Development and demonstration of advanced predictive and prescriptive algorithms to control industrial installation

Abstract

This paper explores the use of sophisticated, predictive AI algorithms for monitoring and optimizing industrial installations of CFB power plant. The effectiveness of the system was shown by applying it to a circulating fluidized bed (CFB 1300) power unit. A customized optimization algorithm was developed to manage the oxygen distribution within the combustion chamber. Implementing the developed control system methodology adjusted the fuel distribution, which in turn impacted the overall performance of the boiler. The approach was evaluated under different boiler operating scenarios, including simulated fuel line malfunctions. The devised methodology enables a reduction of approximately 17% in oxygen distribution imbalance within the combustion chamber when a failure was detected. Furthermore, the optimization algorithms facilitate a seamless adjustment in fuel loads, maintaining the necessary oxygen and temperature distribution at the control plane. Moreover, the capabilities of this system were demonstrated for the automatic identification of malfunctions within two crucial parts of the power unit. The initial issue pertains to a fault in the internal phase insulator of the block transformer, and the second occurrence involved the proactive identification of a membrane wall leak over 13 h before its failure.