[Paper Review] Parameterising continuum level Li-ion battery models & the LiionDB database

Ange Wang, Simon E. J. O’Kane+10 more|arXiv (Cornell University)|Oct 19, 2021

Advanced Battery Technologies Research210 references4 citations

TL;DR

This paper presents a comprehensive review of parameterisation methods for the Doyle–Fuller–Newman (DFN) continuum-level lithium-ion battery model, a widely used framework for simulating cell-scale electrochemical behavior. It introduces LiionDB, a searchable, open-access database aggregating experimentally derived parameters and their metadata, enabling consistent, reproducible, and physically accurate model calibration across diverse battery chemistries and designs.

ABSTRACT

The Doyle-Fuller-Newman framework is the most popular physics-based continuum-level description of the chemical and dynamical internal processes within operating lithium-ion-battery cells. With sufficient flexibility to model a wide range of battery designs and chemistries, the framework provides an effective balance between detail, needed to capture key microscopic mechanisms, and simplicity, needed to solve the governing equations at a relatively modest computational expense. Nevertheless, implementation requires values of numerous model parameters, whose ranges of applicability, estimation, and validation pose challenges. This article provides a critical review of the methods to measure or infer parameters for use within the isothermal DFN framework, discusses their advantages or disadvantages, and clarifies limitations attached to their practical application. Accompanying this discussion we provide a searchable database, available at www.liiondb.com, which aggregates many parameters and state functions for the standard Doyle-Fuller-Newman model that have been reported in the literature.

Motivation & Objective

• To address the critical challenge of accurate and consistent parameterisation in the isothermal DFN model, which is essential for reliable battery simulation.
• To identify and clarify the assumptions, limitations, and methodological inconsistencies in existing experimental and fitting techniques used to extract DFN parameters.
• To establish LiionDB (www.liiondb.com) as a centralized, community-driven repository for DFN model parameters with full provenance and metadata.
• To improve model robustness by highlighting error propagation risks and identifiability issues in parameter estimation.
• To guide researchers in selecting appropriate parameters based on experimental conditions and model assumptions, enhancing predictive accuracy in battery design and analysis.

Proposed method

• Systematic review of experimental and fitting methods used to infer DFN parameters, including electrochemical impedance spectroscopy, galvanostatic intermittent titration, and microstructural measurements.
• Development of a structured, relational database schema (LiionDB) to store parameters, associated papers, materials, and measurement methods with primary and foreign key relationships.
• Incorporation of metadata such as input and temperature ranges, thermal scaling, and notes to ensure context-aware parameter usage.
• Application of nondimensionalisation techniques to assess parameter identifiability and determine which parameters can be reliably extracted from specific experiments.
• Use of a multi-layered data model to link parameters to their source literature, materials, and experimental methods, enabling traceability and cross-comparison.
• Implementation of a web-based interface (www.liiondb.com) to allow community contributions, search, and filtering of parameters by material, method, or application context.

Experimental results

Research questions

• RQ1 What are the most reliable and consistent methods for measuring or inferring DFN model parameters such as solid-state diffusivity, ionic conductivity, and electrode porosity?
• RQ2 How do assumptions in experimental techniques—such as electrode geometry or particle shape—impact the accuracy and transferability of inferred parameters?
• RQ3 To what extent do errors in auxiliary measurements (e.g., electrochemically active surface area) propagate into final parameter estimates for the DFN model?
• RQ4 How can a centralized, curated database improve the reproducibility and consistency of battery model parameterisation across the research community?
• RQ5 Which parameters in the DFN framework are identifiable from specific experimental setups, and how can this be used to guide efficient model calibration?

Key findings

• The DFN model remains a powerful and widely used framework for cell-level battery simulation due to its balance of physical fidelity and computational efficiency.
• Many DFN parameters are highly method-dependent, and their values can vary significantly based on experimental assumptions, such as particle geometry or electrode loading conditions.
• The LiionDB database now hosts a curated collection of DFN parameters with full provenance, including source literature, material type, measurement method, and operational ranges.
• Error propagation is a significant concern, particularly for parameters like solid-state diffusivity, which depend on secondary estimates such as electrode surface area.
• Nondimensionalisation reveals that not all DFN parameters are identifiable from a single experiment, highlighting the need for multi-method calibration strategies.
• Community contributions to LiionDB are encouraged and expected to enhance the database’s comprehensiveness and utility over time.