Multiphysics Modelling of Energy Storage Devices Using Pore Scale Approaches

Abstract

The advancement in high-resolution X-ray tomography image acquisition techniques has enabled imaged-based modelling of pore-scale transport processes to better understand structural performance relationship in porous media. The porous components in electrochemical energy storage devices such as lithium-ion batteries, fuel cell and redox flow batteries are subject to intense research to maximize performance and hence reduce the cost of energy storage systems. The image-based pore-scale modelling approaches such as direct numerical simulation (DNS) are, however, very computationally expensive and it gets infeasible to simulate a representative element volume of porous structure on a standard workstation or laptop machine. Pore network modelling (PNM) approach has been previously used to simulate large size porous domains of fuel cell and redox flow batteries at substantially lower computational cost, however, its application in lithium-ion batteries has not been attempted due to the multiphysics and transient nature of transport mechanism involved during charging and discharging process. Lithium-ion batteries are considered as the top candidate for electrochemical energy storage, so modelling their structure-performance relationship at less computational cost will enable development of efficient numerical pore network modelling framework. Therefore, this thesis aims towards developing pore network modelling framework for lithium-ion batteries to study the impact of microstructure on multiphysics transport processes occurring inside battery electrodes. The development of lithium-ion battery pore network model requires enhancements in the current implementation of pore network modelling algorithms. For example, current pore network extraction algorithms only extract a single phase from a tomography image (usually the pores). On the other hand, lithium-ion battery electrodes contain three phases, namely active material (e.g. NMC), carbon binder, and electrolyte filled void phase. To resolve this issue, multiphase pore network extraction algorithms were developed that connect any two phases via interconnections. This allowed investigating inter- and intra-phase transport processes between phases which are common in lithium-ion battery. The extraction algorithms were tested on random sphere packings and three-phase lithium-ion battery cathode and found to agree well with experimental data and DNS model. Computational performance of PNM model was also compared with other modelling approaches and found to give appreciable performance gain on large size porous domains, while yielding similar or equivalent results.Although modelling of transport process using the PNM approach is computationally efficient, extracting pore networks from tomography images is a computationally expensive task. Also, image resolution plays a vital role to determine the relative accuracy of extracted geometrical properties and hence simulation accuracy. To remove these bottlenecks, an efficient, parallelized network extraction technique was developed that enabled pore network extraction from massive size images. A geometric domain decomposition technique was adopted to reduce the computational cost of extraction. The network extraction was observed 7 times faster and consumed 50% less RAM when used in parallel and serial mode respectively. Finally, a case study was performed to reduce the effect of resolution during pore network extraction. This enabled more reliable extracted pore networks for pore network modelling studies. Finally, pore network modelling of lithium-ion batteries cathodes was performed to study galvanostatic discharge behaviour of half-cells. A massive reduction in computational cost was observed when compared with DNS approach. The structural features of two electrodes were investigated to understand the performance-structural relationship. Also, particle-to-particle and pore-to-pore analysis was performed to analyze the state of lithiation, solid-phase potential distribution and lithium-ion concentration distribution, electrolyte phase potential distribution in solid and electrolyte phase respectively. The study enabled modelling of large size lithium-ion electrodes to analyze the impact of internal microstructure on the overall performance of the cell. The presented work in this thesis is focused on developing, validating, and applying a pore network modelling framework for lithium-ion battery discharge. It has enabled the study of structural performance relationship of battery electrodes on a particle to particle basis without estimating effective transport properties using empirical or experimental data. The excellent computational performance of PNMs has allowed multiphysics modelling on standard workstation or laptop with minimal computational resources. Although developed for lithium-ion battery cathodes the developed framework can be used for any anode structure or study thermal performance-structure relationship as well.