The unscented Kalman filter (UKF) is a commonly used algorithm for estimating the state of charge (SOC) of lithium-ion batteries. However, in practical applications, due to uncertainties such as external environmental temperature variations, battery capacity degradation, and non-Gaussian process noise, further improvements in algorithm performance are required to ensure more accurate estimation. Thus, an improved unscented Kalman filter algorithm (PO-RUKF) is proposed. Firstly, H∞ filtering is introduced into the UKF to enhance robustness, mitigating the effects of various disturbances. Secondly, the parrot optimization algorithm is employed to adaptively adjust the process noise covariance matrix of the UKF, addressing the issue of prior determination of filter noise parameters and thereby improving filtering accuracy. Finally, experimental validation is conducted using two publicly available datasets from the university of Maryland under FUDS and HPPC conditions. The results demonstrate that under varying temperatures, battery capacity degradation states, and different operating conditions, the improved algorithm achieves higher SOC estimation accuracy compared to traditional UKF and robust UKF algorithm, with an average absolute error of less than 0.50% and a root mean square error of less than 0.56%. Additionally, the improved algorithm exhibits stronger robustness and universality. It is proved that the proposed method can provide more reliable and effective technical support for SOC estimation of lithium ion batteries.