AN IMAGING NEURAL-NETWORK COMBUSTION CONTROL-SYSTEM FOR UTILITY BOILER APPLICATIONS

Advances in sensor and control system strategies for utility boiler combustion systems are required in order to meet increasingly stringent standards on emission. In many instances, this may involve selective, single-burner monitoring and control. In this study, a novel control system was developed and demonstrated for this purpose. Using an electronic camera to image the time-resolved, chemically specific emission patterns in the flame and a neural network to process and analyze the resultant images, a model utility load controller was demonstrated in a laboratory liquid-fueled spray flame facility. Characterization tests of the system's dynamic stability were used to optimize the controller's image processing, neural network, and control algorithms. A series of closed-loop control tests characterized the system's response to a variety of simulated load excursions. With the purely spatial information from the imaging sensor, the control system successfully interpreted the images and guided the burner through the imposed excursions. Stable closed-loop control was demonstrated.