Accelerating Electric Vehicle Battery Innovation with Multiphysics Simulation
|Figure 1: Contours of velocity on mid-plane of a 20-cell module. In this example, water is being used as a coolant for the module, flowing in through the top channel and out the bottom channel.|
Rising to the challenge, General Motors (GM) is collaborating with ANSYS and ESim LLC as part of the Computer-Aided Engineering for Electric Drive Vehicle Batteries (CAEBAT) program championed by the DOE’s National Renewable Energy Laboratory (NREL). Together, this team aims to design cell- and pack-level design tools to aid in the development of safe, reliable and high-performance battery packs. By utilizing simulation tools to accurately represent cell and pack multiphysics phenomena, this group can identify end-user needs, establish requirements, integrate and enhance existing sub models, and perform experimental testing to validate the tools.
Designing a battery cell model
The optimal cell-level model will predict multiphysics responses of large-format (capacity greater than 5 amp-hours) lithium-ion battery cells. Since temperature greatly affects the performance, safety and life of the battery pack, thermal management is critical, as heating and cooling create uneven temperature distribution. Any temperature variation can lead to electrically unbalanced modules, lower performance and shortened battery life. For best performance, temperature variations must be minimized across each cell and between cells — if the temperature is too low, the power/energy output is reduced; if the temperature is too high, the cell life is limited.
The proposed cell-level model will provide a seamless connection among electrochemical, thermal and structural responses. It will demonstrate how design changes can influence overall cell performance, such as thermal, electric power, capacity, safety, state of charge and life, and internal imbalance within the cell due to spatial variations of current density and temperature. One-dimensional electrode-scale models will help engineers understand the physics, such as electrochemical kinetics, lithium diffusion and transport, and charge conservation and transport, while a 3-D model is needed for realistic cell geometry that includes cell-level performance.
Battery service life greatly impacts the total lifecycle cost of electric-drive vehicles, and the complex dependency of the cell capacity on time, temperature, voltage, number of charge/discharge cycles, electrode microstructure, and depth of discharge is often neglected in cell models. To simulate cell aging and degradation, electrode-scale predictions will be enhanced by introducing models that account for capacity fade due to mechanical degradation caused by thermal and mechanical stress and loss of active material due to film formation.
|Figure 2: Advances in cosimulation and reduced-order model technologies will be options for system simulation to guide pack design.|
One primary value of battery computational modeling tools is improving the thermal management of the battery pack, which should maintain an optimum average temperature that operates in surrounding environments ranging from -40°F to 122°F. Lithium-ion batteries operate best between 77°F to 95°F, but this range is difficult to maintain due to the varying environments of normal vehicle operation. Active cooling and heating of the battery pack is a challenge due to constraints on cost, power, weight and volume. Design solutions include heat sinks, air jet impingement, micro-channel cooling, heat pipes, immersion cooling and spray cooling (Figure 1).
At present, these concept evaluations and the resulting battery life predictions are based on very time-consuming hardware build-and-test iterations, but could be streamlined by using effective pack-level simulation. Using existing CAD models is the fastest way to create common cell and pack geometries for analysis. The geometry interface’s capability, as well as the replication and parts library function of simulation software, is helpful when geometry and properties information must be input manually. Simulation platforms, such as ANSYS Workbench, can handle complex 3-D geometry of battery cells, including current-collector tabs, encapsulation materials and structural support details.
To solve electrochemical, thermal and fluid behavior at the cell or module level, fully discretizing the cell and coolant channels and performing field simulation is needed. Direct scaling to the pack level is not practical, particularly with the fluid dynamic complexity of some micro-channel liquid cooling strategies. Although parallel code implementation can reduce execution times with high-performance computing hardware, this strategy alone is insufficient due to long development time and the associated number of pack-level simulations.
System simulation offers a computationally inexpensive alternative, but the current state-of-the-art relies on relatively uncoupled and simplistic cell representations, such as equivalent-circuit models. As battery power densities increase and thermal management becomes more complex, these system simulations have neither the reliability nor the resolution necessary to guide pack design. As such, the pack-level tool will need to combine field simulation from ANSYS Fluent and ANSYS Mechanical, as well as system simulation with ANSYS Simplorer through straightforward run-time coupling, or cosimulation (Figure 2). Additionally, the development of innovative reduced-order models (ROMs) will enable field-simulation models to be automatically distilled into a system model with a controllable balance between fidelity and cost.
|Figure 3: A direct coupling between design modeling tools and vehicle simulator allows feedback from a battery system, such as temperature deviation through a pack, to be used for system control criteria during simulations.|
Just as important as the cell- and pack-level models is the availability of interfaces to enable these new tools to interact with other current and future battery models. By contributing to the open architecture software (OAS) activity led by Oak Ridge National Laboratory and adopting specifications for input and output file formats and standard communication protocols, the team’s cell- and pack-level tools can exchange information with tools developed by other CAEBAT teams.
Simulation software can be the backbone of the entire battery-modeling workflow, which also can connect peer-to-peer with other software environments. This flexibility will be instrumental in the development of CAEBAT OAS interfaces and as the team demonstrates the ability to select different battery models with different physics solvers and scales. The flexibility also will be vital for the integration of newly developed tools into its proprietary process automation (PA) engineering environment.
Process automation and robust engineering
Given the complex development process, it is important that the various evaluation and optimization models are not only accurate and consistent, but also represent the most current design in the product development process. Complementary PA tools will follow after the project’s conclusion to enable rapid and reliable comparison of design alternatives, design exploration, robust assessments and optimization.
PA tools incorporate process guidance, which is useful for implementing enterprise-wide standardized work. Components of standardized work typically include following modeling guidelines, applying correct boundary conditions for analysis, post-processing and reporting analysis results. This is essential to assure decision makers that the evaluation and reporting of analysis metrics is consistently performed. GM, for instance, has established methodologies for the engineering areas of noise and vibration, crash analysis and body structural optimization. The battery design modeling tool developed in this project can be coupled with a vehicle simulator to evaluate thermal and electrical responses of a particular battery pack for a given power load profile associated with a specified vehicle driving condition (Figure 3).
Verification and validation
The accuracy and usefulness of the proposed cell-level tool will be demonstrated through rigorous verification and validation processes. Following a math model validation process developed by the National Institute of Statistical Sciences, a test database will be generated for physical validation of the nominal heat source model, as well as cell-level electrical and thermal performance.
This data will provide thermal characterization with known heat source and thermal boundary conditions for various battery thermal management system operation conditions. The total heat generated from the pack is calculated from coolant in-flow and out-flow rates and temperatures. Pack-level specific heat and thermal conductivities are derived from individual cell thermal properties and other electrical and electronic equipment onboard the pack. In addition, thermistors are placed at every module to monitor hot spots across the pack. These data are crucial for validating the coolant flow field around the battery pack.
At the pack level, the state-of-the-art battery will be significantly advanced by the development of innovative ROMs, derived and calibrated from the cell-level models and carefully validated through experiment. The CABEAT project team is incorporating the latest advances in battery modeling research with software tools that are unsurpassed in their ease of use and workflow automation for robust design optimization. With a strong plan for rapid deployment to the industry, these project results will help accomplish the ultimate goal of accelerating the pace of battery innovation and development for future electric-drive vehicles.
Taeyoung Han is a technical fellow of the Vehicle Development Research Lab at General Motors Research and Development Center. Gi-Heon Kim is a senior research engineer for the Center for Transportation Technologies and Systems at NREL. Lewis Collins is the director of Software Development at ANSYS. They may be reached at editor@ScientificComputing.com.