Machine Learning-Based Optimization Boosts Efficiency of Molten-Salt-Driven Kalina Cycle

Strategic Insights -

Research conducted at the University of Tabriz, Iran, has demonstrated the potential of machine learning-based optimization to improve the efficiency of molten-salt-driven Kalina cycles. The study analyzed a system used for renewable-based power generation and identified areas for improvement. By applying a machine learning-based multi-objective optimization framework, the researchers were able to enhance system performance, reducing exergy destruction and increasing economic attractiveness.

Key Takeaways:

  • The research analyzed a molten-salt-driven Kalina cycle under full-load and part-load conditions, achieving energy and exergy efficiencies of 30.2% and 36.72% respectively in the baseline full-load case.
  • The system's payback period at full load was 7.70 years, while at part-load conditions of 0.9, 0.8, 0.7, 0.6, and 0.5, the payback period progressively increased to 7.83, 8.10, 8.49, 8.99, and 9.99 years respectively.
  • Optimization using machine learning-based multi-objective optimization improved energy and exergy efficiencies by 21.6% and 18.3% respectively at full load, while reducing total exergy destruction by 20.4%.
  • From an economic perspective, the payback period decreased to 6.36 years, with annual income and net present value increasing by 1.27% and 20.5% respectively.
  • The TES heat exchanger had the highest exergy destruction rate under conventional exergy analysis, whereas advanced exergy analysis identified the turbine as the primary optimization target.
  • A machine learning-based multi-objective optimization framework was applied to enhance system performance and improve efficiency.

Statistics:

  • Energy and exergy efficiencies of 30.2% and 36.72% respectively in the baseline full-load case.
  • Payback period at full load: 7.70 years.
  • Payback period at part-load conditions: 7.83, 8.10, 8.49, 8.99, and 9.99 years respectively.
  • Optimization improved energy and exergy efficiencies by 21.6% and 18.3% respectively at full load.
  • Reduction in total exergy destruction: 20.4% at full load.
  • Payback period after optimization: 6.36 years.
  • Increase in annual income and net present value: 1.27% and 20.5% respectively.

Sources:

  • Technoeconomic and advanced exergy analysis of a molten-salt-driven Kalina cycle under full-load and part-load conditions: A machine learning-based multi-objective optimization. Energy Conversion and Management: X, 2025, 28():101300.
  • Journal of Engineering. October 20, 2025; p 3634.
  • University of Tabriz, Department of Mechanical Engineering.