Lithium-ion battery remaining useful life prediction for portable medical device safety: a VMD-RVM based approach

• Propose a new method to predict the RUL of lithium-ion batteries in medical equipment • To reduce the battery safety risks through high-precision life prediction. • This method can improve the RUL prediction accuracy. • It can be applied to monitor the BMS of portable medical equipment. Medical monitors typically come equipped with lithium-ion batteries as a source of energy supply. With the popularity of monitors, battery management has become an increasingly important indicator for evaluating the effectiveness of monitor usage. The Remaining Useful Life (RUL) of a lithium battery is an important parameter that characterizes the functionality of a battery management system. Accurate prediction of the RUL of a lithium battery can ensure the smooth operation of the battery equipment and guarantee the utilization efficiency of the monitor. In response to the problem of low estimation accuracy caused by complex estimation models in current lithium battery RUL prediction, a lithium battery RUL prediction method using Variational Mode Decomposition (VMD) and Relevance Vector Machine (RVM) is proposed by selecting the time when the discharge phase voltage drops to the cut-off voltage as the health indicator. Firstly, the time when the voltage drops to the cut-off voltage during the discharge stage is used as a health indicator, and Spearman correlation analysis is used to determine that the voltage drop time has a strong correlation with the battery's capacity data. Secondly, due to the phenomenon of capacity regeneration in battery capacity, VMD is used to decompose the voltage drop time signal, and RVM is used for training. Finally, the voltage drop time data is input into the VMD-RVM model to obtain the predicted RUL of the battery. Verified using NASA's publicly available dataset, the experimental results show that the VMD-RVM prediction method used in this paper has lower MAE and RMSE compared to EEMD-RVM and EMD-RVM, indicating that the proposed method has higher accuracy.