Forecasting the Mass Activity of Platinum Anode Catalysts on Various Carbon Nanostructures in Direct Methanol Fuel Cells Using Machine Learning
Anode catalyst loading in direct methanol fuel cells (DMFC) is extremely high at around 4.5 mg _PtRu cm ^−2 , which increases cost and inhibits commercialization. Several carbon nanostructures, such as carbon nanotubes, graphene, mesoporous carbon, and carbon quantum dots, are used to support platin...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2024-01-01
|
| Series: | ECS Advances |
| Subjects: | |
| Online Access: | https://doi.org/10.1149/2754-2734/ada04a |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Anode catalyst loading in direct methanol fuel cells (DMFC) is extremely high at around 4.5 mg _PtRu cm ^−2 , which increases cost and inhibits commercialization. Several carbon nanostructures, such as carbon nanotubes, graphene, mesoporous carbon, and carbon quantum dots, are used to support platinum as well as platinum with other metals and metal oxides to reduce the platinum content of catalysts. Optimizing the catalyst composition for DMFC requires extensive trial and error experiments due to the complex electrochemical and thermodynamic processes, which demands considerable time. We present here machine learning-aided models that correlate the composition of platinum-based catalysts on different carbon nanostructures with the mass activity of DMFC. Various machine learning techniques are employed to predict the mass activity of platinum-based catalysts using data from published literature. These models demonstrate a good level of predictive accuracy (R ^2 > 0.85) with the available datasets and show that even basic models can provide reliable forecasts. The SHapley Additive Explanations (SHAP) summary plot reveals that graphene’s weight fraction is the most significant feature among all carbon nanostructures, followed by the weight fractions of cobalt and platinum. Hence, machine learning has demonstrated significant effectiveness in predicting platinum’s mass activity based on catalyst composition and process parameters. |
|---|---|
| ISSN: | 2754-2734 |