Predicting combined-cycle natural gas power plant emissions by using artificial neural networks

Gaseous emission from a chimney is recognized as one of the sources of pollution produced from a typical power plant. Among the pollutants of concern from the chimney of the power plant are NO_x , SO₂ and CO. Commonly, the application of continuous emission monitoring systems (CEMS) is used to measure the emissions directly. It is possible however, to predict stack gases from the combustion chamber indirectly so that a build up of a database on related input and output of various parameters can be generated. From this relationship, the critical points of various parameters can be optimized to limit the pollution from the chimney. An artificial neural networks (ANN) based on a feedforward backpropagation model is selected for this objective. The limited data taken from Lumut Power Plant are used to train the neural network. This prediction from neural network based on training agrees well with the data taken from CEMS