Kan Based SoC Estimation

Abstract

Lithium-ion batteries (LiBs) are most favored storage systems used extensively in electric cars (EVs), energy storage systems for renewables, and mobile phones, providing high energy density, high capacity, and efficiency. Accurate State of Charge (SoC) estimation is essential for the effective operation of Battery Management Systems(BMS). To prevent overcharging, over-discharging, and thermal runaway while maximizing performance and lifespan. In this study, the 2RC Equivalent Circuit Model(ECM) is simulated in MATLAB to generate battery data. The 2RC model includes a source for open-circuit voltage (OCV), a series resistance, and two resistor-capacitor networks, which demonstrate how lithium batteries behave during quick changes and stable situations. Calculate the SoC using EKF, AEKF, and AH methods. Comparing the accuracy of these methods shows that AH provides the best results. This simulated data is used to train and evaluate a KAN-based SoC estimation framework. Voltage, current, and temperature responses from the circuit model serve as the input, and the SoC as the target feature for a KAN model. The proposed KAN-based model estimates the SoC accurately, compared with the neural network model.