Estimation of NOx emissions in thermal power plants using neural networks

In this paper a neural network-based strategy is proposed for the estimation of the NOx emissions in thermal powerplants, fed with both oil and methane fuel. A detailed analysis based on a three-dimensional simulated of the combustion chamber has pointed out the local nature of the NOx generation process, which takes place mainly in the burners' zones. This fact has been suitably exploited in developing a compound estimation procedure, which makes use of the trained neural network together with a classical one-dimensional model of the chamber. Two different learning procedures have been investigated, bath based on the external inputs to the burners and a suitable mean cell temperature, while using local and global NOx flow rates as learning signals, respectively. The approach has been assessed with respect to both simulated and experimental data.