基于异构粒子群算法的LKJ辅助驾驶优化研究

Research on LKJ Auxiliary Driving by Heterogeneous Particle Swarm Optimization Algorithm

目前,LKJ机车主要依靠司机经验进行操纵,存在操纵不合理、晚点、能耗大等问题,针对这一问题,基于分布式的思想,提出一种结合模拟退火的异构粒子群算法,根据线路条件、限速条件等,对LKJ辅助驾驶曲线进行优化。算法以节能、准点为优化目标,首先,根据线路限速信息,利用分布式的思想将列车站间运行区段划分为独立的子区间,对各子区间进行并行优化;其次,通过异构粒子群算法中的学习因子,使算法在执行前期加快全局搜索能力,在算法执行后期加快局部收敛速度;然后,为防止算法过早陷入局部最优,利用模拟退火算法的Metropolis准则,使算法在执行过程中能够跳出局部最优解,从而进一步提升全局搜索能力;最后,以京广线部分区段的实际线路数据对算法进行仿真验证。仿真结果表明:在区间运行时间和限速条件的约束下,本文提出的SA-HPSO算法在迭代至76代时收敛至最佳适应度,该算法具有较强的全局搜索能力和较快的收敛速度,能够在较少的迭代次数内搜索到各子区间最优工况转换点,从而生成能够指导LKJ辅助驾驶的站间优化曲线,对LKJ辅助驾驶曲线优化具有一定的现实意义。

At present,LKJ locomotive mainly relies on the drivers experience to operate,which brings problems such as unreasonable handling,delay and high energy consumption.To solve these problems,based on the distributed principle,a heterogeneous particle swarm optimization algorithm combined with simulated annealing is proposed.According to the line conditions and speed limit conditions,this algorithm aims to optimize the LKJ auxiliary driving curve.Energy saving and punctuality are taken as the optimization objectives.Firstly,according to the speed limit information of the line,the running sections between train stations are divided into independent sub-sections.Each sub-section can be optimized in parallel.Secondly,the learning factor in the isomerized particle swarm optimization algorithm is used to accelerate the global search ability in the early stage and the local convergence speed in the late stage.Thirdly,in order to prevent the algorithm from falling into the local optimum prematurely,the Metropolis criterion of simulated annealing algorithm is used to make the algorithm jump out of the local optimum solution during execution,so as to further improve the global search ability.At last,the algorithm is simulated with the actual line data of some sections of The Beijing-Guangzhou Line.The simulation results show that,in interval operation under the constraints of time and speed conditions,the proposed SA-HPSO algorithm in the iteration to 76 generations of convergence to the best fitness.This algorithm has stronger global searching ability and faster convergence speed,which can search in the fewer number of iterations to the point of each interval of the optimal working conditions.Finally,LKJ auxiliary driving optimization curve is generated to guide the train to operate.The proposed algorism has certain practical significance to LKJ assisted driving curve optimization.