Data-driven Modeling And Analysis Of Electric Vehicle Charging Behaviors And Demands

In order to realize the aim of saving energy and reducing emissions,electric vehicles(EVs)have drawn wide attention from all over the world,and global EV stock is growing rapidly.Many studies related to EVs,for example,the locating and sizing of EV charging stations,and the design of orderly charging strategies,are built on the basis that EV charging behaviors and demands are well aqcuainted.The spatio-temporal distribution of charging demands depends heavily on EVs’ behavior,but the latter is usually stochastic,which means traditional modeling methods depending on empirical assumptions can no longer meet the requirement of accuracy.On the contrary,as data from load measurements and statistics from trip surveys are more widely collected,data-driven methods that are free from assumptions of model parameters can play a better role in reflecting the real characteristics of EV activities.Based on that,the main contents of this paper are as follows:(1)From the angle of charging stations,characteristic variables are specified to reflect the regularity of EVs in the charging process,and the diffusion estimator(DE)is applied to probabilistically model these variables.DE possesses as good nonparametric nature as regular kernel density method,but also has good boundary performance and adaptive smoothing properties,hence a better goodness-of-fit;(2)In order to keep as many distribution characteristics of the probabilistic density function(PDF)obtained through DE as possible,this paper utilizes slice sampling to produce samples,which constructs a Gibbs sampler by introducing an auxilary variable,and therefore converts the sampling from the complicated PDF into the uniform sampling from a two-dimensional area.As for the case study,charging load samples are genetated from charging load PDFs and used to assess the reliability of the modified IEEE 30-bus system considering massive EV charging loads,the result of which verifies the effectiveness and accuracy of slice sampling;(3)From the angle of EVs themselves,by introducing the trip chain theory,key variables closely related to EV trips are modeled probabilistically,and transportation conditions and charging habbits are described quantitively.In the simulation stage,spatio-temporal distributions of charging demands are obtained under different senarios.For a specific senario,the total charging demand is acquired after implementing a simple orderly charging strategy.