A training-free algorithm based on the Dempster-Shafer theory estimated battery state of health with a mean RMS error of <5% and state of charge with <2.5%.
A computationally efficient, training-free algorithm based on the theory of evidence accurately estimates battery state of charge and state of health.
Accurate estimation of the state of charge (SOC) and state of health (SOH) of electric batteries is of high importance. Numerous studies are available on this aspect, employing model-based or data-driven approaches. The model-based methods are battery specific, whereas data-driven methods require a large amount of data for training. Both the above classes of methods are computationally complex. In this paper, a computationally efficient yet universally applicable algorithm is proposed to estimate both the SOC and SOH of the battery with high accuracy based on the Dempster-Shafer theory, also known as the theory of evidence. This would be a huge leap in battery prognostics, as the proposed method can be used online, while the battery is in use, in a non-intrusive manner without any training. The proposed method has been tested on several datasets from six different battery models, including both publicly available and newly acquired data. Results show that the proposed method has a mean RMS error of less than 5% for all the models of batteries for the SOH and less than 2.5% for the SOC which is comparable to the machine learning methods.
Jetti et al. (Thu,) conducted a other in Electric batteries. Dempster-Shafer theory (theory of evidence) based algorithm vs. Machine learning methods was evaluated on Mean RMS error for state of health (SOH) and state of charge (SOC). A training-free algorithm based on the Dempster-Shafer theory estimated battery state of health with a mean RMS error of <5% and state of charge with <2.5%.