Modeling Electric Vehicle Charging Load Using Origin-Destination Data

Background: The increasing adoption of electric vehicles (EVs) presents a significant challenge to the power grid. Accurate energy load prediction is crucial for assessing grid resiliency and determining necessary capacity upgrades or alternative strategies such as vehicle-to-grid integration. Existing methods for modeling EV charging load fall into two categories: data-driven and simulation-based. Data-driven approaches face limitations due to the scarcity of available charging session data. Conversely, current simulation-based charging load prediction models often rely on trip-chain generation models, which typically necessitate detailed road topology and traffic flow data that are frequently inaccessible or time-consuming to acquire. Studies that omit these data encounter difficulties in aligning their findings with actual travel demand reflected by origin-destination patterns (Xiang et al. 2019).

Objective: This study seeks to overcome these challenges by introducing an agent-based trip chain generation model that does not rely on extensive network topology data while still capturing the actual travel demand. The generated trip chain is subsequently employed to predict spatial-temporal EV charging load.

Methodology: In the trip chain generation process, we utilize the Monte Carlo method to sample individual trip parameters from the National Household Travel Survey (NHTS). These parameters exhibit interdependencies; for example, the departure time relies on the specific trip's position within the day's sequence. Consequently, we estimate conditional probability distributions to sample these individual trip parameters for each segment of a trip chain.  These parameters include variables such as trip purpose and driving distance. However, relying solely on NHTS data does not fully capture the intricacies of actual travel demand patterns. To address this limitation, we integrate the NextGen NHTS’s origin-destination (OD) add-on data from Georgia to create a zone-to-zone transition matrix. This matrix is used to identify destinations for each segment of daily trip chains, offering a more accurate reflection of real-world travel patterns.

Numerical experiment: We apply our model to Clarke County, Georgia, utilizing Census Block Groups as the zone units. To expedite the simulation, we assume 100 vehicles generated from each Block Group, resulting in a total of 30,640 trips. This preliminary study focuses on a single scenario. The charging strategy assumes that electric vehicles charge when their battery capacity drops below 30%. Based on literature, the charging power is assumed to be 6.7 kilowatts per hour and electricity loss of 0.35 kilowatts per mile. Energy consumption at each Block Group is estimated on an hourly basis.

Results: Validity checks confirm that the generated trip chains closely match actual demand patterns, with 81.9% of trips being intra-county, compared to the actual 82.3%. Preliminary findings indicate that a majority of energy consumption (54.5%) occurs between 9 PM and 3 AM. To fully validate the proposed methodology, multiple scenarios and external datasets will be incorporated.

References

Xiang, Yue, Shuai Hu, Youbo Liu, Xin Zhang, and Junyong Liu. 2019. “Electric Vehicles in Smart Grid: A Survey on Charging Load Modelling.” IET Smart Grid 2(1):25–33. doi: 10.1049/iet-stg.2018.0053.