Control of flue gas emissions based on models derived from historical plant operation data

Although the last decade there is a remarkable trend towards green energy, thermal power plants still cover most of the energy production and will continue to do so in the near future. Consequently, it is important to apply methods reducing the gas emissions. Towards this goal, an innovative software application that improves the regulation of combustion parameters is presented in this paper. Data mining techniques are used to process historical data and extract models on which regulation rules are based, rather on the thermodynamic theory. Multi Agent Systems are adopted to represent the various tasks that constitute the control operation procedure. The statistical analysis suggest that a reduction of NOx and CO emissions should be expected.