State of health estimation of lithium-ion batteries based on equivalent circuit model and data-driven method

The estimation of the state of health (SOH) of lithium-ion batteries (LIBs) is of great significance to ensure the safety and reliability of the battery management system. Equivalent circuit model (ECM) and data-driven based methods are commonly used to estimate the SOH. Each method has pros and cons, but combining them is challenging. In this paper, a new approach integrating ECM and data-driven methods is proposed for SOH estimation. Firstly, the internal resistance of a first-order ECM of the LIB is identified using particle swarm optimization (PSO). Secondly, a fractional-order three-learning strategy PSO is adopted to optimize a back-propagation neural network (BPNN). Finally, the internal resistance of the ECM, voltage, current and time of the LIB are used as input to the optimized BPNN to predict the SOH. Different battery datasets from NASA and CALCE are used to verify the effectiveness of the proposed technique. The results show that the maximum root mean square error (RMSE) of the new method does not exceed 1.35%, and the error of the best SOH prediction is just 0.39%. Moreover, the highest and lowest prediction interval coverage probability (PICP) are 100% and 85.71%, respectively. Compared with other approaches, the proposed method reveals faster convergence speed, superior accuracy, and better generalization ability.

Automated summary

This paper proposes a new approach that integrates the equivalent circuit model (ECM) and data-driven methods for estimating the state of health (SOH) of lithium-ion batteries (LIBs). By identifying the internal resistance of the ECM using an optimization algorithm and using an optimized neural network to predict the SOH, the proposed method demonstrates fast convergence, high accuracy, and good generalization capability.