Current predictive emission monitoring (PEM) techniques are briefly reviewed and the concept for a general predictive model was favorably evaluated. Utilizing the commercial process simulation software ASPEN PLUS®, a one dimensional model based on fundamental principles of gas turbine thermodynamics and combustion processes was constructed. Employing a set of 22 reactions including the Zeldovich mechanism, the model predicted for thermal NOx formation. It accounted for combustor geometry, dilution air injection along the combustor annulus, convective heat transfer across the liner, flame length, and full-load inlet flows. The combustor was subdivided into slices, each of which was modeled by a plug flow reactor, giving insight into profiles of NOx formation, species concentration and temperature along the combustor’s length, as well as quantifying the residence time in the combustor. The simulation predicted the levels of NOx for a particular gas turbine combustor and determined the effects of various parameters, such as flame length, hydrocarbon conversion ratio and recycle zones.

This paper presents a successful demonstration of application of neural networks to perform various data mining functions on an RB211 gas-turbine-driven compressor station. Radial basis function networks were optimized and were capable of performing the following functions: (a) backup of critical parameters, (b) detection of sensor faults, (c) prediction of complete engine operating health with few variables, and (d) estimation of parameters that cannot be measured. A Kohonen SOM technique has also been applied to recognize the correctness and validity of any data once the network is trained on a good set of data. This was achieved by examining the activation levels of the winning unit on the output layer of the network. Additionally, it would also be possible to determine the suspicious, faulty or corrupted parameter(s) in the cases which are not recognized by the network by simply examining the activation levels of the input neurons.

This paper presents a successful demonstration of application of Neural networks to perform various data mining functions on an RB211 gas turbine driven compressor station. Radial Basis Function networks were optimized and were capable of performing the following functions: a) Backup of critical parameters, b) Detection of sensor faults, c) Prediction of complete engine operating health with few variables, and d) Estimation of parameters that cannot be measured. A Kohonen SOM technique has also been applied to recognize the correctness and validity of any data once the network is trained on a good set of data. This was achieved by examining the activation levels of the winning unit on the output layer of the network. Additionally, it would also be possible to determine the suspicious, faulty or corrupted parameter(s) in the cases which are not recognized by the network by simply examining the activation levels of the input neurons.

A one dimensional model based on fundamental principles of gas turbine thermodynamics and combustion processes was constructed to quantify the principle of exhaust gas recirculation (EGR) for NOx reduction. The model utilizes the commercial process simulation software ASPEN PLUS®. Employing a set of 8 reactions including the Zeldovich mechanism, the model predicted thermal NOx formation as function of amount of recirculation and the degree of recirculate cooling. Results show that addition of sufficient quantities of uncooled recirculate to the inlet air (i.e. EGR>∼4%) could significantly decrease NOx emissions but at a cost of lower thermal efficiency and specific work. Cooling the recirculate also reduced NOx at lower quantities of recirculation. This has also the benefit of decreasing losses in the thermal efficiency and in the specific work output. Comparison of a ‘rubber’ and ‘non-rubber’ gas turbine confirmed that residence time is one important factor in NOx formation.

A thermodynamic, environmental and economic assessment of an exhaust gas recirculation (EGR) system for NOx reduction has been carried out on an RB211 gas turbine based compressor station. The configured system was evaluated using a commercial process simulation software ASPEN PLUS® for the EGR process, along with a one dimensional model for the prediction of NOx. The assessment was focused on a realistic system of 20% gas recirculation cooled 300 °C with an aerial cooler. Detailed economic analysis based on present value cost per unit mechanical energy (kWh), showed that there is no economic advantage in implementing an EGR system in an existing gas turbine based station. Although the environmental cost was lower with the EGR system, it was offset by the cost of the EGR system itself combined with the additional incremental cost of fuel due to the decrease in the thermal efficiency.