Comparative Study Between Unscented Kalman Filter and Multi-Layer Perceptron Applied in an Electric Vehicle Simulation with Pack Parameters Generated from the Database

In electric vehicles (EVs), monitoring battery cells ensures safety and longevity. The Battery Management System (BMS) is responsible for this task, using variables such as State of Charge (SoC). This paper presents a comparative study between the Unscented Kalman Filter (UFK) and Multilayer Perceptron Neural Network (PMC) methods applied to a pack using a simulation of a vehicle under the FTP-75 cycle with reference SoC using Coulomb Counting (CC). The Neural Network performed better with an RMSE value of 0.418 compared to 0.0593 for UFK.