Alternative combined co-estimation of state of charge and capacity for lithium-ion batteries in wide temperature scope

Abstract

This study develops a combined method for co-estimation framework for state of charge and capacity of lithium-ion batteries considering wide temperature scope. In this framework, a second-order equivalent circuit model incorporating temperature compensation is established to characterize the battery's electrical performance. Next, the particle swarm optimization algorithm integrating data accumulation and dynamic updating technique is exploited to adaptively identify the model parameters. Then, the charging duration is selected as the health feature, and the long short-term memory recurrent neural network is leveraged to predict battery capacity with high accuracy. Moreover, the dynamically updated model parameters and capacity value are inputted to estimate the state of charge based on the square root cubature Kalman filter. Experimental validations are conducted considering different voltage ranges, data lengths, temperatures and aging status to illustrate the feasibility and superiority of the designed estimation framework, and the results highlight that the proposed co-estimation framework can supply precise state of charge estimation with the error of less than 2% under time-varying temperatures, and the capacity can be predicted efficiently with less than 0.071 A h.