ADAPTIVE CONTROL OF NITROGEN OXIDE REDUCTION PROCESSES

RECENT LEGISLATION IN THE U.S. AND ABROAD, DRIVEN PRIMARILY BY ENVIRONMENTAL CONCERNS, HAS LED TO THE DEVELOPMENT OF TECHNOLOGY TO REDUCE THE LEVELS OF NITROGEN OXIDE (NOX) EMITTED FROM COMBUSTION FACILITIES. APPLICATION OF THIS TECHNOLOGY TO COAL-FIRED POWER GENERATION FACILITIES HAS BEEN PARTICULARLY IMPORTANT BECAUSE OF THE RELATIVELY HIGH NITROGEN CONCENTRATION OF TYPICAL COALS (0.5 TO 2%). SELECTIVE CATALYTIC AND NON-CATALYTIC REDUCTION (SCR AND SNCR) ARE FLUE GAS CLEAN-UP PROCESSES THAT USE A REDUCING AGENT (USUALLY AMMONIA OR UREA) TO CONVERT NOX INTO MOLECULAR NITROGEN AND WATER. ALTHOUGH THESE PROCESSES CAN OFTEN REDUCE NOX CONCENTRATIONS BY OVER 90%, VARIATIONS IN PROCESS TEMPERATURES OR IMPROPER CONTROL OF AMMONIA INJECTION RATES CAN LEAD TO INCREASED NOX CONCENTRATIONS OR "AMMONIA SLIP". AMMOONIA SLIP NOT ONLY DEFEATS THE GOAL OF REDUCED EMISSIONS (AMMONIA IS AN AIR TOXIC), BUT CAN ALSO CAUSE OPERATIONAL PROBLEMS IN THE PLANT, SUCH AS CORROSION, AND WILL CONTRIBUTE TO ASH FOULING. THE COMPLEXITY OF THESE PROVESSES, VARIABILITY IN BOILER LOAD AND TEMPERATURE PROFILES, AND THE DIFFICULTIES ASSOCIATED WITH MONITORING AMMONIA SLIP ALL CREATE A DIFFICULT CONTROL PROBLEM FOR NOX REDUCTION USING FLUE GAS CLEAN-UP TECHNOLOGY. THIS PROJECT WILL EMPLOY ARTIFICIAL NEURAL NETWORKS (ANNS) TO ADAPTIVELY MODEL AND CONTROL NOX REDUCTION PROCESSES USING THE BOILER AND PROCESS PARAMETERS AS INPUT AND THE RESULTING NOX AND AMMONIA CONCENTRATIONS AS OUTPUT. THE CAPABILITY OF ANNS TO LEARN FROM OBSERVATION, TOGETHER WITH THEIR INHERENT ABILITY TO MODEL NONLINEARITY, MAKE THEM IDEALLY SUITED TO THE PROBLEM OF PROCESS CONTROL IN COMPLEX PROCESSES. THE PHASE I PROJECT WILL CONCENTRATE ON FLUE GAS CLEAN-UP PROCESSES, ALTHOUGH THIS METHODOLOGY SHOULD ALSO BE APPLICABLE TO MODIFIED COMBUSTION STRATEGIES SUCH AS AIR STAGING. NEW AMMONIA MONITORING TECHNOLOGY CURRENTLY UNDER DEVELOPMENT WILL PROVIDE PREVIOUSLY UNAVAILABLE INPUT DATA FOR MODEL DEVELOPMENT.