空中航路点通行能力评估与优化方法

Evaluation and optimization method for air waypoint capacity

航路作为空中交通流的承载体,其瓶颈主要在于交叉航路节点的通行能力。为提高航路网络的运行安全和效率,优化航路网络的拓扑结构,本文提出基于仿真样本-代理模型-变量优化的航路点通行能力优化框架。根据历史数据统计繁忙航路点的高峰时段航班数量及繁忙高度层流量占比情况,提出基于仿真的通用航路点通行能力评估方法;以航路点构型中的航路输入角度、流量占比和输出角度作为输入特征,将航路点通行能力评估方法得到的通行能力值作为样本标签,训练多层感知机代理模型,实现对航路点通行能力评估的快速计算;基于粒子群算法,以调整航路角度和交通流分布作为优化手段,对航路点通行能力进行优化。实验结果表明,构建的航路点通行能力评估方法能较好地反映实际航路点的通行能力状况,使用多层感知机作为代理模型能够满足航路点通行能力预测代理任务的精度要求,平均绝对误差为1.19,优化的航路点通行能力相比优化前提升了37.9%,验证了优化框架的有效性。

As carriers of air traffic flows,bottlenecks in air routes mainly occur in the traffic capacity of cross-route nodes.To improve the operational safety and efficiency of an air route network and to optimize its topological structure,this study proposes an optimization framework for the capacity of air route points based on simulation sample surrogate model variable optimization.Based on historical data,the number of flights during peak hours and the proportion of traffic at busy flight levels at busy waypoints are statistically analyzed,and a simulation-based universal waypoint capacity evaluation method is proposed.According to the input features of input angle,flow proportion,and output angle of the waypoint configuration,the capacity values obtained from the waypoint capacity evaluation method are used as sample labels,and a multi-layer perceptron surrogate model is trained to quickly calculate the waypoint capacity evaluation.Based on aparticle swarm optimization algorithm,the capacity of waypoints is optimized by adjusting the route angle and traffic flow distribution as optimization methods.Experimental results demonstrate that the constructed waypoint capacity evaluation method can better reflect the actual capacity status of the waypoint.Using a multi-layer perceptron as a surrogate model can meet the accuracy requirements of the waypoint capacity prediction surrogate task with an average absolute error of 1.19.The optimized waypoint capacity is increased by 37.9%as compared with that prior to optimization,thus verifying the effectiveness of the optimization framework.