| OCR Text |
Show CONTROL OF NOx USING MODIFIED RESPONSE SURFACE METHODOLOGY (MRSM) Joseph Colannino, P.E. Technology Development Leader John Zink Company, Tulsa, Oklahoma ABSTRACT Today, stringent regulatory limits demand new techniques for better controlling N O x. Large furnace volumes and extractive sampling processes inherently delay N O x response to process changes. This frustrates and destabilizes conventional feedback control. To maintain stability one must decrease controller gains. This results in sluggish control with frequent excursions above regulated limits. Feedforward control offers an attractive solution. It can compensate for N O x before the process departs from ideal. However, feedforward control requires an accurate predictive emissions model. This paper presents such a model, describes the important process variables, and gives effective N O x control algorithms n o w in successful use. One may easily incorporate the technique, known as modified response surface methodology ( M R S M ) , into programmable loop controller (PLC) platforms or distributed control systems (DCS). What is Wrong With Traditional Feedback NOx Control? Feedback N O x control suffers from several maladies. First, N O x production is highly nonlinear. Small perturbations of inlet conditions may make large changes in outlet N O x . This makes N O x control difficult and unstable. Second, large furnace volumes tend to integrate, blur, and delay N O x response. The analysis system itself also requires sample extraction and conditioning. So some minutes must elapse before the control system can perceive a bonafide change. Delays increase the order of the system and can make an otherwise stable control loop, unstable. To combat this instability one must slow down the controller response. This makes for sluggish control. Regulated emissions frequently run out of limits. Feedforward NOx Control Feedforward control makes use of advance information to adjust the variables before the process has time to respond. However, feedforward control requires an accurate N O x - predicting algorithm. One derives the algorithm as follows. The Explicit NOx Relation From experience w e know that combustion-related N O x is a function of the following variables: 1. firing rate ( A ) ; 2. excess oxygen concentration (y); 3. oxygen concentration in the plenum or windbox (yw) for the case of flue gas recirculation; 4. degree of fuel staging, fuel blending, diluent injection (e.g. water or steam) |
