Simulating the Temperature Spread within a Commercial Li-Ion Battery Module – A performant and non-destructive Characterization and Modeling Process

Summary:

Modeling the effects of thermal cell-to-cell variations within a so-called digital twin running in parallel to the physical battery system offers various advantages, especially for large-scale and investment-intensive automotive, marine or stationary systems. In this context, FEM/CFD approaches as well as thermal equivalent circuit models represent the current state of the art in calculating the temperature distribution. However, due to their demanding parameterization and high computational effort, those existing approaches are rarely applied to complex systems consisting of thousands of cells as common in the named applications. This presentation therefore presents a novel model simplification approach for thermal equivalent circuit models, which allows the calculation of the average, minimum and maximum temperature of the system at a fraction of the computational cost of the full-scale model, which calculates cell-specific temperatures for every cell in the system. The investigations are conducted within the EU-funded research project “Nautilus”, which investigates hybrid battery and fuel cell energy systems for next-generation cruise ships.
During the investigation, a commercial Li-ion battery module with 64 Ah NMC pouch cells in 14s2p configuration is used. In the first step, a full-scale thermal equivalent circuit model is derived and parameterized. For the characterization of the heat transfer between the cells inside the module, a novel external heating experiment, which allows the determination of the cell-to-cell thermal resistance in a quick and non-destructive manner is conducted. The validation results for the full-scale model show a high model accuracy in the relevant temperature range. Running the full-scale model on a performant desktop pc results in a simulation time of 122.3 seconds per hour profile time. This computational effort, while being acceptable for the simulation of a single module, is not sufficient for the real-time, parallel simulation of a large-scale system consisting of hundreds of such modules. Therefore, in the second step a novel simplification approach reducing the full-scale model to a three-cell equivalent circuit model is applied. The simplification aims at representing the heat transfer between the outer (coldest) and center (hottest) cell using simple differential equations instead of simulating the intermediate cells. As shown during the validation, the resulting simplified model represents the minimum, maximum and average temperatures of the full-scale model with high precision while reducing the computational effort to 1.78 seconds per hour profile time. This severe reduction in computational effort allows the parallel simulation of large-scale systems within cloud-based digital twins.