Identificação de parâmetros para um modelo de circuito elétrico para bateria de segundo uso

Abstract

The increasing adoption of electric vehicles and, consequently, of lithium-ion batteries raises concerns about battery disposal, as they are removed from electric vehicles once they reach 80% of their nominal capacity. In this context, second-life batteries emerge as sustainable, circular economy applications, because the removed battery modules can be classified and rearranged as storage systems, primarily for stationary applications. Nevertheless, the lack of information on second-life batteries, combined with the need to understand their behavior, particularly regarding aging cycles, represents a significant concern. Within this context, equivalent circuit models can be employed to study battery behavior. However, there is a gap concerning the evolution of model parameters over the battery’s operating cycles. Aiming to understand these aspects, the present work has the overall objective of comparing three techniques used to identify the parameters of a first-order model for a lithium-ion battery and evaluate the changes in these parameters and the battery capacity as a function of the degradation resulting from the number of second-life operating cycles. Laboratory tests were carried out on an NMC lithium battery sourced from an electric vehicle to determine its capacity, discharge tests to characterize the voltage curve, and aging tests through charge and discharge cycles. The model parameters were determined at every 30- cycle interval. Three parameter-identification techniques were implemented and compared: the leastsquares method, curve analysis, and curve fitting. Among them, all three techniques presented a relatively low mean absolute percentage error, with values below 2%, the least-squares method presented the greatest accuracy. Battery degradation corresponded to 1,26% of the total capacity when compared with the first reference test, with a total loss of 1,42Ah over the cycles. The curve-analysis method showed greater sensitivity to the behavior associated with changes in the circuit model parameters over the evaluated operating cycles. For the leastsquares method, as well as for curve fitting, because they involve optimization problems, the parameters found do not always adequately represent the physical conditions of the battery, that are reflected in the analysis of parameter changes.