算法的寻优精度。经过对多个Benchmark函数的测试、分析和比较,结果表明该算法具有较高的收敛精度和较快的收敛速度。最后将该算法用于火箭发动机涡轮气动优化,以较小的计算成本将涡轮气动效率提高了2.5%。应用结果表明该算法适用于快速仿真优化问题,能有效地节约计算成本。

Modified Differential Evolutionary Algorithm for Fast Simulation Optimization and Its Application

A modified differential evolutionary algorithm(MDE) is proposed. MDE adopts "Position" varying scale factor, which calculates the scale factor linearly according to the position of each individual after arranging by the fitness. To maintain good diversity, normal distribution function is used to disturb the parameters of MDE. A new mutation operator is proposed too, which can enhance the exploration efficiency and precision associating with basic mutation operator. The benchmark function result shows that the algorithm not only has good performance of exploration precision, but also has faster convergence speed than basic DE. At last, MDE is applied in aerodynamic optimization of turbine in LRE, and the result shows aerodynamic efficiency is increased by 2.5% with low computational cost. The application instance indicates good applicability of MDE for simulation optimization problem.