AI-Powered Battery Charging System Using Machine Learning Algorithms for EVs

Abstract

This paper presents an intelligent charging system powered by AI to boost performance and improve the State of Charge (SoC) and State of Health (SoH) of batteries in Electric Vehicles (EVs). Despite advancements in design and electrical specifications, the significance of effectively charging EV batteries is essential for ensuring safety, efficiency, and longevity. The prototype features a Raspberry Pi as its main controller, along with sensors that monitor key battery metrics including voltage, current, temperature, and pressure. The proposed system employs machine learning algorithms to forecast critical battery metrics such as SoC, SoH, and charging capacity based on the collected sensor data. These metrics are gathered in real-time, allowing the implementation of adaptive charging strategies that maintain consistent battery performance while adjusting to real-time variations in battery conditions to avoid issues like overheating, overcharging, and deterioration. This method not only prolongs battery life but also minimizes safety hazards. The proposed system is scalable, cost-efficient, and superior to traditional charging systems. Furthermore, it can be paired with renewable energy sources to further improve energy efficiency. This paper aims to promote the creation of an environmentally sustainable and smart battery charging system for electric vehicles.

Introduction

Overview

In 2024, Electric Vehicles (EVs) made up 20% of global car sales. Despite advances in EV technology, battery charging systems remain a critical focus due to their role in battery health, safety, and performance. Traditional charging methods (like CC/CV) are rigid and can cause overcharging and faster battery degradation.

Challenges with Conventional Charging

• Fixed algorithms don’t adapt to real-time battery conditions.

• Frequent charge/discharge cycles degrade State of Charge (SoC) and State of Health (SoH).

• Inefficiencies increase safety risks and shorten battery life.

AI-Based Smart Charging Solution

• Machine Learning (ML), especially the RandomForestRegressor model, is proposed to predict SoC and SoH dynamically.

• IoT-connected sensors (voltage, current, temperature, and pressure) collect real-time data.

• Raspberry Pi processes the data and adapts charging parameters in real-time using MQTT protocol and Python.

System Architecture

• Data Collection: INA219 (voltage/current), DS18B20 (temperature), FS-L-0055 (pressure).

• Data Processing: Noise removal and feature extraction.

• Model Training: Uses scikit-learn to train and deploy a model predicting SoC and SoH.

• Control Logic: Adjusts charging based on predictions using dynamic control algorithms.

Prototype and Testing

• Data gathered from sensors at regular intervals.

• Visualizations (graphs and tables) show trends in sensor data.

• SoC is estimated using a voltage-based formula.

• Predictions from the ML model help guide optimal charging strategies.

• The system uses a Flask-based web interface for monitoring and diagnostics.

Security and Future Considerations

• Cybersecurity concerns in smart grid systems (e.g., WINSmartGrid™) highlight the need for secure, resilient infrastructure.

• AI integration offers predictive analytics, real-time adaptability, and long-term cost savings.

Conclusion

The suggested system improves Electric Vehicle (EV) charging by combining real-time sensor data, machine learning-driven battery state evaluation, an adaptive multi-stage charging control strategy, and IoT-enabled remote oversight. By collecting key variables such as voltage, current, temperature, and pressure, a robust data foundation is established. Using scikit-learn, the system effectively forecasts the battery\'s State of Charge (SoC) and State of Health (SoH). Additionally, its adaptive charging control, which utilizes a pulse charging method, continuously modifies the charging profile in real-time to guarantee optimal efficiency, safety, and a longer battery life. By integrating MQTT and a Flask-based dashboard, the system provides an economical solution for real-time remote monitoring and control. This intelligent charging system represents a major leap forward in EV battery management, delivering a scalable, efficient, and secure solution, with opportunities for future enhancements such as advanced predictive analytics and integration with renewable energy sources.

References

[1] D. Pant, N. Singh and P. Gupta, \"Charging System for Electric Vehicles,\" 2022 IEEE Delhi Section Conference (DELCON), New Delhi, India, 2022. [2] M. S. Hashim, J. Y. Yong, V. K. Ramachandaramurthy, K. M. Tan and M. Tariq, \"Coordinated Vehicle-to-Grid Scheduling to Minimize Grid Load Variance,\" 2019 International Conference on Electrical, Electronics and Computer Engineering (UPCON), Aligarh, India, 2019. [3] A. Haraz, K. Abualsaud and A. Massoud, \"State-of-Health and State-of-Charge Estimation in Electric Vehicles Batteries: A Survey on Machine Learning Approaches,\" in IEEE Access, vol. 12, pp. 158110-158139, 2024. [4] D. Reeh, F. Cruz Tapia, Y. -W. Chung, B. Khaki, C. Chu and R. Gadh, \"Vulnerability Analysis and Risk Assessment of EV Charging System under Cyber-Physical Threats,\" 2019 IEEE Transportation Electrification Conference and Expo (ITEC), Detroit, MI, USA, 2019. [5] A. Haraz, K. Abualsaud and A. M. Massoud, \"Ensemble Learning for Precise State-of-Charge Estimation in Electric Vehicles Lithium-Ion Batteries Considering Uncertainty,\" in IEEE Access. [6] Lan-Ron Dung and Jieh-Hwang Yen, \"ILP-based algorithm for Lithium-ion battery charging profile,\" 2010 IEEE International Symposium on Industrial Electronics, Bari, Italy, 2010 [7] Dhairya Parikh, Raspberry Pi and MQTT Essentials: A complete guide to helping you build innovative full-scale prototype projects using Raspberry Pi and MQTT protocol ,Packt Publishing, 2022. [8] Monk, Simon. Raspberry pi cookbook. \" O\'Reilly Media, Inc.\", 2022. [9] A. K. P, P. K. S, M. K. A, S. B. S, P. B. D and D. K. R, \"Recurrent Neural Network based Data-Driven SOC Estimation in Lithium-Ion Battery,\" 2023 International Conference on Distributed Computing and Electrical Circuits and Electronics (ICDCECE), Ballar, India, 2023. [10] M. El-Hajj and D. Vorotilov, \"Testing the Security and Performance of MQTT Protocol on Raspberry Pi for IoT Applications,\" 2024 IEEE Asia Pacific Conference on Wireless and Mobile (APWiMob), Bali, Indonesia, 2024. [11] Hanlei Sun, Jianrui Sun, Kun Zhao, Licheng Wang, and Kai Wang. Data-driven ica-bi-lstm-combined lithium battery soh estimation. Mathematical Problems in Engineering, 2022:1–8, 2022. [12] B. Wukkadada, K. Wankhede, R. Nambiar, and A. Nair, “Comparison with http and mqtt in internet of things (iot),” in 2018 International Conference on Inventive Research in Computing Applications (ICIRCA). IEEE, 2018, pp. 249–253. [13] D. B. C. Lima, R. M. B. da Silva Lima, D. de Farias Medeiros, R. I. S. Pereira, C. P. de Souza, and O. Baiocchi, “A performance evaluation of raspberry pi zero w based gateway running mqtt broker for iot,” in 2019 IEEE 10th Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON), 2019, pp. 0076–0081. [14] M. El-Hajj and D. Vorotilov, \"Testing the Security and Performance of MQTT Protocol on Raspberry Pi for IoT Applications,\" 2024 IEEE Asia Pacific Conference on Wireless and Mobile (APWiMob), Bali, Indonesia, 2024. [15] P. Vogel, T. Klooster, V. Andrikopoulos and M. Lungu, \"A Low-Effort Analytics Platform for Visualizing Evolving Flask-Based Python Web Services,\" 2017 IEEE Working Conference on Software Visualization (VISSOFT), Shanghai, China, 2017,

Copyright

Copyright © 2025 Abisheak S, Muthuram G. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.