DATA‐DRIVEN PROGNOSTICS FOR BATTERIES SUBJECT TO HARD FAILURE

This chapter considers data‐driven remaining useful life (RUL) of a battery prediction, which is typically made on the basis of projecting the trajectory of the system's health indicator, often called the degradation signal. Two most commonly used health indicators of batteries are capacity and internal resistance, while other health‐dependent variables such as battery self‐discharge rate may also be considered. By analyzing the evolution paths of the health indicating variables/degradation signals, it is possible to infer not only the current but also the future health status of the unit being studied. The chapter introduces a method specifically developed for battery RUL prediction under hard failure. In this method, a joint modeling scheme is used to take into consideration both the degradation data and the time‐to‐failure data. To better assess the performance of the prognostic algorithm, alternative interval prediction, the maximum power interval (MPI), is introduced as opposed to confidence intervals and mean/median‐based intervals.