Citation

Siow, Jat Shern (2025) Forecasting and optimisation of electric vehicle charging systems using XGBOOST Machine Learning model. PhD thesis, Multimedia University.

Full text not available from this repository.

Abstract

Electric vehicle (EV) charging optimisation is increasingly important for managing rising energy demand and supporting Malaysia’s transition to sustainable mobility. This thesis proposes an AI-based smart charging optimisation framework developed using real-world data collected from four 47 kW DC chargers at The Curve, Mutiara Damansara. A dataset of 1,000 charging sessions over seven days was compiled, containing parameters such as energy consumption (kWh), charging duration, voltage, current, State of Charge (SOC), and time-of-use. All data were cleaned, normalized, and preprocessed to ensure reliability and consistency. Six machine-learning models Random Forest, XGBoost, Gradient Boosting Regressor, Support Vector Regression, LightGBM, and LSTM were trained and tested using an 80:20 split with repeated runs to ensure robustness. The XGBoost model achieved the best performance, obtaining an RMSE of 0.094, MAE of 0.067, and MAPE below 4%, significantly outperforming the other models. The predicted charging demand closely matched actual patterns, demonstrating strong accuracy and stability. These results were integrated into an optimisation framework that improved charging load management and reduced peak demand fluctuations in simulated scenarios. A case study at Seringin Residence validated the model’s practical application, showing improved load distribution and potential cost savings from more efficient charging scheduling. The findings confirm that AI-driven forecasting can support residential and commercial EV charging infrastructures by enhancing operational efficiency and energy planning. The study acknowledges limitations such as reliance on a single-site dataset, a short data collection period, and the absence of renewable energy integration. Despite these constraints, the research makes a novel contribution by providing Malaysia’s first real-world AI-based EV charging optimisation framework. Future work will focus on expanding datasets, incorporating traditional optimisation baselines, integrating solar and ESS systems, and scaling the framework for broader implementation.

Item Type:	Thesis (PhD)
Additional Information:	Call No.: TL220.5 .S56 2025
Uncontrolled Keywords:	Battery charging stations (Electric vehicles)
Subjects:	T Technology > TL Motor vehicles. Aeronautics. Astronautics > TL1-484 Motor vehicles. Cycles
Divisions:	Faculty of Artificial Intelligence & Engineering (FAIE)
Depositing User:	Ms Nurul Iqtiani Ahmad
Date Deposited:	19 Jan 2026 05:29
Last Modified:	19 Jan 2026 05:29
URII:	http://shdl.mmu.edu.my/id/eprint/15193

Downloads

Edit (login required)