Research on EV ownership and load forecasting methods based on improved Gray Markov models

With the rapid large-scale adoption of electric vehicles (EVs) and the ongoing expansion of charging infrastructure, power system stability is increasingly. To address the limitations in current EV ownership forecasting accuracy and outdated load prediction methodologies, this study proposes an enha...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	International journal of electrical power & energy systems Ročník 173; s. 111323
Hlavní autoři:	Tang, Minan, Sheng, Wenxin, Zhang, Kaiyue, Li, Hanting, Wang, Mingyu
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Elsevier Ltd 01.12.2025 Elsevier
Témata:	Electric vehicle Gray wolf optimization algorithm Grey Markov model Load forecasting Retention projections Electric vehicle Grey Markov model Gray wolf optimization algorithm Load forecasting Retention projections
ISSN:	0142-0615
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Popis
Shrnutí:	With the rapid large-scale adoption of electric vehicles (EVs) and the ongoing expansion of charging infrastructure, power system stability is increasingly. To address the limitations in current EV ownership forecasting accuracy and outdated load prediction methodologies, this study proposes an enhanced grey Markov model for EV ownership prediction, incorporating an improved Grey Wolf Optimization (GWO) algorithm, integrated with a Monte Carlo-based approach for EV load forecasting. First, to better capture the underlying patterns and long-term growth trends in ownership data, a grey model (GM(1,1)) is employed for initial forecasting, followed by residual correction using a Markov chain model to improve prediction accuracy. Second, to resolve the issue of suboptimal background value selection in the conventional grey model, an improved GWO algorithm is applied to optimize key model parameters. Subsequently, probabilistic and statistical models are utilized to simulate critical EV battery parameters and user driving behaviors, with distributional goodness-of-fit tests and sensitivity analyses conducted to ensure robustness and reliability of input data for load forecasting. Finally, the Monte Carlo simulation is implemented to generate probabilistic load profiles for the period 2024–2026. The proposed framework is validated using real-world EV ownership and charging load data from Wuhan City. Case study results demonstrate that the proposed model achieves high predictive accuracy and practical applicability, yielding a minimum Mean Absolute Percentage Error (MAPE) of 2.4036%. Furthermore, the daily load curve peaks consistently at 8:00 p.m., aligning with typical evening charging behavior. [Display omitted] •Combine GWO, GM(1,1), and Markov models to enhance stability with limited ownership data.•The improved GWO better optimizes GM(1,1) background values for enhanced prediction.•Comparative study validates the proposed hybrid model’s effectiveness•EV results by type, considering travel habits & ownership forecasts.
ISSN:	0142-0615
DOI:	10.1016/j.ijepes.2025.111323