Agent–cellular automata model for the dynamic fluctuation of EV traffic and charging demands based on machine learning algorithm

Abstract

Electric vehicles (EV) comprise one of the foremost components of the smart grid and tightly link the power system with the road network. Spatial and temporal randomness in electric charging distribution will exert negative impacts on power grid dispatch. Existing research focuses mainly on mathematical inferences from statistical data, and the dynamic movement of an individual vehicle traveling in a traffic system is rarely taken into account. Machine learning algorithm can take the EV dynamic condition into consideration. Based on machine learning algorithm, this paper proposes a charging demand simulation method based on the Agent–cellular automata model to describe the changes in location and the state of charge of a moving EV. CRUISE software is used to analyze power consumption in different scenarios. Then, the Monte Carlo algorithm models the dynamic fluctuation of EV traffic and charging demands. Case studies are conducted on a typical composite system consisting of a 54-node distribution system and a 25-node traffic network, and the simulation results demonstrate the effectiveness of the proposed method.

Appendices

Appendix A

See Tables 3, 4, 5, 6, and 7.

Table 3 Peak load data of the 54-node distribution system

Table 4 Load rates of substations

Table 5 Energy losses of the distribution system

Table 6 Branch data of the 54-node distribution system

Table 7 Weights of nodes in the 25-node traffic system

Appendix B: Calculation of the charging load in a charging station

When applying the method in this paper to predict the charging load of EVs at a given charging station, the location selections made by EV users should be taken into account.

By considering charging time and charging cost, we can predict the charging decision for each vehicle agent once the demand occurs. The charging power and the maximum charging capacity for each time period can be determined using the formula below, and this method can be applied in the multiobjective optimization model for coordinated planning of the power distribution system and EV charging network.

$$\begin{array}{*{20}l} {P_{k,t}^{\text{FCS}} = \rho_{k,t} z_{k} p^{\text{FCS}} } \hfill & {\quad \forall k \in \varOmega^{K} ,\forall t \in T} \hfill \\ {P_{k}^{\text{FCS,max}} = z_{k} p^{FCS} } \hfill & {\quad \forall k \in \varOmega^{K} } \hfill \\ {\rho_{k,t} = \frac{{\lambda_{k,t} }}{{z_{k} \mu }}} \hfill & {\quad \forall k \in \varOmega^{K} ,\forall t \in T} \hfill \\ \end{array}$$

where $\lambda_{k,t}$ is the average arrival rate (the number of vehicles arriving at a charging station for charge per unit time) of vehicles waiting for charge at charging station k in time period t. $z_{k}$ is the number of charging service devices at charging station k. $P_{k,t}^{\text{FCS}}$ and $P_{k}^{\text{FCS,max}}$ are the charging power and maximum charging capacity of charging station k in time period t, respectively. $p^{\text{FCS}}$ is the nominal charging power. $\rho_{k,t}$ is the average equipment usage rate of charging station k during time period t.