Unlocking Insights into Battery Systems: A Data Science Approach to Impedance Analysis

Abstract

Developing next generation battery systems requires new insights into how complex materials interact to provide the performance characteristics demanded by tomorrow’s energy storage applications. Electrochemical impedance spectroscopy (EIS) provides a tool for noninvasively probing these systems to uncover a signature of the underlying physicochemical processes. In particular, EIS measurements are useful for characterizing material properties1, studying degradation pathways2, and providing feedback for material design and optimization3. Typically, experimental impedance spectra are related to the underlying physics of the system by fitting an equivalent circuit analog. These models are comprised of any number of electrical circuit elements and make fitting the data to extract system properties relatively straightforward; however, the lumped parameters and degenerate nature of these circuit analogs can limit the available information in complex systems.4 While extensions to EIS, such as nonlinear EIS, can increase the informational content of impedance diagnostics by increasing the sensitivity to different physical parameters and interactions,5,6 taking full advantage of impedance-based techniques ultimately requires moving beyond simply fitting equivalent circuit analogs. For lithium-ion batteries, several rigorous physics-based models have laid the groundwork for a more sophisticated analysis of the impedance response.7,8Here we build on this work to analyze impedance spectra by leveraging insights from the pseudo 2-dimentonal (P2D) battery model. Using a large dataset of simulated impedance spectra for combinations of parameters across a physically meaningful space, the physics-based model can be fit to experimental spectra in a similar method to equivalent circuits (Fig. 1). This data driven approach enables a more robust fitting than trying to optimize the model directly and allows the sensitivity of each parameter to be explored around the fit. Finally, we show that this method is easy to use and invite you to use our open-source, web-based impedance analysis tool to visualize, fit, and interpret your own impedance data!