The Neural Network Controller for the Dry Low Emission Combustor of Gas-Turbine Power Plants

Abstract

The study is devoted to the development and testing of the harmful substances' emission controller for a gas turbine power plant with a capacity of 16 MW (GTP-16) based on the built-in neural network mathematical model of a dry low emission (DLE) combustor. The developed algorithms for the neural network controller of the emission of nitrogen oxides, carbon monoxide and pressure pulsations in the DLE-combustor flame tubes are implemented in MATLAB R2018b Simulink and integrated into the GTP-16 automatic control system (ACS) on the hardware/software platform PXI NI. The efficiency of the emissions controller was checked during bench tests on the GTP-16 simulator, with the DLE-combustor neural network model performing the functions of a virtual emissions sensor. The errors in the estimation of emissions and pressure pulsations that meet the accepted requirements are determined. The normal error distribution of the developed neural network model of the combustion chamber is proved. The resulting emission control quality corresponds to the desired one. The conclusion about the possibility and prospects of using neural networks for the development of an adaptive emission control system for DLE-combustors of the gas turbine power plants was made.