Simulating Battery Systems with Surrogate Models

When simulating a battery system, using a surrogate model instead of a full-fledged finite element model can offer both speed and accuracy. Investing in data generation and surrogate model training is a great approach for teams looking to optimize design efficiency.

Learn the key considerations of modeling batteries and get an overview of surrogate models and their various use cases in this archived webinar. In the presentation, we explore the workflow of training surrogate models and describe the following examples of functions included in the COMSOL Multiphysics^® software:

• Deep neural network (DNN)
• Polynomial chaos expansion (PCE)
• Gaussian process (GP)

After a high-level overview of neural network architecture optimization, the webinar concludes with a demonstration in which we compute Li-ion battery voltage behavior in COMSOL Multiphysics^® and use a surrogate model for quicker evaluations.

Suggested Resources

• Gain a deeper understanding of surrogate model usage
• Check out these resources on training surrogate models for faster simulation apps
• Try building a deep neural network surrogate model