Surrogate Model Training of a Battery Rate Capability Model
Application ID: 119181
This app demonstrates the usage of a surrogate model function for predicting the rate capability of an NMC111/graphite battery cell. The rate capability is shown in a Ragone plot. The surrogate function, a Deep Neural Network, has been fitted to a subset of the possible input data values. Three input data values can be set: the thickness of the negative electrode, the active material volume fraction of the negative electrode, and the active material volume fraction of the positive electrode. The low computational cost of evaluating the surrogate function allows sliders to be used to interactively adjust the input values and predict the Ragone plot for any combination of input values. Once a promising combination of values has been identified, the actual physical Li-ion battery model can be computed for those input values, to verify the predictions of the surrogate model. In addition, the then computed physical data can be used to further improve the surrogate model.
This application example illustrates applications of this type that would nominally be built using the following products:
however, additional products may be required to completely define and model it. Furthermore, this example may also be defined and modeled using components from the following product combinations:
The combination of COMSOL® products required to model your application depends on several factors and may include boundary conditions, material properties, physics interfaces, and part libraries. Particular functionality may be common to several products. To determine the right combination of products for your modeling needs, review the Tabella delle Funzionalità and make use of a free evaluation license. The COMSOL Sales and Support teams are available for answering any questions you may have regarding this.