On the utilization of artificial intelligence for studying and multi-objective optimizing a compressed air energy storage integrated energy system

• Published in: Journal of energy storage 2024-04, Vol.84, p.110839, Article 110839
• Main Authors: Yun, Pengyu, Wu, Huiping, Alsenani, Theyab R., Bouzgarrou, Souhail Mohammed, Alkhalaf, Salem, Alturise, Fahad, Almujibah, Hamad
• Format: Article
• Language: English
• Subjects: Artificial intelligence; Energy storage; Energy utilization; Machine learning; Optimization; Regression analysis
• ISSN: 2352-152X; 2352-1538
• DOI: 10.1016/j.est.2024.110839

The field of utilizing machine learning algorithms and artificial intelligence for studying and optimizing compressed air energy storage integrated energy systems with solid oxide fuel cells is of utmost importance. Further studies in this field are of great significance and should be pursued to unlock the full potential of these integrated energy systems. This study proposes an integrated energy system combining compressed air energy storage (CAES) and solid oxide fuel cell (SOFC) to generate compressed air, power, and heating. The SOFC generates electricity, part of which powers the CAES system for compressed air production. Flue gases from the SOFC activate domestic heat recovery, resulting in heating air capacity. Machine learning techniques predict system performance and optimize it for best results. Machine learning algorithms developed using regression analysis have high accuracy with R-squared values >98 % for all outputs and they perform well to predict the new observation with predicted R-squared values mostly >99 %. Also, its act in optimizing the system performance is significant. By choosing a utilization factor of 0.795 and a current density of 4300 A/m2, the energy efficiency can reach 63.4 % while the exergy efficiency can reach 32.5 %. These values align with the predicted ranges given by the machine learning models. •An integrated energy system combining CAES and SOFC is proposed.•System generates compressed air for energy storage as well as power, and heating.•Machine learning algorithms are established to predict and optimize performance.•Models have high accuracy with R-squared values >98 %.•Models perform well to predict the new observation with predicted R-squared>99 %.