Demand Modeling of Electric Car Fast Charging Station

Abstract

This paper presents the modeling and simulation of power loads due to plug-in electric vehicles’ (EVs) charging events at a dc fast charging station. Two algorithms for modeling the loads are introduced and compared, based on sampling and one based on statistical distribution build from the sample database. Simulation of load versus time was performed using a horizon of 7 days using both techniques. The cause for the difference in the result of these two approaches is explored. Regardless of the method used, the results show that a dc fast charging station with 6 fast chargers potentially serving 700 plug-in EVs generally gets 105 charging events per day with a peak load of 375 kW.

Index Terms--Demand forecasting, load modeling, electric vehicles, dc fast charging station.

Introduction

Climate change, energy and environmental issues are the long-term problems faced by society. In the transportation sector, plug in electric vehicle (EV) technology is evolving to help relieve greenhouse gas emissions, since electric vehicles are greener and more efficient than traditional vehicles. Projections by the Bloomberg New Energy Finance indicate that by 2040, EVs will account for 54% of all new sales [1].

With the number of EVs on the road growing, the demand for EV charging is on the rise. Electric utilities are now installing and operating their own publicly available charging stations and taking a leading role to ensure there are enough charging stations to support the anticipated significant growth in EV ownership. These programs include a handful of strategically placed direct current (dc) fast chargers. Presently the most popular dc fast chargers (three most popular types are: CHAdeMO with a power rate of 50 kW; SAE, 60 kW; and Tesla, 120 kW), also known as Level 3 charging units, can offer 60 to 100 miles of range for an electric car in just 20 minutes of charging.

The use of dc fast chargers, in concentrations of 6 or 10 in a fast charging station, brings challenges to utilities with higher levels of demand and energy consumptions and near instantaneous changes in power demand. What utilities prefer is to not have the additional consumption occur during peak periods of electricity demand (generally hot summer afternoons and cold winter days). To better handle the challenge that utilities are facing, it is necessary to model the EV driving and charging behaviors and their impact on distribution systems.

Several researchers have explored statistical methods applied to modeling power load due to EV fast chargers [2]- [5]. Ref. [2] characterized potential usage patterns of a single fast charging station with two charging ports by using a decision tree to simulate the charging behaviors of EVs. They made some important assumptions of people’s charging behavior in the decision tree that are used later in this paper. Ref. [3] simulated work of a fast charger’s activity by exploiting empirical data associated with EV drivers’ behaviors. They assume each EV takes only one trip a day. Ref. [4] used Bayesian temporal distribution models derived from real-world driving data in Michigan and Markov-chain technique to model EVs’ driving schedules. In [5], charging demand is modeled in different charging locations and different control strategies using probabilistic distributions derived from U.K. National Statistics. Use of real-world vehicles’ travel data is becoming the norm [2]-[5]. In this paper, we incorporated more recent and nationwide travel data from National Household Travel Survey (NHTS) 2017 [6] collected by US Department of Transportation and develop statistical travel models that allow more than one trip per day. In this paper we will use this database to develop a multi-trips model to better represent real world driving behavior.

The objective of this research is to develop a model of demand for fast charging stations. In section II, we develop two methods (based on database sampling and on sampling generated from statistical distributions) to model EV driving behaviors and integrated them into our algorithm of modeling EV charging behavior. In section III, we develop simulated demand for a dc fast charging station using the two approaches described in section II and show the station demand versus time. We then explore the origin of the differences in the results.

Cite this page