Sustainable design of solar chimney power plants: A hybrid neural network approach for thermo-economic optimization

The optimal design of geometrical features in solar chimney power plants enhances performance but often increases costs, creating a need for economical design approaches. This study proposes an artificial intelligence-driven multi-objective optimization framework for thermoeconomic solar chimney power plant design, integrating numerical simulations with neural networks and genetic algorithms. The investigation considered a high-dimensional input feature space consisting of collector inlet height, collector diameter, chimney diameter, chimney height, and solar radiation, modeling their effects on system performance to develop high-fidelity neural networks for predicting actual power, overall efficiency, and total cost targeting Manzanares plant conditions. Numerical simulations using finite volume methods were conducted with ANSYS, generating comprehensive datasets based on 136 sets of geometrical parameters. The developed neural networks are deployed as objective functions in a multi-objective genetic algorithm framework for performing Pareto optimality that simultaneously maximizes power and efficiency while minimizing cost. The optimization study yielded a remarkable improvement in both power and efficiency, with power output increasing by a factor of 3.82 and efficiency rising by 4 times, all while maintaining almost same cost as the reference plant. Further analysis showed that power generation was 3.65 times higher, and efficiency 3.55 times greater, at just 87 % of the cost of the reference plant. Notably, a 10 % higher investment resulted in a substantial gain—power was enhanced by 4.51 times and efficiency improved by 5.73 times. These gains were achieved through a strategic design approach that involved enlarging the collector and chimney diameters, while reducing the chimney height. This approach enables rapid exploration of complex design spaces that would be computationally prohibitive using traditional computational fluid dynamics-based optimization methods and can be extended for optimizing any solar chimney-based energy system.