基于改进GH-SOM的可接受航班编队调度优化

Scheduling optimization of acceptable flight formation based on improved GH-SOM

针对可接受航班编队调度优化问题,从最大当量航程约束和最大允许延误约束出发,推导出可接受编队模式的统计判别边界,将编队调度优化问题转化为最优层次聚类问题,利用改进的层次生长型自组织映射(GH-SOM)神经网络实现对可接受编队调度聚类的递归求精。仿真结果表明:与经验判别边界相比,基于可接受编队统计判别边界的识别量提高了92.14%,平均扁率和平均时间同步偏差分别降低了25.00%和26.23%,扁率标准差和时间同步偏差标准差分别降低了12.50%和18.75%;与自组织映射、标准GH-SOM相比,基于改进GH-SOM的识别量分别提高了303.49%、162.87%,平均扁率分别降低了34.25%、22.58%,平均时间同步偏差分别降低了47.06%、36.62%,扁率标准差分别降低了45.10%、6.67%,时间同步偏差分别降低了46.94%、3.70%,因此,可接受编队模式的统计判别边界与改进GH-SOM是有效性的。

Aiming at the scheduling optimization problem of acceptable flight formation,the maximum equivalent range constraint and the maximum allowable delay time constraint were considered,and the statistical decision boundaries of acceptable formation pattern were derived.The formation scheduling optimization problem was transformed into the optimal hierarchical clustering problem,and an improved growing hierarchical self-organizing map（GH-SOM）neural network was used to realize the scheduling clustering recursive refinement of acceptable flight formation.Simulation result shows that compared with the empirical boundaries,the recognition quantity based on the statistical decision boundaries of acceptable formation increases by 92.14%,the mean flat rate and mean time synchronization deviation decrease by 25.00% and 26.23% respectively,and the standard deviations of flat rate and time synchronization deviation decrease by 12.50% and 18.75% respectively.Compared with self-organizing map（SOM）and standard GH-SOM,the recognition quantities based on the improved GH-SOM increase by 303.49% and 162.87% respectively,the mean flat rates decrease by 34.25% and 22.58% respectively,the mean time synchronization deviations decrease by 47.06% and 36.62% respectively,the standarddeviations of flat rates decrease by 45.10% and 6.67% respectively,and the standard deviations of time synchronization deviations decrease by 46.94% and 3.70% respectively.Therefore the statistical decision boundaries of acceptable formation pattern and the improved GH-SOM proposed in this paper are effective.