Balancing resistor-based online electrochemical impedance spectroscopy in battery systems: opportunities and limitations

Battery management systems (BMS) are crucial for maintaining the health and performance of battery systems, particularly in applications like electric vehicles. A key function of a BMS is cell balancing, which aims to equalize the state-of-charge (SoC) across individual cells to ensure uniform operation and extend lifespan. Dissipative balancing, where excess energy is converted into heat via a resistor, is a common and cost-effective approach. This research focuses on leveraging the inherent capability of dissipative balancing systems – the balancing resistor – for conducting online electrochemical impedance spectroscopy (EIS). EIS is a powerful diagnostic tool that provides insights into the internal state of a battery, including temperature, aging, and state-of-health. The ability to perform online EIS directly through the existing balancing resistor would eliminate the need for additional hardware, simplifying the BMS design and potentially lowering cost. However, the accuracy and reliability of EIS measurements performed in this manner need careful investigation and optimization. The research question is thus to evaluate whether the signals from existing dissipative balancing systems can be used for impedance analysis, and to propose methods for achieving the necessary signal-to-noise ratio (SNR). The current literature extensively discusses both dissipative and non-dissipative balancing techniques, and the diagnostic value of EIS is well-established in various battery-related research. However, there is a gap in understanding the combined application of dissipative balancing resistors for online EIS. This study addresses this gap, aiming to provide a practical and cost-effective solution for enhancing BMS diagnostics.

The existing literature extensively covers the various aspects of battery management systems, cell balancing techniques, and electrochemical impedance spectroscopy. Studies have thoroughly examined the advantages and disadvantages of dissipative versus non-dissipative balancing [1, 2]. The benefits of EIS for in-depth battery diagnostics such as temperature or state-of-health monitoring have been established and are gaining recognition in industry [3]. Various methods for performing EIS have been explored, including passive and active excitation techniques [4]. The study also references works investigating the self-discharge behavior of battery cells and its correlation with various parameters, such as cell voltage and specific surface area [5, 6, 7]. Furthermore, the literature addresses the impact of fast-charging strategies on battery health [12], and the challenges of implementing online EIS on embedded systems [23]. However, a comprehensive analysis of using the existing dissipative balancing resistors for online EIS is lacking, making this study a significant contribution to the field. The study also draws upon existing knowledge regarding thermal management of balancing resistors [13, 14, 15] and the impact of signal processing techniques on SNR in measurement systems [32, 37, 38].

The research involved both theoretical analysis and experimental validation. The theoretical part focused on sizing an active dissipative balancing system, considering parameters like battery self-discharge rate, operating strategy, charging system characteristics, thermal limitations of resistors, and sense wire sizing. A formula was derived for normalizing the balancing resistor resistance to the battery capacity (Equation 1), allowing for comparison across different battery systems. The impact of balancing time and minimum balancing voltage on resistor size was also analyzed (Equation 4). The relationship between the balancing resistor and the charging system was explored, highlighting the importance of aligning the resistor value with the charger's minimum power output. The study also detailed the influence of sense wire resistance on voltage measurements, emphasizing the need for four-terminal pair connections to improve accuracy. The theoretical section then delved into the signal-to-noise ratio (SNR) of impedance measurements. Various sources of noise and distortion were considered, including quantization noise from the analog-to-digital converter (ADC) (Equation 8, 9, 10), jitter (Equation 11), thermal noise (Equation 12), and pink noise. Equations were derived for calculating the SNR contributions from the ADC (Equation 9, 10), jitter (Equation 11), thermal noise (Equation 12), and the effects of signal processing techniques, such as filtering, averaging, and the Fast Fourier Transform (FFT) (Equation 13, 14, 15). Finally, a combined equation (Equation 17) was formulated to estimate the overall SNR considering all the factors involved. This analysis revealed the relationship between different parameters, illustrating that quadrupling the FFT size, sampling rate, or the number of averaging values is equivalent to increasing the ADC resolution by one bit or reducing the filter bandwidth by a quarter (Equation 18). The experimental work involved the design and construction of a demonstrator for testing the feasibility of online EIS using an existing balancing resistor. The demonstrator was integrated into the battery pack of an electric van, with a second balancing resistor added in parallel to the standard BMS balancing system. The hardware setup is described in detail, including the choice of microcontroller, ADC, DAC, and other components. The experimental procedure included conducting EIS measurements at various frequencies over a range of battery SoC. The measurements were compared with reference measurements obtained using commercial laboratory equipment. Temperature monitoring of the board was also performed using a thermal camera.

The research demonstrated the feasibility of online EIS measurements using a balancing resistor within a BMS. The theoretical analysis provided quantitative estimates of the SNR at different stages of the measurement process. The formula derived for normalizing the balancing resistor resistance to battery capacity allows for a consistent comparison across various battery systems. The experimental results from the demonstrator, although showing some deviations from reference measurements, demonstrated good correlation, particularly at higher SoC. The SNR of the impedance measurements decreased at lower frequencies, suggesting that pink noise is a significant factor influencing measurement accuracy. The peak temperature reached during the experiment was approximately 47°C, suggesting a need for careful thermal design when implementing this approach in real-world applications. For a specific battery (lithium-iron-phosphate CATL 271 Ah), a large excitation current would be needed to achieve a typical target voltage response for EIS in laboratory settings; however, lower voltages are possible if the SNR permits. The study also found that increasing the number of cells measured in parallel reduces voltage SNR by approximately 26 dB, illustrating the challenges of performing online EIS on large battery packs. The results showed that using signal processing techniques such as increased sampling rate, FFT size, or averaging could mitigate the impact of lower analog resolution in the ADC, highlighting a path towards cost-effective implementation. The analysis confirmed that using the balancing current to actively excite the system, instead of just using it as the current source, would improve the SNR of impedance measurement by making it only dependent on the voltage SNR. The authors suggest that further investigations are needed to refine the self-discharge model and improve the system design through more refined signal processing and thermal management.

The findings address the research question by demonstrating that online EIS using a balancing resistor is technically feasible. The significance of this research lies in its potential to provide a cost-effective method for enhancing battery diagnostics within existing BMS designs. The ability to perform online EIS without requiring additional hardware represents a major advantage, potentially improving the safety and reliability of battery systems. The observed decrease in SNR at lower frequencies suggests that mitigation of pink noise will be important in future system designs. The correlation between the experimental results and reference measurements indicates that online EIS using balancing resistors can provide valuable insights into battery behavior. However, it also demonstrates the need for careful consideration of thermal management to avoid excessive heat generation. Future research should focus on developing more sophisticated signal processing techniques to further improve SNR and accuracy. The optimization of balancing strategies to minimize energy loss and heat generation while maintaining sufficient excitation for EIS is crucial for practical application.

This study successfully demonstrated the feasibility of online EIS using existing balancing resistors in battery systems. The research provides a method for sizing balancing resistors and quantifies the impact of various parameters, including self-discharge rate, operating strategy, and signal processing techniques, on the accuracy of the measurements. While challenges remain in terms of heat generation and achieving high SNR at low frequencies, the presented methodology offers a path towards a cost-effective solution for enhancing battery diagnostics. Future research could focus on optimizing signal processing algorithms to further enhance SNR, developing advanced thermal management strategies, and extending the approach to larger battery systems.

The demonstrator used in this study was not optimized for size, weight, or cost, limiting its immediate applicability to real-world applications. The experimental setup involved testing a single cell or small battery pack; further investigations are needed to verify the performance of this approach in larger battery systems. The noise analysis focused primarily on quantization noise, jitter, thermal noise, and pink noise; other potential sources of noise and distortion were not fully investigated. Finally, the interpretation of the impedance data was limited; future work will involve detailed modeling and analysis of the impedance spectra to extract meaningful information about battery health.