Dual Active Bridge Converter Output Current Ripple Prediction in EV Battery Chargers Through Multimodal Adaptive Spatio-Temporal Graph Neural Network

The prediction of converter output current ripple in Electric Vehicle (EV) battery chargers stands as a vital condition to maintain stable charging with safe battery operation by reducing voltage fluctuations and delivering steady currents. The prediction of converter output current ripple faces difficulties because the complex dynamics of converter behavior changes according to various load conditions and environmental factors. This measurement inaccuracy threatens the capability to make optimal power transfers and reduces loss management ability thereby affecting the life span and operational performance of the EV battery. This research presents a framework to predict the amount of current oscillation in chargers for EV battery systems powered by Dual Active Bridge (DAB) Converter. The proposed technique utilizes Multimodal Adaptive Spatio-Temporal Graph Neural Network (MASTGNN) as its method. The proposed method aims at both enhancing current ripple prediction accuracy and optimizing converter performance together with improving EV battery lifespan during charging. A MAST-GNN process predicts output current ripples in DAB converters used for EV battery charging through modeling which evaluates the behavior changes of voltage and current and power when operational parameters vary. Implementation of the proposed method in MATLAB will undergo comparison testing with CNN as well as BESO-ANFIS and ANN together with other existing methods. By achieving an accuracy of 97% in current ripple prediction and maintaining a high converter efficiency of 98.5%, the proposed MASTGNN method outperforms traditional approaches without compromising the stability and performance of the EV battery charging process.