Enhancing State-of-Health Estimation for Electric Vehicles Using Transfer Learning Techniques

Accurate State of Health (SOH) estimation is crucial for optimizing lithium-ion battery (LIB) performance and ensuring safety in electric vehicles and renewable energy systems. Traditional model-based methods often fail to capture LIBs' complex ageing dynamics, so this research focuses on data-driven approaches, using machine learning and Transfer Learning (TL) techniques. Neural Networks (NN) such as Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRUs) excel in handling the temporal dynamics needed for accurate SOH estimation. Utilizing three different datasets the study employs pre-Trained models and techniques to assess different transfer learning techniques and find a trade-off between compu-Tational complexity and accuracy in SOH estimation. Seventeen TL techniques were explored across NN architectures. LSTM networks generally showed the lowest mean absolute error (MAE), while GRUs were the most computationally efficient. Results indicate significant improvements over baseline models, emphasizing the potential of transfer learning in battery health estimation.