Review on Machine Learning Methods for Remaining Useful Lifetime Prediction of Lithium-ion Batteries

Electric cars are considered as the most ecologically friendly and low-cost means of transportation in the future. As a result, battery technology advancement is of interest for many researchers. Lithium-ion batteries are mostly used for electric vehicles. However, if the Remaining Useful Lifetime (RUL) drops below capacity degradation, devastating device failure will occur. Hence, it is important to predict the RUL to prevent such problems. Data-driven methods are demonstrated to be superior to model-based methods for this reason. This paper provides a review on Machine Learning (ML), one of the data-driven methods, and summarizes various approaches that have been used in lithium-ion (Li-ion) batteries RUL prediction. In addition, comparison of model-based and ML methods are discussed. In particular, the comparison of three ML methods,i.e., Support Vector Machine (SVM), Neural Networks (NN), and Deep Learning(DL) are also presented. Simulation results show that SVM is able to provide higher RUL accuracy than LSTM and ANN.