Enhancing green hydrogen forecasting with a spatio-temporal graph convolutional network optimized by the Ninja...

Area of Science:

• Environmental Science
• Chemical Engineering
• Computer Science

Background:

• Sustainable infrastructure development is crucial for combating climate change.
• Green hydrogen, produced via renewable electrolysis, is a key component of this shift.
• Accurate forecasting of green hydrogen production is challenging due to variable environmental and system factors.

Purpose of the Study:

• To develop an improved system for forecasting green hydrogen production.
• To address the variability of temporal and spatial factors influencing production.
• To enhance the accuracy and efficiency of renewable energy modeling.

Main Methods:

• Utilized a Spatio-Temporal Graph Convolutional Network (STGCN).
• Introduced a novel algorithm, the Ninja Optimization Algorithm (NiOA), for optimization.
• Employed binary NiOA for feature selection and continuous NiOA for model architecture and variable optimization.

Main Results:

• The STGCN model achieved an R² of 0.8769 and MSE of 0.00375.
• The STGCN integrated with NiOA demonstrated significantly improved results with an R² of 0.9815.
• The STGCN-NiOA model achieved a substantially lower MSE, indicating enhanced prediction accuracy.

Conclusions:

• The integration of STGCN with NiOA offers a robust framework for accurate green hydrogen production forecasting.
• Adaptive metaheuristics, like NiOA, show significant promise for improving renewable energy system modeling.
• This strategy provides a dependable approach for modeling renewable energy systems and bolstering green hydrogen initiatives.