Comparative Analysis of SoC Estimation in Batteries using SFRA and EIS

—The paper proposes a performance comparison between two different techniques for estimating the State of Charge (SoC) of batteries. In particular, Electrochemical Impedance Spectroscopy (EIS) and Sweep Frequency Response Analysis (SFRA) are analyzed. Although the excellent performance achievable with EIS on SoC estimation is known, this technique suffers from an intrinsic disadvantage linked to application time. In fact, by using even very low stimulus frequencies, in the order of mHz, the times required to execute the techniques are incompatible with real-time applications; unless particular feature selection strategies are implemented, as the authors have already demonstrated. For this reason, in this paper, the use of SFRA for battery monitoring is proposed for the first time. Through SFRA, application times are significantly reduced thanks to the high stimulus frequencies used. SFRA has been widely used in various application scenarios related to fault diagnostics and prediction. The adopted approach to compare the two techniques involved the design and implementation of an experimental setup and a related measurement campaign to obtain a dataset relating to the EIS and one relating to the SFRA. Subsequently, several Machine Learning models were tested for performance comparison between the two techniques. The obtained results confirmed the goodness of the SoC estimate based on EIS, however equally satisfactory results were not obtained using the SFRA. This indicates that further investigations will need to be conducted, given the successful application of SFRA in other areas, including changing estimation targets such as health status or battery failure prediction.