An inhomogeneous hidden Markov model for state recognition of lithium battery

Timely state recognition of battery system is significant to prevent safety accidents of electric vehicles. The aging process of lithium battery is a random process, the aging speed is changing with time, and its aging degree is difficult to measure. Therefore, a state recognition model of lithium battery is established by an inhomogeneous Hidden Markov Model (IHMM). In order to further improve the state recognition effect, an IHMM with multiple observation variables is established. By analyzing degradation data of No.5 lithium battery from NASA data set and numerical experiments, the results show that IHMM is superior to Hidden Markov Model (HMM) in the state and state sequence recognition effect. At the same time, the IHMM with multiple observation variables has higher recognition effect.