RSPF-based Prognosis Framework for Estimation of Remaining Useful Life in Energy Storage Devices

This paper presents a case study where a RSPF-based prognosis framework is applied to estimate the remaining useful life of an energy storage device (Li-Ion battery). A comparison based on prognosis performance metrics indicates that the RSPF-based prognostic approach is more suitable than classic PF methods to represent rare events such as capacity regeneration phenomena between charging periods, in terms of accuracy of the state estimate and steadiness of the RUL estimate. We surmise that the existence of particles in the tails of the state pdf allow the RSPF-based prognostic algorithm to generally provide a more conservative estimate of the RUL of the faulty piece of equipment. We surmise that it also helps to incorporate the probability of rare and costly events in the evolution of the fault condition in time.