基于多粒子群协同的城轨列车速度曲线多目标优化  

Multi-objective Optimization of Speed Profile of Urban Rail Train Based on Multiple Particle Swarms Co-evolutionary

作　　者：徐凯[1,2] 杨飞凤 涂永超 吴仕勋 XU Kai;YANG Feifeng;TU Yongchao;WU Shixun(Information Science and Engineering College,Chongqing Jiaotong University,Chongqing 400074,China;Chongqing Public Transport Operation Big Data Engineering Technology Research Center,Chongqing 400074,China)

机构地区：[1]重庆交通大学信息科学与工程学院,重庆400074 [2]重庆市公共交通运营大数据工程技术研究中心,重庆400074

出　　处：《铁道学报》2021年第2期95-102,共8页Journal of the China Railway Society

基　　金：重庆市教委科学技术研究项目(KJQN202000703);重庆市研究生教育教学改革重点项目(yjg172004)。

摘　　要：在满足安全原则和各类约束条件下,为实现城市轨道交通列车运行能耗低、行驶时间短和停车精度高三个目标,建立了列车运行控制模型。在Pareto原理基础上,考虑列车运行中的各类工况序列,提出一种协同进化的多目标混沌粒子群算法(CMOCPSO),用于优化列车的自动驾驶速度曲线。此算法框架结构分为上下两层,其中基础群位于下层,以目标引导法实现全局搜索,在尽可能发掘边沿解的条件下让解分布更为均匀;精英群则位于上层,将其加以扰动后实现局部精细搜索。此外,为了改良算法的各项性能指标,进一步引入双外部档案来实现上下层的双向交互,在恰当的通信周期参数下,构成一个多方协同的高效寻优粒子群体。通过仿真试验验证表明:相比于多目标粒子群MOPSO算法,所提算法在多样性及收敛性上均具有明显的优势。为获取多种工况下优质的列车自动驾驶曲线,采用模糊隶属度法对Pareto前沿解集进行了筛选。In order to meet the safety principles and various constraints,a train operation control model was established to achieve the three goals of low energy consumption,short running time and high stopping accuracy of urban rail train.On the basis of Pareto principle and considering multiple working conditions of train operation,this paper proposed a co-evolutionary multi-objective chaotic particle swarm optimization algorithm(CMOCPSO)to optimize the automatic driving speed profile of train.The framework of this algorithm was divided into upper and lower layers.The basic group located in the lower layer used the goal-guided method to achieve a global search,and made the distribution of solution space more uniform under the condition of discovering edge solutions as much as possible.The elite group located in the upper layer was disturbed to realize local fine search.In addition,in order to improve the different performance indicators of the algorithm,two external archives were further introduced to realize the two-way interaction between the upper and lower layers.A multi-dimensional collaborative and efficient optimization particle group was formed with the appropriate communication cycle parameters.The simulation results show that the proposed algorithm has more obvious advantages on diversity and convergence than the multi-objective particle swarm algorithm(MOPSO).Ultimately,the fuzzy membership method was adopted to screen the Pareto frontier solution set in order to obtain optimal automatic train operation speed profile with multiple driving patterns.

关 键 词：城轨列车 多种工况序列 多目标 多粒子群协同 速度曲线 

分 类 号：U284[交通运输工程—交通信息工程及控制]