| Abstract [eng] |
Lithium-ion battery packs are widely used in electric vehicles, renewable energy storage systems, and portable electronics due to their high energy density and long cycle life. However, battery packs consist of multiple cells connected in series, where variations in internal resistance, capacity, and thermal characteristics cause cell imbalance during operation. These imbalances reduce the usable capacity of the battery pack, accelerate degradation, and may lead to thermal safety risks. Therefore, effective battery cell balancing and thermal monitoring are essential functions of modern Battery Management Systems (BMS). The aim of this thesis is to investigate battery charging balancing methods and to develop a predictive thermal modeling approach using embedded machine learning techniques. A simulation model of a multi-cell lithium-ion battery pack was developed in MATLAB/Simulink, incorporating charging and discharging cycles, state-of-charge dynamics, and thermal behavior based on resistive heating mechanisms. Using the generated simulation dataset, a lightweight neural network regression model was designed to predict short-term temperature rise of battery cells based on parameters such as cell temperature, current, state of charge, and previous temperature change. The developed model demonstrates accurate prediction of short-term temperature variations and is capable of identifying cells that may experience higher thermal stress during operation. The results show that predictive thermal modeling can provide early indicators of potential overheating conditions and support proactive balancing decisions. Furthermore, the computational and memory requirements of the proposed model are sufficiently low for implementation in embedded microcontroller-based BMS platforms. The research demonstrates that integrating machine learning-based thermal prediction with conventional battery balancing strategies can enhance battery safety, improve operational efficiency, and extend battery lifetime. The proposed approach contributes to the development of intelligent battery management systems suitable for future electric mobility and energy storage applications. |
