Abstract

Authors: Nedka Dechkova Nikiforova^1, Rossella Berni^1, Lorenzo Ciani^2, Gabriele Patrizi^2, Antonio Pievatolo^3
Affiliations: 1. Department of Statistics Computer Science Applications ”G. Parenti”, University of Florence, Florence, Italy
2. Department of Information Engineering, University of Florence, Florence, Italy
3. Institute for Applied Mathematics and Information Technologies “E. Magenes”, National Research Council, Milan, Italy
Abstract: This talk deals with the advancement of experimental planning strategies for studying battery degradation and lifetime performance. In a first battery study, we specifically consider the peculiar battery life characteristics and evaluate the environment in which the experiment is conducted. In this case, the laboratory ensures a steady and controlled situation with respect to environmental factors; moreover, all trials are performed by the same operator. Therefore, we observe two structural levels: an upper level, where the climate chamber and temperature drive the degradation mechanism of the battery life cycle, and a lower level, which enables us to evaluate the impacts of the manufacturer and current on the battery characteristics. A split-plot design is particularly appropriate for this study, given the presence of these two structural levels. That is, the climate chamber and the temperature are whole-plot factors, while the manufacturer and the current are sub-plot factors. The study utilizes a total of sixteen batteries, with no replicates; the discharge capacity is the response of interest.
Expanding upon these findings, in a second study, a non-linear hierarchical mixed-effects framework is considered to accommodate more complex degradation dynamics. In this second study, the experimental planning is broadened to also consider two different laboratories where the trials will be performed, as well as two different battery usage profiles, i.e., an automotive versus a non-automotive one. Therefore, we deal with battery data into a three-level hierarchical structure as follows: i) laboratory (level-1), ii) battery manufacturer (level-2), iii) battery usage profile and current rate (level-3). By considering the experimental planning performed at several hierarchical levels, we aim to build a hierarchical D-optimal design. To this end, a specific non-linear degradation model is also defined, including both fixed and random effects. Therefore, the Fisher Information matrix is derived, also considering the defined hierarchical structure; the results are currently under development, also including the specific algorithm for obtaining the final hierarchical D-optimal design. This comprehensive approach provides a valuable framework for studying and improving the reliability of battery lifetime. Finally, it is worth emphasizing that this presentation focuses exclusively on the experimental planning; subsequent phases, such as data analysis and modelling, fall outside the scope of this contribution.
Acknowledgements: The study has been supported by Bando PRIN 2022- 2022WBN75S - E3DM - Experimental Design and Maintenance, a Decision-Making approach driven by Degradation Models; CUP master: B53C24006390006; CUP: B53C24006400006.