Research on state of charge estimation of lithium battery based on adaptive forgetting factor recursive least squares and multi-innovation adaptive unscented Kalman filter

State of charge (SOC) estimation is critical to battery management and mileage prediction of electric vehicles. Model-based SOC estimation supply an effective solution to computation efficiency, but its accuracy highly dependent on model parameters. Three key technical contributions are made. First, the traditional model parameter identification method based on recursive least squares with forgetting factor (FFRLS) often sets its forgetting factor to be a constant during the iterative process, while disregarding estimated voltage accuracy. In this study, an adaptive forgetting factor recursive least squares method (AFFRLS) is proposed based on the adaptive theory. The core idea is that the forgetting factor is adaptively adjusted according to preset adaptive formula of the voltage error. In this way, both robustness and identification accuracy of the algorithm can be improved effectively. Second, a multi-innovation adaptive unscented Kalman filter (MIAUKF) is proposed to utilize historical voltage data and weaken the impact of voltage cumulative error on SOC accuracy. Finally, the performance of the proposed approach is evaluated by comparison simulations considering three different temperature scenarios based on two standard driving cycles. The results show that the proposed approach can obtain higher accuracy in SOC estimation with maximum relative error below 2%.