Abstract
In this work an Adaptive Neuro-Fuzzy Inference System (ANFIS) model of the voltage of a fuel cell is developed. The inputs of this model are the fuel cell temperature, current density and the carbon monoxide concentration of the anode supply gas.
First an identification experiment which spans the expected operating range of the fuel cell is performed in a test station. The data from this experiment is then used to train ANFIS models with 2, 3, 4 and 5 membership functions.
The performance of these models is then compared and it is found that using 3 membership functions provides the best compromise between performance and fast model evaluation. This model has a mean absolute error of 0.70%. It is concluded that the developed ANFIS model is suitable for optimization of fuel cell systems and as the steady state component in larger dynamic system models.
First an identification experiment which spans the expected operating range of the fuel cell is performed in a test station. The data from this experiment is then used to train ANFIS models with 2, 3, 4 and 5 membership functions.
The performance of these models is then compared and it is found that using 3 membership functions provides the best compromise between performance and fast model evaluation. This model has a mean absolute error of 0.70%. It is concluded that the developed ANFIS model is suitable for optimization of fuel cell systems and as the steady state component in larger dynamic system models.
| Original language | English |
|---|---|
| Journal | International Journal of Hydrogen Energy |
| Volume | 40 |
| Issue number | 46 |
| Pages (from-to) | 16814–16819 |
| Number of pages | 6 |
| ISSN | 0360-3199 |
| DOIs | |
| Publication status | Published - Dec 2015 |
Keywords
- High temperature PEM fuel cells
- ANFIS modeling
- CO influence in fuel cells
- Empirical fuel cell modeling