A Bayesian approach to modeling fast chargers functionality for grid frequency support

As governments around the world commit to achieving net zero emissions, the upcoming years will witness a significant increase in fully electric vehicles (EVs), predominantly supported by DC fast charging (DCFC) infrastructure. While DCFCs are primarily used to provide rapid and convenient charging, their widespread adoption highlights the need to integrate them into power system operations, such as frequency control. Thus, it is crucial to estimate the level of support that DCFCs can provide for frequency control, especially in future EV adoption plans subject to data uncertainty. Existing methods, including individual EV modeling and clustering-based approaches, fall short due to high data requirements and incompatibility with transmission network models. Aggregated modeling and averaging techniques, while simpler to apply, overlook critical factors such as EV owner preferences and primarily focus on the vehicles rather than the chargers. Additionally, these methods are primarily designed for low-power chargers intended for prolonged charging sessions with more predictable plug-in patterns and are not suitable for DCFCs, where the behavior of EV owners is more dynamic and subject to higher uncertainties. EV owners are also more sensitive to meeting their expectations when using DCFCs rather than low-power chargers. To address these limitations, this paper developed a Bayesian probabilistic equivalent capacity model for DCFCs. This model uniquely incorporates deep discharge vulnerability, mobility requirements, and owner preferences, providing a comprehensive assessment of DCFC frequency support. A novel concept called mileage loss (ML) is also introduced, enabling DCFCs to contribute to frequency control. It also allows system operators and EV aggregators to analyze the risk-taking and risk-averse behaviors of EV owners in vehicle-to-grid (V2G) mode. Furthermore, the proposed model is validated using the Australian frequency regulation framework, demonstrating its scalability and applicability. The case study results demonstrate that the proposed modeling method can achieve a high level of accuracy, with an estimation precision of up to 97.7 % for the aggregated DCFC power capacity.