Artificial intelligence for control and optimization of boilers’ performance and emissions: A review

Burning fossil fuels is a major concern for global warming control. In Saudi Arabia, steam power plants that relay on boilers to produce the steam accounted for around 50% of the total electricity generation capacity in 2021, in addition to 15% capacity from co-generation power plants. Fossil-based energy production produces harmful emissions, mostly NOx. The case of combustion efficiency and NOx emissions is complicated and cannot be addressed by simple theoretical model. This paper investigates the existing literature exploring data driven approaches via Artificial Intelligence (AI) and machine learning (ML) in a variety of applications, including mainly boiler optimization for efficient and clean energy production. The study discusses the importance of energy intelligence for industries in supporting decision making and prediction considering supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. Expert systems, artificial neural networks (ANN), genetic algorithms (GA), fuzzy logic (FL), and different hybrid systems, which incorporate two or more techniques, are all included in AI systems. A condensed review of some recent literature is presented on the applications of support vector machines, ANN, GA, FL, and hybrid systems to combustion systems for emphasizing boilers' performance and emissions. An overview of boiler dynamics and control are presented and the use of AI for trade-off between boiler's efficiency and NOx emissions is discussed, with special focus being made on hydrogen combustion. The reported findings in this study demonstrate the promise of AI in control of emissions and optimization of performance in boilers with the least amount of construction and operating expenditures. The complexity and a number of restrictions present in emissions control systems, however, may reduce the likelihood of success. The specific perspectives are found to be data availability, data reliability, validity of methodology, and system complexity. It is concluded that wide variety of AI approaches have been used in a diversity of disciplines for modelling, predicting performance and controlling emissions of combustion processes. It is indicated that data representing the history and performance of real systems along with the proper selection of the AI model are crucial for setting up such systems. Based on this study, it can be stated that AI models are great tools towards better performance of boilers at minimum emissions.