Considering the Unscented Kalman Filter (UKF) typically assumes zero system noise based on existing knowledge or experience, it may be difficult to model actual noise in general situations. Moreover, using a single innovation sequence can lead to bias in state estimation, affecting the precision in estimating. To address this, utilizing Particle Swarm Optimization (PSO) for parameter recognition in second-order RC model. Incorporating an adaptive filtering algorithm to modify the noise covariance matrix dynamically during real-time updates, as well as utilizing multi-innovation identification theory to variably update the state to propose the MI-DAUKF algorithm to achieve coordinated estimation between State of Charge (SOC) and State of Health (SOH) of lithium-ion batteries. Simulation experiments demonstrate that the MI-DAUKF algorithm achieves a maximum error of 1.16%, an average error (AE) of 0.56%, and a root mean square error(RMSE) of 0.67% in SOC estimation. The estimated capacity error is less than 0.005 Ah, with an error percentage maintained within 0.15%, thus confirming the accuracy and robustness of the algorithm.