Monitoring of Air Emissions Using a Multivariable Model and Process History

Abstract

Energy producers are facing a challenging task in trying to monitor the energy conversion processes due to their complexity, nonlinear dynamics, and a large number of affecting factors. There are several methods available which can deal with multidimensionality and which could be used in industrial monitoring systems, but it seems that the methods used by the industry are not necessarily fully compatible with the requirements of modern energy production. A system capable of handling the large amount of available measurement data, extracting the essential pieces of information, and presenting the condition and evolution of the process in an easily understandable manner could be advantageous when monitoring and analyzing energy processes. In this paper, we demonstrate the use of self-organizing maps (SOM) and existing measurements in monitoring air emissions in an industrial circulating fluidized bed (CFB) boiler. The chosen approach should be able to take the multivariate and dynamical characteristics of the process into account, and therefore provide a flexible and efficient platform for monitoring.