Research On Lithium-ion Battery Rul Model Driven Prognosis Method

As lithium-ion battery have been widely used in many systems, it’s more and more important to build a health monitoring for battery. As a particularly crucial part of the lithium-ion battery health monitoring, more and more methods are used in the field of the remaining useful life prediction in order to achieve promotion RUL prediction accuracy. As a kind of complex electrochemical system, the performance of lithium-ion battery will degrade under continuous charging and discharging. When its degradation decreases to the failure threshold of lithium-ion batteries, it’s generally considered the lithium-ion batteries can’t be used anymore.How to build a model to accurately describe the capacity degradation has always been a crucial point to provide more precise in RUL prediction for model driven methods. The polynomial model and exponential model with the cycle number and capacity, and the double exponential model with the discharge time and capacity are commonly used to describe the capacity degradation as a smooth curve which is unable to accurately describe the phenomenon of the charging capacity degradation process. In recent years, more and more models are built based on seeking for the relationship between lithium-ion batteries and other health indicator, but it still have no a suitable model to describe the capacity degradation.Therefore, in order to solve the above problems, this paper is committed to build a model that can accurately describe the lithium-ion battery capacity degradation for a more precise and better robustness in RUL prediction.First, in order to analysis the relationship between capacity and other health indicators, this paper extracted the impedance and the temperature changing rate(TR) and temperature difference(DT). Considering the close relationship between battery’s health degradation and temperature, especially the highly linear correlation between capacity and TR and DT, two new models are proposed with TR and DT to describe the capacity degradation.Second, a new RUL prediction is proposed based on these two models with the result of these two health indicators TR and DT based on the SVM method.Finally, in order to provide a better performance in RUL prediction, these two models are presented on the ensemble learning algorithm, the method of the weighted linear regression and a kind of clustering algorithm are applied in this paper.