%0 Journal Article %A Skreiberg, 0. %A Hustad, J. E. %A Garnaes, E. I. %A Reinermann, P. %A Bjorge, T. %D 2007 %I Begell House %N 3 %P 221-237 %R 10.1615/InterJEnerCleanEnv.v8.i3.30 %T COMPARING EMPIRICAL, STATISTICAL, AND NEURAL NETWORK MODELS CALCULATING NOx EMISSIONS FROM GAS TURBINES %U https://www.dl.begellhouse.com/journals/6d18a859536a7b02,1fdbfb0345386dda,5f0f5f845eb2e1a6.html %V 8 %X Measurements of NOx levels and process data have been performed on two offshore gas turbine installations in Norway, a standard aeroderivative system and a dry low-emission system. The data have been used to develop statistical and neural network NOx emission models, and comparisons with existing empirical models, i.e., so-called physical or first principles models, have been made. Model comparison results show that the neural network models perform best, but also, some statistical models are able to reproduce the measured emission levels quite well. Empirical models are not able to reproduce the emission levels from the two gas turbines satisfactorily. %8 2008-05-03