Capacity loss suppression in lithium-ion batteries is essential to make electrifed technologies (grid energy storage, electric, and hybrid vehicles) more reliable. Further, on-the-fly prediction of the life of individual batteries is a necessary requirement to develop battery management systems (BMS) to establish strategies that allow to extend their life, schedule replacements, and even anticipate catastrophic failures, very much ahead of the development of a failure event, particularly in commercial electric vehicles. Available models possess limited accuracy and are conned to a narrow set of selected battery chemistries, architectures, and operating conditions. In this context, the development of practical, on-the-fly physical descriptions has the potential to provide the optimal background to deliver dramatically advanced on-board strategies to further reduce costs, extend life, and maximize performance and device reliability. A physics-based, reduced order framework was developed to calculate the charge capacity loss contributions from spatially homogeneous and heterogeneous degradation mechanisms, chemomechanical cycling, and initial capacity recovery. The formulation goes well beyond prevalent coulomb-counting models and is tuned solely based on experimentally measurable parameters, although it neglects any buildup of internal resistance. The model was compared against the largest data set available to date for commercial LFP|graphite cells and shows less than 10% error for 92% of the cells. Results suggest that in most cells the charge capacity increases through the first 50 cycles, beyond which homogeneous SEI growth dominates capacity loss up to 500 cycles. Above 500 cycles and at high current densities the model attributes capacity loss primarily to heterogeneous SEI growth proceeding at microstructurally favored locations, further assisted by chemomechanical failure of the graphite anode particles towards the end of cell life. A Monte Carlo analysis of the initial capacity increase as a function of the cycle number demonstrates the existence of two battery populations, enabling the prediction of battery degradation with 99% confidence. The developed model sets the stage for on-the-fly capacity loss calculations in hybrid and electric vehicles, especially at low currents and constant current voltage holds and could be extended to capture the deleterious effects of high current densities in fast-charging scenarios by including side reactions and resistive losses.

jupyter; python

Aniruddha Jana, A. Surya Mitra, R. Edwin García

Many thanks for the support to the Toyota Research Institute, USA.

Researchers should cite this work as follows:

• Aniruddha Jana, Surya Mitra Ayalasomayajula, Edwin Garcia (2020), "romdegradation: Physics-based, Reduced Order Degradation Model of Lithium-ion Batteries," https://nanohub.org/resources/romdegradation. (DOI: 10.21981/KGET-D846).

  BibTex | EndNote

1. batteries
2. battery degradation
3. battery modeling and simulation
4. lithium ion batteries
5. rechargeable batteries