Real-Time Under-Load Electrochemical Impedance Spectroscopy (EIS) Analysis and Modeling

Abstract

All batteries degrade, but the challenge comes in quantifying that degradation and ultimately predicting a battery's failure and/or end-of-life. Battery health and monitoring have become more imperative with larger deployment of lithium batteries in electric vehicle (EV) and energy storage system (ESS) applications. Research shows a strong correlation between battery degradation and AC impedance measured using electrochemical impedance spectroscopy (EIS). This indicates that AC impedance can be used as a proxy for battery health. Additional information, like a battery's internal temperature can also be extracted from EIS measurements. As such, there is a strong push to measure AC impedance of a battery in field diagnostic applications. However, EIS is typically performed at rest under very stable environmental conditions. This traditional method of interpreting EIS must be modified to account for changing environmental conditions and changing current loads like those seen in EV and ESS applications. This paper explores the prospect of using EIS in a real-time battery management system (BMS) with the primary focus on determining the limits for interpreting EIS under loaded current conditions. An experiment is conducted to evaluate using EIS measurements taken on a loaded battery for the purpose of temperature estimation.A set of experiments were conducted where EIS was performed on Samsung 18650 Nickel Manganese Cobalt (NMC) cells while the cell was subject to loaded conditions. These experiments consisted of load currents ranging from charging rates (C-rate) between C/5 to 2C, and discharging rates (D-rate) between D/5 to 2D. Test were conducted at temperatures in the range of 5°C to 35°C. Closer inspection of Figure 1 shows EIS spectra diverging for charging conditions at lower frequencies. This is believed to be caused by a drift in the battery’s SOC. During charging, lower frequencies in AC impedance tend to have lower magnitude. Conversely, higher AC impedance magnitude is observed during discharging. The EIS frequency at which the loaded impedance deviates from the unloaded impedance by >10% is determined. The curve in Figure 2 represents the lowest EIS frequency at which under load EIS spectra is deemed usable, meaning that the shaded region below the curve represents the unusable portion of the EIS spectra. Moreover, this relationship serves as groundwork upon which to correct for EIS under loaded conditions. It is believed that the relationship of frequency limit as a function of load current shown in Figure 2 is slightly skewed by battery heating as an artifact of the test procedure. As a result, this experiment will be repeated prior to the final paper to gather data that accounts for these heating effects. A method to correct for shifts in an under-load EIS spectra will also be presented in the final paper.The prospect of using an under-load EIS spectra for internal temperature prediction is also demonstrated on a lab scale. In this experiment, the Samsung cell was soaked in a thermal chamber at constant temperature for several hours. A series of 3D discharge pulses are then applied to the battery for various durations. EIS is performed back-to-back during the current pulses (i.e., under load) and during rest times. The EIS spectra are fed into a cell temperature estimation model based on a non-zero interrupt frequency (NZIF) method. Figure 3 shows how EIS spectra changes with temperature for both loaded and unloaded conditions. The details of the NZIF model are beyond the scope of this paper. Thermistors are placed on the surface of the battery to measure a reference temperature. The results of the NZIF-predicted internal temperature are contrasted with measured thermistor battery surface temperatures, and those results are shown in Figure 4. These results indicate that NZIF-predicted internal temperature rises and falls during discharge and rest events, consistent with expected temperature rise and thermistor-measured battery surface temperature.Preliminary analysis suggests that constant current loads only slightly skew no-load EIS measurements, and that most of the EIS spectra taken under load is still usable. Furthermore, the usable portion of an EIS spectra can be determined by an EIS frequency vs load current relationship. This paper demonstrated that EIS can be used for internal temperature prediction while the battery is subject to a loaded condition, which is potentially useful to improve safety and performance of battery packs in field applications. In future work, more EIS measurements will be taken under different loaded conditions to improve NZIF predictions of internal temperature. Nevertheless, these early results are extremely promising for the use of EIS in real-time BMS applications. Figure 1