Abstract
Metal-supported solid oxide fuel cell (MS-SOFC) is an emerging technology suitable for vehicular applications. Utilizing the off-gas from MS-SOFC to realize methane steam reforming is an attractive approach to avoid the challenges of hydrogen storage and transportation. However, the operation temperature of such a pre-reformer will be lower than conventional high-temperature reformer. How to optimize the geometric parameters of such a pre-reformer needs to be clarified. In this study, a data-driven multi-objective optimization method is proposed. The critical parameters of a methane steam pre-reformer are determined using the principal component analysis based on Taguchi method and a data-driven multi-objective optimization based on artificial neural network (ANN) and NSGA-II genetic algorithm. A heterogeneous mathematical model is established based on the Langmuir-Hinshelwood method. The effects of critical geometric parameters of the pre-reformer on the performance under intermediate-temperature conditions are studied using Taguchi orthogonal experimental design and analysis of variance (ANOVA). A surrogate ANN model for predicting the performance of the methane pre-reformer is setup. Then a multi-objective genetic algorithm optimization is performed based on the surrogate model to maximize the methane conversion rate and minimize the total cost. The results indicate that the significances of the four parameters are different. The number of tubes and tube length demonstrate the most significant effects on the methane conversion rate、hydrogen production yield and total heat transfer cost, whereas the bed porosity and tube inner diameter exhibit the lowest degree of influences. Using a surrogate ANN model during the multi-objective optimization process can alleviate the computation load significantly while maintains a high prediction precision. The methane conversion rate increases by 9.5% and the total cost declines by 18.4% compared to the original design.
Original language | English |
---|---|
Article number | 134114 |
Journal | Fuel |
Volume | 385 |
DOIs | |
Publication status | Published - 1 Apr 2025 |
Keywords
- Data-driven method
- Heterogeneous model
- NSGA-II genetic algorithm
- Steam methane reforming
- Taguchi method