A Comparative Analysis of Different Types of SoC Estimation Using Machine Learning Techniques for Li-ion Battery Management System (BMS)

The increasing demand for using electric vehicles (EVs) has emphasized the need for efficient and reliable lithiumion batteries, which form the basis of modern EV technology because of their high energy density, longevity and reusability. However, to achieve these performances and the security of these batteries, SoC measurement is fundamentally important as a key parameter defining battery recharge level compared to its nominal capacity. Accurate estimation of SoC is again crucial as overcharging leads to improved SoC estimation, profound releasing and waste energy concerns of encouraging, leading to reduced battery life, less safety, and poor EV performance tantamount to the degradation level. In response to these difficulties, this work focuses on the plan and development of the progressed SoC estimation techniques that combines three different approaches; Coulomb counting (CC), Extended Besides, we have such systems as Extended Kalman Filter (EKF), Artificial Neural Network (ANN). This technique Coulomb Counting, though simple and easy to implement, accrues aggregate errors when used over a long time. Integrating the proposed ANN and EKF model results in a high-performance SoC estimation model capable of overcoming non-linear and weakness associated with lithiumion battery. MATLAB simulations are made to measure the effectiveness of all the plans, with come out illustrating the predominant accuracy and reliability of the hybrid model irrespective of conditions. This work is relevant regarding the enhancement of battery technologies and supports further efforts to push the growth of electric vehicles. Furthermore, accurate estimation of other key battery states such as State of Health (SoH), State of Energy (SoE), State of Power (SoP), and Remaining Useful Life (RUL) is crucial for efficient battery management and longevity.