Group Learning Algorithm-Guided Hybrid Learning Framework for Battery Degradation and Lifetime Prediction in Electric Vehicles

Abstract

Accurate forecast of battery deterioration and lifetime is essential for the reliability, safety and efficiency of EV. Lithium-ion batteries can age, but if you understand how they age, you can manage them better and make them last longer. Battery degradation is affected by factors such as the number of cycles, depth of discharge, charge/discharge rates, temperature and electrochemical processes. This work presents a GRLA-ConvBiLST Hybrid Learning Framework for accurate prediction of battery deterioration and lifetime. Firstly, CEEMDAN decomposes battery capacity data into high and low frequency components. Then, the battery degradation patterns are characterised by health-related measures such as DC (dI/dV), DV (dV/dQ) and ICA (dQ/dV). The ConvBiLST network captures the spatial and temporal dependencies and then the network hyper-parameters are optimised by the GLA for better learning efficiency and prediction accuracy. The experimental findings reveal that the proposed framework achieves the best State-of-Health (SOH) prediction accuracy of 94.12%, outperforming CNN, LSTM, CNN-LSTM and GRU models. The results show that the suggested approach can efficiently and reliably solve the problems of EV battery deterioration and lifespan prediction, which supports the intelligent battery management systems, extends the service life of batteries and enhances the performance and sustainability of EV.