Elman neural network in the soft sensor modelling for the unburned carbon in fly ash from utility boilers

Abstract

Unburned carbon in fly ash is an important parameter affecting combustion efficiency of coal-fired boiler. In view of the deficiency of feed-forward neural network soft sensor modeling on unburned carbon in fly ash from the power plant, in this paper, we make use of recurrent Elman neural network to realize dynamic modeling of the boiler combustion process. A set of operating data from a 300MW power plant boiler is used here to train and validate the soft sensor model. Then this is compared with the results of BP network. The results after comparing show that Elman network can better achieve soft sensor modeling for unburned carbon in fly ash.