A multi-timescale framework for state monitoring and lifetime prognosis of lithium-ion batteries

Battery State-of-charge (SOC) estimation and lifetime prognosis play important roles in its fuel-gauging and predictive maintenance. However, the inevitable modeling biases and measurement noises caused by the complexity of battery dynamics in nonlinearity, temperature sensitivity, and degradation, still challenge the accuracy and robustness. To bridge this research gap, this paper presents a multi-timescale framework for SOC estimation and lifetime prediction of lithium-ion batteries. In the fast timescale, a recursive-least-square-based parameter identification algorithm is first employed to eliminate the effects of identification errors on SOC estimation and reduce the dimension of the SOC observer. Then, a robust observer and dual extended-Kalman-filter (DEKF) are combined to achieve simultaneous SOC and capacity estimation against bounded modeling errors. In the slow timescale, the battery lifetime is predicted through a particle-filtering based on the capacity. The effectiveness and superiority of the proposed method are demonstrated by both simulation and experimental results conducted on LiCoO2 cells. The results indicate that the root-mean-square-error of the SOC and capacity can be achieved within 3.5% and 0.5%, respectively, in the presence of the model mismatch. In addition, the lifetime prediction framework can also ensure accurate lifetime prediction.