Sign in Subscribe
Artificial intelligence

Artificial Intelligence-based Surrogate Modeling for Hydrogen Liquefaction Process Optimization

Strategic Insights

— 1 min read

Researchers at Dalian University of Technology, in collaboration with King Saud University, have made significant breakthroughs in optimizing the hydrogen liquefaction process using artificial intelligence-based surrogate modeling. The study, published in the International Journal of Hydrogen Energy, aims to tackle the complex challenge of hydrogen liquefaction optimization, which has been an open issue due to convergence issues and multiple design variables. The researchers developed an artificial neural network-based surrogate model, which was applied to the hydrogen liquefaction process, resulting in a significant reduction in computational effort and improved accuracy.

Key Takeaways:

  • The researchers developed an artificial neural network-based surrogate model to optimize the hydrogen liquefaction process, which reduced the time needed for particle swarm optimization by more than 99.99%.
  • The percentage error of prediction of the ANN-based surrogate model compared to particle swarm optimization (PSO) was 4% for the minimum internal approach temperature and 0.04% for specific energy consumption.
  • The optimization of the proposed model significantly reduced the computational effort and improved accuracy in predicting the hydrogen liquefaction process.
  • The study used particle swarm optimization (PSO) as a benchmark to compare the performance of the ANN-based surrogate model.
  • The researchers cite that the 'shallow' neural network model, with one hidden layer used in the present study, can be used to analyze the whole liquefaction process as an extension of the current work.
  • The study found that the optimization of the proposed model can be applied to other complex processes in the field of chemical engineering.

Statistics:

  • The ANN-based surrogate model reduced the time needed for PSO optimization of the existing model by more than 99.99%.
  • The percentage error of prediction of the ANN-based surrogate model compared to PSO was 4% for the minimum internal approach temperature and 0.04% for specific energy consumption.
  • Over 99.99% reduction in computational effort compared to the existing particle swarm optimization model.
  • 4% error in prediction of minimum internal approach temperature using the ANN-based surrogate model.
  • 0.04% error in prediction of specific energy consumption using the ANN-based surrogate model.

Sources:

  • Artificial Intelligence-based Surrogate Modeling for Computational Cost-effective Optimization of Hydrogen Liquefaction Process. International Journal of Hydrogen Energy, 2025;137:819-829.
  • Dalian University of Technology, Ningbo Inst, Ningbo 315200, Zhejiang, People's Republic of China.
  • Bo Zhang, Ali Rehman, Amjad Riaz, Kinza Qadeer, Seongwoong Min, Moonyong Lee, Ashfaq Ahmad, Fatima Zakir, and Mohamed A. Ismail.

Read more