Maximum Correntropy Filter-Based Adaptive Fusion Method for SOH Estimation Considering Capacity Regeneration Phenomenon

As a key indicator of aging, accurate state of health (SOH) estimation of lithium-ion batteries helps to keep the safety and reliability of electrical vehicles. Capacity regeneration phenomenon, which unavoidably occurs in the battery aging, will change the degradation rate of batteries and reduce the SOH estimation accuracy. To solve this issue, this article proposes a maximum correntropy filter-based adaptive data-model fusion SOH estimation considering capacity regeneration. First, original capacity degradation data is decomposed into a residual reflecting degradation trend, and uncertain term containing local fluctuation and regeneration. Then, a data-driven SOH model is established by adopting a long short term memory neural network and Lévy process to fit the residual and uncertain term, respectively. Hence, the nonlinearity and long-term time dependence can be captured, and the uncertainty caused by regeneration can be well modeled. Then to reduce the accumulated error, a maximum correntropy filter is designed to achieve adaptive fusion prediction of the data-driven and empirical models. Experiment and simulation results shows the effectiveness of the proposed fusion method considering regeneration.