Electric Taxi Operation Simulation Research Based On Multi-Agent Technology

With the decrease of global fossil energy stock,the energy crisis is becoming more and more serious.As an effective mitigation measure,electric vehicle(EV)has attracted wide attention because of its remarkable advantages,such as high efficiency,energy saving,low noise and zero emission.In the future,large-scale electric vehicles have become a trend.However,the access of large-scale electric vehicles will bring a lot of challenges to the operation and dispatch of the power grid.If not properly handled,it will even affect the safe and stable operation of the power system.The electric taxi has the highest driving density and the most complex behavioral characteristics among many types of electric vehicles,and its charging characteristics can best represent the charging characteristics of electric vehicles.Therefore,it is of great significance to study the charging characteristics of electric taxis.Firstly,based on multi-agent technology,a multi-agent simulation system for electric taxi operation is built on JADE platform,which includes eight agents,such as electric taxi agent,power grid agent and charging station agent.The functions of each agent model and the interaction process between agents are introduced in detail.According to the characteristics of different types of agents,it divides several modules and distributes them to run on different hosts to realize the distributed simulation of electric taxi operation.In addition,aiming at charging station agent,this paper designs a charging station M/M/c/N/?queuing model based on queuing theory,which considers the distribution of arrival time and charging time of passengers,and adds a charging guidance strategy considering the spatial distribution of charging load.Secondly,after synthesizing and comparing several main learning algorithms,this paper in-troduces the Q_?learning algorithm and establishes the learning model for the behavior decision of electric taxi,including the reinforcement learning algorithm model from several perspec-tives:state space,behavior decision space,behavior strategy selection,probability updating and reward and punishment function.The effectiveness of the learning algorithm is verified by simulation,and the effectiveness of charging and guiding strategy is studied on this basis.The results show that the introduction of charging guidance strategy is helpful to balance the spatial distribution of charging load of electric taxis.At the same time,it can improve the income of electric taxi owners and reduce the average queuing time.Finally,this paper studies the location and capacity of charging station.Based on the multi-agent simulation system built in this paper,the effects of adding charging station,changing the location of charging station and changing the capacity of charging station on the spatial distribution of charging load,charging queuing time and charging seeking distance of electric taxi are analyzed and compared.It provides a new practical method for the experimental method of charging station location and capacity determination.