چكيده به لاتين
The water-gas shift (WGS) reaction is a pivotal step in reducing carbon monoxide content in reformate stream (containing hydrogen, carbon monoxide, and carbon dioxide), aiming to produce pure hydrogen for fuel cells. In this research, after evaluating various synthesis methods (mechanical-chemical, co-precipitation, and impregnation), co-precipitation was selected as the optimal synthesis route. Subsequently, the catalytic activity of CuO.Al₂O₃ catalysts (with x values of 10, 20, 30, 40, and 50 wt%) was assessed in the intermediate temperature range of the WGS reaction (150-400 °C).The highest carbon monoxide conversion (64.5%) was achieved with CuO(40%)-Al₂O₃, using a steam-to-dry gas ratio of 0.3 and a gas hourly space velocity of 18000 ml/(hr.g). However, higher copper oxide loadings (50 wt%) resulted in lower CO conversion due to reduced dispersion and weaker interactions between copper oxide and alumina. Increasing the calcination temperature from 400 to 500 °C did not significantly alter the specific surface area of the selected catalyst, but CO conversion decreased from 64.5% to 55.4% at 400 °C. This phenomenon can be attributed to thermal sintering of copper species and weakened interactions between copper oxide and alumina, as confirmed by XRD, TPR, and FESEM analyses.
In the final phase, to achieve optimal values for the influential synthesis parameters (pH, temperature, and aging time), the Box-Behnken design (BBD) under the Response Surface Methodology (RSM) was employed. The impact of each variable was evaluated using analysis of variance (ANOVA). The convergence values of R², adjusted R² (R²adj), predicted R² (R²pred), and signal-to-noise ratio were 0.9801, 0.9443, 0.7481, and 22.6708, respectively, indicating the adequacy of the model in fitting the experimental data and its sufficient accuracy in data analysis. The optimal values proposed by the model for aging temperature (°C), aging time (hr), and pH were 60.6, 10, and 4, respectively, resulting in a 51.9% CO conversion at 300 °C. The comparison between the predicted (51.9%) and experimental (52.8%) values showed a model error of approximately 1.5%.
