Joint SOH and RUL estimation for lithium-ion batteries via optimal deep belief network with Bayesian algorithm

This article proposes an innovative method for assessing the state of health (SOH) and remaining useful life (RUL) of lithium batteries . The innovation lies in the integration of an optimal deep belief network with a Bayesian algorithm (ODBN-BA) for joint estimation, coupled with intrinsic computing expressive empirical mode decomposition with adaptive noise (ICEEMDAN) for in-depth feature extraction. Through precise parameter tuning using BA, the model achieves significant enhancements in prediction accuracy, robustness, and generalization capabilities. The experiment was validated using publicly available data from the national aeronautics and space administration (NASA) center for excellence in forecasting and the center for advanced life cycle engineering (CALCE), and compared with various advanced algorithms. This article uses SimuNPS for simulation verification, the results showed that ODBN-BA method proposed in this paper performed well in both SOH and RUL estimation, with high accuracy, strong robustness, and good generalization. Especially when dealing with noisy data, this method can still maintain excellent estimation performance, providing an effective solution for online monitoring of lithium battery SOH and accurate prediction of RUL. In the experimental data, SOH estimated MAE value of B0005 battery was as low as 9.7261E-05, and RUL estimated AE value was 0, further proving the excellent ability of ODBN-BA model in reducing estimation errors and improving estimation accuracy, demonstrating its huge potential for application in battery health management. • Create a new estimation model ODBN-BA; • Perform noise processing and compare with the original data • Perform joint estimation of SOH and RUL; • Use six types of data to ensure the applicability of the method