State of Health Diagnosis and Remaining Useful Life Prediction for Lithium-ion Battery Based on Data Model Fusion Method

Accurate state-of-health (SOH) diagnosis and remaining useful life (RUL) prediction of lithium-ion batteries (LIBs) play an extremely important role in ensuring safe and reliable operation of electric and hybrid vehicles. However, due to the complex electrochemical properties, it is difficult to achieve the goal of accurate diagnosis and prediction. Here, we propose a novel data-model fusion method to perform accurate SOH estimation and RUL prediction for LIBs, which considers nonlinear dynamics of not only discharging process but also charging process. A long short-term memory (LSTM) network is first employed to model battery SOH dynamics. A neural network (NN) model is then developed to describe battery capacity degradation mechanism according to the prior knowledge extracted from the charging process. Finally, an unscented Kalman filter (UKF) algorithm is incorporated with the LSTM network and NN model to filter out the noises and further reduce the estimation errors. Different from the traditional model fusion approaches, this proposed method uses full information from all sensors, and with no need for any physical model. Experiments and verification demonstrate both the effectiveness of this proposed method and its superior modeling performance as compared with several commonly used methods.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.