Battery total capacity estimation based on the sunflower algorithm

Abstract

Battery total capacity estimation is crucial for battery management systems (BMSs). Accurate knowledge allows BMSs to anticipate potential failures in advance, through indicators such as the state of health. In this paper, we suggest a new capacity estimation framework based on the sunflower optimization algorithm (SFO). The technique reported accounts for capacity error sources, including measurement and estimation noises, in addition of being recursive. SFO is applied to find the best candidate that has the minimum objective function value in a population of sunflowers. A reduction strategy of the search space is used to enhance the solution accuracy. The best candidate is then used through a forgetting factor to update battery capacity. The algorithm's accuracy was verified using NASA Prognostic Data Repository, in addition to three scenarios of a battery in electric and plug-in hybrid electric vehicle applications. As far as we can tell, this is one of the first attempts to use SFO to estimate battery capacity. SFO showed great performance: SFO recorded a maximum error of 1.26% in the worst-case scenario. Moreover, all predictive performance indicators (root mean square error RMSE, mean absolute error MAE, and mean absolute percentage error MAPE) did not exceed 0.4% in all tests.