Machine Learning-driven Optimization of a Multigeneration Solar Power Plant

Researchers at Kharazmi University in Tehran, Iran, have made significant progress in optimizing a multigeneration solar power plant using machine learning techniques. Their study, published in the International Journal of Hydrogen Energy, aimed to improve the plant's efficiency and productivity while reducing costs and environmental impact. The research team, led by Mostafa Esmaeili, utilized a genetic algorithm combined with the CatBoost algorithm to optimize three key performance metrics: hydrogen generation rate, total cost rate, and overall exergy efficiency.

Key Takeaways:

• The solar power plant consists of nine distinct subsystems, including a solar cycle, a conventional Rankine cycle, two organic Rankine cycles, and units for hydrogen, ammonia, and methane production, as well as a reverse osmosis desalination system and a desiccant-based cooling cycle.
• The plant achieves a net electrical generating capacity of 5742 kW and generates 224.2 kW of cooling energy, with a hydrogen production rate of 131.2 kg/h.
• The system's overall energy and exergy efficiencies were quantified at 53.5% and 50.7%, respectively, with an overall exergy destruction rate of 18404 kW.
• The levelized costs for electricity and hydrogen generation were established at 0.025 USD/kWh and 1.136 USD/kg, respectively.
• Optimization using machine learning techniques led to a 5.7 USD/h drop in total cost rate and a 0.8 kg/h decrease in hydrogen production rate, while increasing exergy efficiency by 1% to 51.7%.
• The study demonstrates the potential of machine learning-driven optimization for improving the performance of multigeneration solar power plants.

Statistics:

• The solar power plant produces 5742 kW of electricity and 224.2 kW of cooling energy.
• Hydrogen production rate: 131.2 kg/h.
• Overall exergy destruction rate: 18404 kW.
• Total cost rate: 158.5 USD/h.
• Levelized costs for electricity: 0.025 USD/kWh.
• Levelized costs for hydrogen: 1.136 USD/kg.

Sources:

• Esmaeili, M., Moradi, A., Vojdani, M. M., Karami, M., & Rosen, M. A. (2025). Machine Learning-driven Optimization of a Multigeneration Solar Power Plant: a 4e Framework for Hydrogen and Energy Generations. International Journal of Hydrogen Energy, 147.
• NewsRx. Data on Machine Learning Discussed by Researchers at Kharazmi University (Machine Learning-driven Optimization of a Multigeneration Solar Power Plant: a 4e Framework for Hydrogen and Energy Generations). Energy Weekly News. August 1, 2025; p 57.