Loading...
Interdisciplinary applications of mathematical methods for overcoming coarse experimental datasets in energy storage and particle dynamics models
Date
2025-12
Abstract
The global transition to net-zero emissions is heavily reliant on the advancement of lithium-ion battery (LiB) technology, which is crucial for renewable energy integration and electric vehicle adoption. Enhancing the performance, safety, and life-cycle efficiency of LIBs requires sophisticated modelling approaches across the battery’s lifecycle. This thesis presents two integrated studies that address key modelling challenges at different stages: the manufacturing process and operational state of charge (SoC) monitoring.
The first study focuses on operational performance by evaluating machine learning (ML) architectures for accurate SoC estimation, a critical function of battery management systems. Using an experimental dataset from commercial LIB cells, this work compares the performance of a multilayer perceptron (MLP), a long short-term memory (LSTM) network, and a nonlinear autoregressive with exogenous input (NARX) neural network. The results indicate that MLP and LSTM models perform effectively with smaller training datasets, while the NARX model requires more extensive data for accuracy and shows greater sensitivity to hyperparameter tuning. This study has already been published in the Journal of Power Sources [Rae, Ottaviani, Capkova, Kazda, Santa Maria, Ryan, Passerini & Singh 2025]. The interdisciplinary application and adaptation of these models is mentioned and linked to the development of forecasting models in seaweed nutrition and modelling of pharmaceutical powders in twin screw granulation.
The second study investigates semi-analytical approaches to particle dynamics modelling for the slurry preparation phase of LIB manufacturing, a particulate system governed by complex coagulation and breakage dynamics. Coupled population balance models (PBMs) and computational fluid dynamics (CFD) are a powerful tool for such systems, but their computational cost is often prohibitive. This research provides a comprehensive review of recently developed semianalytical methods for solving the nonlinear Smoluchowski equation, which are mesh-free and offer efficient coupling with CFD. The analysis evaluates these approaches based on computational cost, truncation errors, and temporal behaviour, concluding that the modified variational iteration method (MVIM) and He’s variational iteration method (VIMHe) generally demonstrate superior performance. This review serves to guide experienced researchers by outlining the merits and demerits of leading methods, minimising unnecessary trial and error. The study presented in this component of the thesis has been extensively revised and is ready for submission to scientific journals during the submission of this thesis.
Collectively, this thesis provides significant insights into computational modelling for both the manufacturing and real-time battery management systems (BMSs) of LIBs. By advancing efficient PBMCFD coupling for manufacturing and establishing guidelines for robust ML-based SOC estimation, this work contributes to the development of safer, higher-performance, and more sustainably produced energy storage systems
Supervisor
Description
Publisher
University of Limerick
Citation
Files
ULRR Identifiers
Funding code
Funding Information
Sustainable Development Goals
External Link
License
Attribution-NonCommercial-ShareAlike 4.0 International