Electric Vehicle Demand Forecasting Using Machine Learning and Charging Infrastructure Suitability Analysis in Thailand

Abstract

This research employs a range of data analysis techniques, including Genetic Algorithm (GA), Artificial Neural Networks (ANN), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), the Grey Model (GM), a hybrid Grey Model with Genetic Algorithm (GM–GA), and the Residual Revised Grey Model with Genetic Algorithm (R2GM–GA), to forecast future demand for electric vehicles (EVs) in Thailand. The findings are intended to support strategic planning and development of EV charging infrastructure. The dataset was obtained from the Automotive Information Center of Thailand and comprises 22,288 records of new vehicle registrations from 2019 to 2023. After filtering for EV-specific entries, 5,320 records were retained for analysis. Additional data on EV charging station locations were collected from official websites and mobile applications of charging service providers in Thailand, yielding 3,174 records. Data on EV charging usage, comprising 3,740 records, were also incorporated; although some information was partially restricted due to confidentiality, the dataset remains sufficiently comprehensive and current for analysis. The selected models were evaluated based on forecasting performance using error-based metrics. The results indicate that the R2GM–GA model achieved the highest predictive accuracy, producing the lowest error and estimates most consistent with observed values. The forecasting error was 2.84%, which is considered highly reliable according to standard accuracy evaluation criteria. Furthermore, the forecasting results were used to examine the relationship between EV adoption and charging infrastructure requirements. The analysis suggests that Thailand would require approximately 53,543 charging connectors and at least 16,818 charging stations nationwide to adequately support daily EV usage. These findings provide practical insights for policymakers and relevant stakeholders in guiding the efficient planning and development of EV charging infrastructure in the future. .