Soft computing analysis of a compressed air energy storage and SOFC system via different artificial neural network architecture and tri-objective grey wolf optimization

Highlights: • A detailed 4E modeling and tri-objective optimization of a hybrid energy system is presented. • To find optimal system design and performance, ANN and MOGWO are employed. • ANN is evaluated for the best performance by different architects and training algorithms. • According to the results, the LM training algorithm is more efficient. • The ERTE and total cost rate are obtained 45.7% and 34.2 $/h, respectively. In the present study, a novel combined system consisting of solid oxide fuel cell (SOFC), organic Rankine cycle (ORC), and compressed air energy storage (CAES) is proposed, investigated, and optimized. The SOFC and CAES models are validated individually to ensure the accuracy of the results. Here, the grey wolf multi-objective optimization (MOGWO) approach is applied to find the optimal system design and performance. For this, a trained neural network is provided to the MOGWO algorithm as a fitted function, and multi-objective optimization is carried out on it. The most significant benefit of the suggested method is time-saving. The proposed system's thermodynamic performance is investigated from the energy, exergy, economic, and environmental (4E) points of view at three periods, including full-time, charging, and discharging periods. The results indicate that the Levenberg-Marquardt training algorithm has the best performance among all of the algorithms. The value of exergetic round trip efficiency (ERTE), total cost rate, and CO₂ emission at the best optimum point are obtained as 45.7%, 34.2 $/h, and 0.22 kg/kWh, respectively.