Remaining Useful Life Prediction Driven by Multi-source Data for Batteries in Electric Vehicles

Abstract

Predicting battery remaining useful life (RUL) is used for early warning of battery aging failure and providing instructions of battery maintenance and recycling. The existing RUL prediction focus too much on decreasing the dependence of aging tests, neglecting the value of test data. In this regard, a battery RUL prediction method driven by multi-source data is proposed for EVs to make full use of the aging test data from other cells. Six lithium-ion batteries were used to verify the effectiveness of the method. The results show that the prediction error is less than only 1 cycle in the case of capacity ‘diving’. In conclusion, the proposed method effectively improves the performance of RUL prediction by using multi-source data, and provides a solution for battery management in the era of big data.