Improved LSTM based state of health estimation using random segments of the charging curves for lithium-ion batteries

Abstract

Battery State of Health (SOH) estimation is crucial for providing valuable information for optimizing battery usage and improving battery efficiency. Considering the uncertainties in battery charging behavior during practical usage, this paper proposes an ensemble model based on an improved long short-term memory (LSTM) neural network. The model takes random segments of charging curves as input to estimate the SOH of lithium-ion batteries. In this paper, a multi-layer LSTM network with attention mechanism is proposed as the base learner. Then, multiple base learners are trained using different parts of the charging curve segments. The BP neural network is used to integrate the SOH estimation value of each base learner to obtain the final SOH estimation value. The accuracy and robustness of the proposed method are validated using the Oxford lithium-ion battery dataset and NASA battery degradation dataset. Additionally, this paper investigates the influence of charging segment length and the number of sampling points on the estimation results.