The innovation provides enhanced health management routines for batteries. A mathematical model has been developed to describe battery behavior during individual discharge cycles, as well as over the cycle life. Different prognostic modes for estimating the state of charge, state of life, end of discharge, and/or end of life of a battery are provided. It employs a mathematical, rigorous reasoning framework for better understanding and representation, manipulation, and management of the various sources of uncertainty inherent in the prognostics of the remaining useful life in a battery. The models used to estimate the remaining useful life of batteries are linked to the internal electrochemical processes of the battery. The effects of load (and, by extension, temperature) have been incorporated into the models. The model is used in conjunction with a particle filtering framework to make state estimations and probabilistic predictions of remaining useful life for individual discharge cycles, as well as for battery life. The model fidelity improves when the influence of factors like temperature, discharge C-rate, end of discharge, state of charge after charging, etc., are explicitly incorporated. Model validation studies were conducted using data from a series of battery cycling experiments at various thermal and electrical loading conditions. In addition, the models and algorithms were integrated on an electric UAV and subsequently flown on numerous test flights.
The sort of information obtained from the battery prognostics can be used for making decisions about usage within the current discharge/charge cycle to ascertain that primary usage goals can be accomplished. The information can also be used to conduct a tradeoff in long-term durability versus short-term usage needs.
While this model was developed with Li-ion battery chemistries in mind, it can be applied to other batteries as long as effects specific to those chemistries are modeled as well (e.g. the memory effect in Ni-Cd rechargeable batteries).
This work was done by Bhaskar Saha and Kai F. Goebel of Ames Research Center.