EXTENDED ABSTRACT: Lithium-ion batteries (LIBs) have become the most-used energy storage device in modern hybrid and electric vehicles, resulting in a huge market in China. However, the design of LIB depends strongly on a method of “test and error”, which is time-consuming and labor-intensive. Computer aided engineering is a popular means to accelerate R&D in industrial issues in nowadays, which also sheds light on the design of LIB. Mathematical models for LIB dynamics mainly fall within two categories: Equivalent Circuit Models (ECMs) and Electrochemical Models (EMs). ECMs use equivalent electrical components to represent the dynamic behavior of the battery, which therefore is widely used for battery management system but not battery design. In contrast, EMs explicitly consider the chemical and physical processes that take place in the battery, which is more suitable in battery design. The most widely used EM in the literature is the pseudo-two-dimensional (P2D) model [1] based on porous electrode theory, which is described by a set of tightly coupled and highly nonlinear partial differential-algebraic equations. Although great success that has been achieved by the P2D model, the application of it is limited in a lot of scenarios. For example, the real complex porous electrode structure is simpliffed as a homogeneous medium which makes the inhomogeneous spatial information neglected. Besides, many complex physical processes have been modeled by simple mathematical descriptions (e.g., Fick diffusion model to describe diffusion inside active particles), and thus a more rational physical model is still absent. On the other hand, understanding, modeling and coding of such kind of model is not easy, makes the using of EMs to design battery an impossible mission for battery engineers. As a consequence, a software based on the real physical processes as well as complex electrode structures is proposed. The software is developed with two goals in mind to facilitate the design of LIB: 1) simple and easy to use; 2) providing as more as possible options to manipulate the electrode structure to achieve better LIB performance. Figure 1 is an example of our software interface that has been installed inside 21C Lab of CATL. Users can modify the electrode structure involving length, porosity, particle size etc. easily.To satisfy the goal 2), the architecture of our software is divided into two categories: homogeneous model and heterogeneous model. The former model is a P2D type one, which incorporates more physical facts than the original P2D model but still keep a high computation efffciency. The latter model is a 3D one considering real complex electrode structures, which can capture the most comprehensive spatial information by paying a much higher computation cost. In the following we will introduce these two models in details. In the homogeneous model, heat production and transfer have been incorporated, which therefore can be used to describe temperature variation during charging, discharging or cycling. Besides, both concentration gradient driven and chemical potential driven lithium diffusion inside active particles have been considered, supporting a fast prediction of experimental results and also mechanism analysis. Figure 2 shows an example. In this case, charging was found to terminate at earlier SOC as the charging rate increases, and the reason was revealed by the simulated concentration distribution in particles. Ascribing to the phase separation behavior introduced by chemical potential driven lithium diffusion, Li concentration at surface arrives upper limit earlier as rate increases, which suppresses further chemical reaction at the particle/electrolyte interface and leads to the phenomenon in Figure 2a. Although the features are more complicated, our software still shows a far higher computation efffciency as compared to COMSOL.The heterogeneous model is a modiffed electrochemical model with the so-called smooth boundary operation [2]. In this framework, the chemical potential driven diffusion, anisotropic diffusion as well as stress ffeld are incorporated and coupled. Since the real complex electrode structure has been considered, the analysis of the distribution information of concentration, voltage etc. become possible, as shown in Figure 3. Therein, the processes of Li transferring from anode to cathode during discharging can be “seen” directly. Three kinds of phase separation: interfacial, bulk and inter-particle are found, which however cannot be totally predicted by a homogeneous model.The software is still far from comprehensive in its current state. In the future, we will dive deeper in mainly these directions: 1) integrating with multi-physical field solver in our software, pave the way to more complicated battery service environments; 2) developing a multiscale framework from microscale for intrinsic properties of materials, to mesoscale for cell properties and ffnally to macroscale for battery modules and systems; 3) combining physical models with AI techniques, to improve the usage efffciency of ffnite (or few) experimental results. By interweaving multi-physics, multiscale and complex physical processes under real work conditions of batteries, we convince our software can be a real computer aided tools for LIBs.

Keywords: Lithium-ion battery; Electrochemical; Phase separation; Software; Physical model

REFERENCES:

[1] M. Doyle, T. F. Fuller, and J. Newman, J. Electrochem. Soc., 140(1993), 1526 [2] Hui-Chia Yu, Hsun-Yi Chen and K. Thornton, Modelling Simul. Mater. Sci. Eng. 20 (2012) 075008