Indirect remaining useful life prognostics for lithium-ion batteries

The capacity is often used as health index(HI) for predicting the remaining useful life (RUL) of a lithium-ion battery. However, it is quite challenging to monitor the capacity of battery on-line. There exist errors of precision and accumulation, which can not be measured in the process of using accurately. In order to solve the problems, discharge voltage and relative parameters are selected as indirect health index, such as time interval of equal discharge voltage. Correlation coefficient method is applied to prove strong relevant between the real capacity and time interval of equal discharge voltage. In this paper, RVM_PF_AR fusion algorithm, which RVM is used to train the degradation tendency of batteries, PF is used to multi-step prognostic and AR is used to forecast the long-term trend, is proposed to estimate the remaining useful life indirectly. Furthermore, B5 from NASA and J9 from a company are used to experiment, showing that the fusion algorithm can predict RUL effectively. Finally, the validity and universality of this algorithm are also proved.