UConn team devises chemistry-adaptive battery fuel gauge based on probabilistic data association
29 September 2014
Researchers at the University of Connecticut have developed a chemistry-adaptive battery fuel gauge based on a probabilistic data association (PDA) algorithm. Their paper, available online in the Journal of Power Sources, addresses the problem of tracking the state-of-charge (SOC) in Li-ion batteries when the battery chemistry is unknown.
It is desirable for a battery fuel gauge (BFG) to be able to perform without any offline characterization or calibration on sample batteries. All the existing approaches for battery fuel gauging require at least one set of parameters, a set of open circuit voltage (OCV) parameters, that need to be estimated offline. Further, a BFG with parameters from offline characterization will be accurate only for a “known” battery chemistry. A more desirable BFG is one that is accurate for “any” battery chemistry.
—Avvari et al.
In their paper, the researchers showed that by storing finite sets of open circuit voltage parameters of possible batteries, they could derive a generalized battery fuel gauge using a probabilistic data association (PDA) algorithm.
The PDA algorithm starts by assigning prior model probabilities (typically equal) for all the possible models in the library and recursively updates those probabilities based on the voltage and current measurements. In the event of the need to gauge an unknown battery, the PDA algorithm selects the most similar OCV model to the battery from the library.
The team also demonstrated a strategy to select the minimum sets of OCV parameters representing a large number of Li-ion batteries.
Resources
G.V. Avvari, B. Balasingam, K.R. Pattipati, Y. Bar-Shalom (2015) “A battery chemistry-adaptive fuel gauge using probabilistic data association”, Journal of Power Sources Volume 273, Pages 185-195, doi: 10.1016/j.jpowsour.2014.09.006
Comments