Accurately estimating the state of health (SOH) of the lithium-ion batteries (LIB), for battery management systems, plays an important role in an ensuring reliable system operation and reducing maintenance costs. Because of the complicated degradation mechanism and the complex internal electrochemical reactions, accurate SOH estimation remains challenging for battery energy management and applications. In this study, we proposed a new SOH estimation method. First, the data is preprocessed and multiple features are extracted to simulate the aging process of the LIB, and the battery capacity is selected as the state variable. Because of the strong capability to fit complex nonlinear problems, a regression model based on gradient boosting decision tree (GBDT) is proposed to estimate the SOH. In addition, a new hybrid optimization algorithm based on quantum particle swarm optimization (QPSO) algorithm and Nelder-Mead simplex (NMS) algorithm is proposed for parameter optimization of the GBDT model. Several lithium-ion battery test data sets from the NASA Ames Prognostics Center of Excellence were selected to validate the proposed method. Compared with other SOH estimation methods and other parameter optimization algorithms, the experimental results show that the proposed method is superior in terms of accuracy, generalization performance and reliability.
Keywords Lithium-ion batteries, state of health, feature selection, parameters optimization, gradient boosting decision tree