Selective catalytic reduction: testing, numeric modeling, and control strategies

Department Head: Allan Thomson Kirkpatrick.2010 Spring.Includes bibliographical references (pages 96-99).Selective Catalytic Reduction (SCR) catalysts respond slowly to transient inputs, which is troublesome when designing ammonia feed controllers. An experimental SCR test apparatus was installed on a Cooper Bessemer GMV-4 natural gas engine. Transient data was taken of commercially available SCR Catalysts. These transient tests are used to quantify SCR catalyst response. Space velocity, catalyst temperature, inlet NOx concentration, and ammonia to NOx molar feed ratio were varied. A Simulink numeric model was created to examine the SCR transient phenomena. The Simulink numeric model showed in-catalyst ammonia and NOx concentration as a function of length in the direction of exhaust flow. This helped explain the SCR transient results. Transient testing showed a fifteen minute delayed response in NOx reduction from ammonia transitions. Ammonia slip succeeded ammonia transitions by thirty minutes. Simulink modeling revealed that these delays are caused by large quantities of ammonia stored in the catalyst. Due to ammonia storage, ammonia waves propagate through the catalyst, front to back. Emission of these constituents through the catalyst is delayed because the wave takes time to propagate through the entire catalyst length. Ammonia feed rate control testing was done on the experimental setup to improve ammonia and NOx emissions from the catalyst. Three control algorithms were used: feed forward control, using a pre ammonia injection ceramic NOx sensor; a feed forward plus feedback control, using a pre ammonia injection ceramic NOx sensor and post catalyst ceramic NOx sensor to generate feed signals; and a feed forward plus feedback algorithm that used a pre ammonia injection ceramic NOx sensor and a mid catalyst ceramic NOx sensor to generate feed forward and feedback signals. The feed forward controller used molar ratio as the control variable, and the feedback system used a technique that minimized the post catalyst ceramic NOx sensor signal. Ammonia to NOx molar ratio was stepped every five or fifteen minutes, and the algorithm made decisions, based on the catalyst response to the step. The decisions were made to minimize the post catalyst ceramic NOx sensor. Feed forward testing revealed that the lack of pressure compensation on ceramic NOx sensors causes errors in feed forward NOx readings, and sub optimal ammonia feed. Feedback testing revealed that a minimization technique can be used successfully with a feedback step rate of one step per fifteen minutes, and a step size of 5% ammonia to NOx molar ratio. The feedback algorithm, with the feedback ceramic NOx sensor located one third the way through the catalyst length, worked poorly. The technique approached a lean ammonia to NOx molar ratio, and stabilized slower than the post catalyst feedback ceramic NOx sensor technique. These phenomena are explained with the Simulink numeric model