Optimization of pressure swing adsorption in a three-layered bed for hydrogen purification using machine learning...

Area of Science:

• Chemical Engineering
• Materials Science
• Computational Chemistry

Background:

• Hydrogen purification is crucial for various industrial applications, including steam-methane reforming (SMR).
• Traditional adsorbents face limitations in achieving high purity and recovery rates for hydrogen from SMR off-gas.
• Optimizing pressure swing adsorption (PSA) cycles is essential for efficient hydrogen production.

Purpose of the Study:

• To develop and evaluate a novel three-layered adsorbent bed for hydrogen purification from SMR off-gas.
• To investigate the performance of different adsorbent combinations using PSA cycle modeling.
• To optimize PSA operational parameters using machine learning for enhanced hydrogen production.

Main Methods:

• A three-layered adsorbent bed was designed using UTSA-16, activated carbon, and zeolite 5A.
• Adsorption isotherms and breakthrough curves were simulated and validated for SMR off-gas.
• Pressure swing adsorption (PSA) cycle models were developed to analyze purification performance.
• A genetic algorithm (GA) was used to optimize a backpropagation neural network (BPNN-GA) for predicting hydrogen production.

Main Results:

• The three-layered bed with UTSA-16 achieved a hydrogen purity of 99.99%, with 62.08% recovery and 7.2209 mol/(kg·h) productivity.
• The BPNN-GA model demonstrated superior prediction accuracy (error of 0.0173) compared to the BPNN model (error of 0.0513).
• The BPNN-GA model showed a high correlation coefficient (R ≈ 1) with Aspen model targets, indicating excellent performance.

Conclusions:

• The novel three-layered adsorbent bed effectively purifies hydrogen from SMR off-gas.
• The BPNN-GA model provides a precise and accurate method for optimizing PSA operational parameters.
• This approach enables efficient determination of optimal conditions for maximizing hydrogen production and purity.