Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation
Proton exchange membrane water electrolysis is an effective method for producing hydrogen required for advancing the transition to greener sustainable energy. However, the complex dynamics associated with the process requires predictive tools for large-scale deployment. Despite advancements in machi...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-07-01
|
| Series: | Cleaner Engineering and Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2666790825001636 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850106706535645184 |
|---|---|
| author | Okorie Ekwe Agwu Saad Alatefi Ahmad Alkouh |
| author_facet | Okorie Ekwe Agwu Saad Alatefi Ahmad Alkouh |
| author_sort | Okorie Ekwe Agwu |
| collection | DOAJ |
| description | Proton exchange membrane water electrolysis is an effective method for producing hydrogen required for advancing the transition to greener sustainable energy. However, the complex dynamics associated with the process requires predictive tools for large-scale deployment. Despite advancements in machine learning-based models, previous studies often lack explainability, diminishing user trust in their deployment. This study addresses this deficiency by developing an accurate and explainable hydrogen yield rate model using Bayesian regularized neural network. The dataset utilized comprises nine input variables and 231 data points for each variable. The results from the model development show that the model demonstrates reasonable precision, with a mean square error of 0.0588, root mean square error of 0.24, mean absolute error of 0.1057 and a coefficient of determination of 0.95. The connection weights algorithm applied to the model enhances its explainability by illustrating the relative contributions of each input variable and their impacts on hydrogen yield. It was found that stack voltage and water pressure have the most significant impacts on the electrolysis process accounting for 23 % and 17.6 % respectively while the lower explosive limit had the least impact with a 4 % importance factor. The model's applicability domain was established using the Williams plot, while trend analyses indicated that the model aligns with the physical trends associated with water electrolysis phenomena. Overall, the model can be used in two modes: online, by integrating it into software programs, and offline, by simply entering parameter values into the explicit model without having to run long lines of code. |
| format | Article |
| id | doaj-art-3f3c4fa182504b028f11dc3904fafe8d |
| institution | OA Journals |
| issn | 2666-7908 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Cleaner Engineering and Technology |
| spelling | doaj-art-3f3c4fa182504b028f11dc3904fafe8d2025-08-20T02:38:46ZengElsevierCleaner Engineering and Technology2666-79082025-07-012710104010.1016/j.clet.2025.101040Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimationOkorie Ekwe Agwu0Saad Alatefi1Ahmad Alkouh2Petroleum Engineering Department, University Teknologi PETRONAS, 32610, Seri Iskandar, Perak Darul Ridzuan, Malaysia; Center of Reservoir Dynamics (CORED), Institute of Sustainable Energy, Universiti Teknologi, PETRONAS, 32610, Seri Iskandar, Perak Darul Ridzuan, Malaysia; Corresponding author. Petroleum Engineering Department, University Teknologi PETRONAS, 32610, Seri Iskandar, Perak Darul Ridzuan, Malaysia.Department of Petroleum Engineering Technology, College of Technological Studies, PAAET, Kuwait City, 70654, KuwaitDepartment of Petroleum Engineering Technology, College of Technological Studies, PAAET, Kuwait City, 70654, KuwaitProton exchange membrane water electrolysis is an effective method for producing hydrogen required for advancing the transition to greener sustainable energy. However, the complex dynamics associated with the process requires predictive tools for large-scale deployment. Despite advancements in machine learning-based models, previous studies often lack explainability, diminishing user trust in their deployment. This study addresses this deficiency by developing an accurate and explainable hydrogen yield rate model using Bayesian regularized neural network. The dataset utilized comprises nine input variables and 231 data points for each variable. The results from the model development show that the model demonstrates reasonable precision, with a mean square error of 0.0588, root mean square error of 0.24, mean absolute error of 0.1057 and a coefficient of determination of 0.95. The connection weights algorithm applied to the model enhances its explainability by illustrating the relative contributions of each input variable and their impacts on hydrogen yield. It was found that stack voltage and water pressure have the most significant impacts on the electrolysis process accounting for 23 % and 17.6 % respectively while the lower explosive limit had the least impact with a 4 % importance factor. The model's applicability domain was established using the Williams plot, while trend analyses indicated that the model aligns with the physical trends associated with water electrolysis phenomena. Overall, the model can be used in two modes: online, by integrating it into software programs, and offline, by simply entering parameter values into the explicit model without having to run long lines of code.http://www.sciencedirect.com/science/article/pii/S2666790825001636HydrogenNeural networksElectrolysisProton exchange membraneExplainable AI |
| spellingShingle | Okorie Ekwe Agwu Saad Alatefi Ahmad Alkouh Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation Cleaner Engineering and Technology Hydrogen Neural networks Electrolysis Proton exchange membrane Explainable AI |
| title | Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation |
| title_full | Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation |
| title_fullStr | Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation |
| title_full_unstemmed | Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation |
| title_short | Modelling the future of cleaner energy: Explainable artificial intelligence model for green hydrogen production rate estimation |
| title_sort | modelling the future of cleaner energy explainable artificial intelligence model for green hydrogen production rate estimation |
| topic | Hydrogen Neural networks Electrolysis Proton exchange membrane Explainable AI |
| url | http://www.sciencedirect.com/science/article/pii/S2666790825001636 |
| work_keys_str_mv | AT okorieekweagwu modellingthefutureofcleanerenergyexplainableartificialintelligencemodelforgreenhydrogenproductionrateestimation AT saadalatefi modellingthefutureofcleanerenergyexplainableartificialintelligencemodelforgreenhydrogenproductionrateestimation AT ahmadalkouh modellingthefutureofcleanerenergyexplainableartificialintelligencemodelforgreenhydrogenproductionrateestimation |