Recurrent neural network-enabled adaptive sliding mode control for fast charging of Li-ion batteries in electric vehicles

Abstract

The widespread adoption of electric vehicles requires advanced charging control strategies capable of achieving accurate current regulation and improved transient performance. This paper presents a recurrent neural network–Integrated adaptive sliding mode controller (RNN-ASMC) for the control of a DC–DC buck converter in a five-stage constant-current charging system for lithium-ion batteries. The proposed RNN-ASMC controller is evaluated against conventional proportional-integral (PI), fuzzy-PI-derivative (PID), NN-optimized PID, and adaptive neuro-fuzzy inference system (ANFIS-PID) controllers using the same converter configuration, charging profile, battery parameters, and simulation conditions. The controller combines adaptive sliding mode control with RNN-based estimation to improve charging-current tracking and reduce transient oscillations. Mathematical analysis and MATLAB/Simulink simulation results demonstrate improved transient current regulation, reduced settling time, and lower steady-state error under the investigated operating conditions. The analysis based on Lyapunov method ensures the convergence characteristics of the designed control algorithm in the presence of bounded uncertainty in the model. The analysis performed does not involve any evaluation of system performance at the system level; the only evaluation that is done is that of the controller performance, which is based on a first-order ECM. As a result, the improvements achieved are confined to the regulation of charging current and transient dynamic performance. Other aspects like electro-thermal effects, battery degradation, state-of-health estimation, etc, have not been considered.