Implementation of Discrete Wavelet Combined Bayesian-Optimized Gaussian Process Regression-Based SOC Prediction of EV Battery

The precise state of charge assessment of lithium-ion battery packs used in electric vehicles is crucial to their dependable and energy-efficient operation. The performance of estimation algorithms reduced due non-Gaussian measurement noise generated due to electromagnetic interference and sensor errors. In order to address this, DWT is proposed in this study together with Bayesian optimized Gaussian process regression. By lowering the noise and expanding the dimensions of the battery's data being measured, the DWT is employed to extract the valuable data characteristics. At first the mRMR method will be utilized to choose significantly correlated input parameters. The retrieved features are utilized to train the BO-GPR model and accurately predict the SOC. The model is simulated in Matlab 2023a Version. The obtained result indicates that the compared to existing literatures, proposed (DWT-B0-GPR) represents the better relationship between measured data and SOC with reduced RMSE of 0.0054.