Bitte wählen Sie ihr Lieferland und ihre Kundengruppe
NO(x) models depend on specific spatial configuration of fuel-air lines supplying pulverized-fuel boiler. Therefore, the setting elements (air dampers, coal feeders) get significant input variables of the models. Validation procedures of uncertainty air dampers and air flows are important stage of data processing for on-line mode of software flue gas analyzer. Models of CO2 result essentially from the balance of fuel and air streams fed into combustion chamber. These models demonstrate the highest sensibility on air fans control signals, the sum of coal feeder's settings and electric power. CO is created in case of oxygen deficiency, usually in local areas of combustion chamber, where oxygen concentration is not measured. Therefore, quality of created CO models is worse. Investigations show, that better results can be achieved by training CO model with the data obtained by application of moving average procedure. It is sufficient in case of environment pollution estimation. Similar accuracy as for MLP (Multi Layer Net) neuronal nets was obtained for radial basis function networks with about few hundreds of radial neurons.