Evaluation of low-complexity algorithms for assessing lithium-ion battery charging based on state of health metrics

Abstract

Lithium-ion batteries are crucial for portable devices like smartphones and laptops, as well as electric vehicles like e-bikes and cars. However, commercial products often opt for simple charging methods without considering the specific demands of different battery states of health. This study evaluates five simple charging methods under varying battery health conditions, based on six performance indicators: maximum temperature rise, average temperature rise, charge capacity, discharge capacity, charge rate, and charge efficiency. The five methods include constant current-constant voltage charging, constant power-constant voltage charging, and constant loss-constant voltage charging. The study also proposes a states of health estimation method for the charging techniques, using a neural network to build a battery states of health estimator. The results show a maximum relative error of 4.12 %, a minimum relative error of 0.1 %, an average relative error of 0.98 %, and a root mean square error of 1.35 %.