基于客流大数据的公交发车班次仿真优化

Simulation and Optimization of Bus Departure Frequency Based on big Data of Passenger Flow

通过优化发车班次,可以有效提高公共交通的服务质量从而提升吸引力。发车班次的优化需要综合考虑车辆的行驶与客流到达的不确定性规律,很难使用解析的数学模型进行建模。本文通过对客流大数据进行分析,构建了一个能够再现现实随机客流的客流生成仿真模型,结合公交运营过程产生的GPS轨迹、进出站信息等数据,建立了整条公交线路的仿真程序。并将响应面分析法中的最速上升思想融入机器学习领域的贝叶斯优化中,建立SA-BO算法模型以提升优化效率。以客流仿真模型和基于SA-BO算法的仿真优化构成了整体的基于客流大数据的公交发车班次仿真优化。结果显示基于客流大数据的仿真模型MAPE指标为0.577%,能够良好的再现现实随机客流;SA-BO算法相比较传统BO算法优化效果提升了26.7%。

The quality of public transportation services can be effectively improved to increase its attractiveness by optimizing the departure frequency.The optimization of departure frequency requires comprehensive consideration of the uncertainty laws of vehicle travel and passenger flow arrival,and it is difficult to use analytical mathematical models for modeling.By analyzing the big data of passenger flow,this paper first built a passenger flow generation simulation model that can reproduce the actual random passenger flow.Then,based on the GPS trajectory,inbound and outbound information generated by the bus operation process,a simulation program for the entire bus line was established.Finally,the fastest-rising ideas from the response surface analysis method were incorporated into Bayesian optimization in the field of machine learning,and the SA-BO algorithm model was established to improve the optimization efficiency.The passenger flow simulation model and the simulation optimization based on the SA-BO algorithm constitute the overall simulation of bus departure frequency optimization based on passenger flow big data.The results show that the MAPE index of the simulation model based on big data of passenger flow is 0.577%,which can reproduce realistic random passenger flow;SA-BO algorithm improves the optimization effect by 26.7% compared with traditional BO algorithm.