Research On Rebalancing Of Bike-sharing System Based On Model Predictive Control

In recent years,with the rapid development of global urbanization and industry,climate warming has intensified,and the international community has paid more and more attention to climate change and urban mobility.Bike-sharing has become popular in many regions and cities around the world,thanks to its low carbon,environmental protection,ease of operation,and health benefits.It not only provides urban residents with an environmentally friendly means of travel to ease traffic congestion,but also forms a good connection with other means of travel,such as subway and bus,on the "last mile" issue.However,the irregular behavior of users and the unbalanced demand between stations will affect the normal use of users,greatly reduce the operating efficiency of the system,if not solved,even block the original road system.The root cause of this imbalance is the lack of efficient rebalancing strategy,so it is important to study the rebalance problem of bike-sharing system.In this thesis,the network dynamic model of the bike-sharing system is set up with the method of demand forecast and balance dispatch of the bicycle share station.The method of constructing the bicycle sharing network is put forward.A quantitative analysis is made on the influence characteristics of the numb of shared bicycle stations,the usage and demand characteristics of users and station are analyzed based on the historical data,and a demand forecast model of shared bicycle stations is put forward.Rebalancing strategy based on maximizing scheduling efficiency is proposed,and a rebalancing equalization control algorithm is designed according to the dynamic model of shared bicycle network and the characteristics of objective function.Finally,simulation and experimental verification are carried out.The main work of the paper is as follows:(1)Analyzing the use characteristics of users and stations from the perspective of big data,summarizing the rules of using shared bikes and the reasons for the imbalance between stations.According to the travel rules of users,the demand forecast of shared bike site borrowing and repaying is forecasted based on the time series prediction model.Therefore,the disturbance caused by the user’s renting and returning operation is solved.(2)The construction principle of shared bicycle network is put forward,a dynamic model of share bicycle network is established based on graph network theory,and the rebalancing problem of shared bicycle is abstracted as a dynamic equilibrium problem of network.The user’s renting and returning operation is regard as the disturbance in the network equalization process.A multi-objective equalization efficiency in the rebalance process is proposed,including the site equalization index and the scheduling economic cost.The equilibrium problem of shared bicycle network is transformed into an efficiency optimization problem to be solved.(3)In view of the discrete,multi-constraint,non-linear,strong interference and strong randomness characteristics of the bicycle-sharing system,the goal is to maximize the equilibrium efficiency of multi-objective rebalancing.This paper presents an optimal strategy of forecasting control based on site demand forecasting,and proves the convergence and stability of the model predictive control method proposed in this paper.The effectiveness of the model and model predictive control equalization algorithm is verified by testing the solution in ideal and practical networks of different sizes and different connection networks.