How to Create Thermal Models of Automotive Batteries in Ansys Fluent

Hello, my name is Mert Berkman. In this webinar, I will be talking about building thermal models of automotive batteries using ANSYS Fluent and TwinBuilder. Today, I will first introduce Ozen Engineering, which I am a part of, and then go into an overview of the modeling strategy.

The heart of the presentation will be a step-by-step model creation. We will see five different tasks to build the model, and I will talk in detail under each of these tasks. Ozen Engineering is a group of experts in simulations of structural, thermal, fluid, and electromagnetic fields.

We provide ANSYS support and licensing, training classes, mentorship, and consulting. We are located on the west and southeast coasts of the United States and do a lot of work with ANSYS. We are a channel partner and will show some ANSYS software tools throughout the presentation.

ANSYS offers strong engineering tools for structures, fluids, electronics, simulations, optics, VR, botanics, design, semiconductors, and many more engineering software options. Let's start talking about battery thermal modeling.

This chart, taken from the ANSYS Learning Hub, shows ANSYS's approach to battery thermal modeling. We will be working inside the red box, focusing on a module level.

We will start building our model in TwinBuilder, create a pull-and-circuit model, and then move on to the thermal side of the problem using ANSYS Fluent as our thermal solver.

We will build an LTI ROM, a reduced order model, and hook it up back to our TwinBuilder model, making it easier to use as a systems tool. The strategy for building the model involves a lumped electro-thermal modeling approach.

We will create an electric pull-and-circuit model for our cell, move on to a module ECM, set up the model in Fluent, generate and execute the CFD model, generate a thermal step response in Fluent, and convert this response into an LTI ROM.

This will generate quick thermal responses within our systems model in TwinBuilder. Once we couple the ROM and the ECM, we will have our full battery system model.

To build an ECM model, we need HPC data, which provides information on battery discharge versus time and involves voltage profile during discharge. We will start by creating a cell equivalent circuit model using TwinBuilder and its battery wizard.

We will open TwinBuilder, go to the battery wizard, and use the cell configuration tool. This will pop up a screen, allowing us to create our cell body model. We will have four parameters: SOC and temperature, which we will put in using HPPC data.

We will then generate a new component, the battery cell, ready for use. Next, we will move from the cell module to the cell to our module battery. We will go back to the battery wizard, move to the module configuration for the battery, and select our battery cell, the ECM that we have generated.

Our configuration for this particular example is 4S1P, using external loads. We will then generate a new battery module and drag it into our canvas. Now, we want to save our TwinBuilder model with an appropriate name.

The Fluent portion or the workflow is the most involved, as we will use Fluent as our energy equation solver and provide cell temperature information from Fluent back to the ECM model. This will create a two-way coupling.

The cell heat source will be defined by the module ECM, go to the energy equation, and provide feedback on the cell temperature. Our goal is to generate a transient thermal response using ANSYS Fluent.

We will open ANSYS Fluent, read in the readily available mesh file, and set up the model using the left side view. We will turn the battery model on and use conjugate heat transfer. Fluent will solve for joule heating.

We will define materials, assign them to the cell zones, and set up boundary conditions. For this particular case, we have a cold plate on one side with strong convection cooling due to water cooling. We will define our boundary condition accordingly.

We will then initialize the solution, define convergence criteria, run a steady-state analysis, and check our results. This is the midway point of task three.

We have built our module battery model inside Fluent and will now generate the thermal step response model from Fluent to be used in TwinBuilder. We will go back to our battery model, set the model parameters to zero, and use the battery ROM toolkit.

This wizard will guide us on how to build our thermal response in Fluent. We will select the LTI option, select the cells, and define our input as volume heat. Fluent will put out temperature, and we will define our transient setup.

Once we define the transient setup, we will click "run training" and generate the thermal response. Fluent will create four new files and a folder called LTI. We will then pass these files on to TwinBuilder, where they will be used as inputs for the Joule heat and volume heat.

Now that we have our LTI ROM information, we want to process it into TwinBuilder. We will go back to TwinBuilder, use the thermal model identification tool, fill in the responses, and hit "generate." This will add a new item under the project components, containing the thermal ROM.

We will define time versus voltage, tie a joule heat source as a function of time, prepare a lookup table, and refer to the data files. We will then define outputs, create rectangular plots, and define a new report. This completes the setup.

We will do a transient analysis, solve for the time steps using the reduced order model, and check the results. If the results match our expectations, we have successfully built our ROM model.

We will open our model created at the start, add details to build further and connect to our ROM model, take the twin builder model from task two, copy it, and place it into the same canvas as our ROM model.

We will then do the coupling process, giving a source for the joule heating part and tying the terminal input to the ECM. The temperature information out of the four cells from our ROM model will be tied back into the ECM model. We have now made the connections and coupled the two separate models.

We will hit the analyze button, and the run should be very fast because we are not doing any CFD. We are just marching through the 1800 seconds of time steps using the reduced order model. This concludes my presentation. Thank you for your attention.