Modeling the Nitrogen Oxides Reduction in Selective Catalytic Reactors and Identification of the Kinetic Parameters

Modeling the Nitrogen Oxides Reduction in Selective Catalytic Reactors and Identification of the Kinetic Parameters

Copyright @ IF AC Advanced Control of Chemical Processes, Pisa, Italy, 2000 MODELING THE NITROGEN OXIDES REDUCTION IN SELECTIVE CATALYTIC REACTORS AN...

2MB Sizes 0 Downloads 7 Views

Copyright @ IF AC Advanced Control of Chemical Processes, Pisa, Italy, 2000


s, Bittanti(o)O, A. De Marco(O), G. Fittipaldi(O), S. Malloggf), W. Prandoni(A>, L. Spinelli(O) (·Politecnico di Milano DEI, piazza Leonardo da Vinci 32, 20133 Milano, Italy (°'ENEL Ricerca, via A. Pisanol20, 50122 Pisa, Italy (A'ENEL Ricerca, via Volta 1, 20093 Cologno Monzese MJ, Italy

Abstract: In this paper. the problem of abatement of Nitrogen Oxides by Selective Catalytic Reduction (SCR) is considered. A dynamic model of the process is developed for control purposes. A distinctive feature of this model is that it takes into account the effects of possible low levels of oxygen concentration in the flue gas. The kinetic relationships have been estimated from data taken from an experimental rig designed for this specific purpose. The model has been used for the investigation of the dynamic and static behaviour of the plant as a function of the operating variables. in particular oxygen concentration and temperature. CopyrightC 2000 IFAC Keywords: Nitrogen Oxides abatement; Selective Catalytic Reduction (SCR); chemical kinetic constants; dynamic modelling.

nitrogen oxides abatement and low level of ammonia Usually slip. in various operating conditions. conventional thermal plants operate with a relatively high level of concentration of oxygen in flue gas. In such conditions. the SCR process is not affected by the oxygen concentration. The model of (Belli et al., 1996) was developed under this working assumption. Another characteristic of that model is the fact that the kinetic relations and their constants were taken from the literature. in particular from Anderson et al. (1994). The mentioned model was mostly used and appreciated in many contexts. However. as already said, that model is not valid at low oxygen concentration in the flue gas. However. the Italian Electricity Power Board (ENEL). which promoted the research activity. was also interested in the investigation of the operation of SCR in those conditions. for instance in order to use orimulsion as fuel. Recently. using an experimental apparatus at its own R&D laboratories in Livomo. Italy. an extensive testing campaign has been carried out. The collected data refer to all SCR plants used by the Company, and the data cover also the case of low oxygen. These data are at the basis of this paper, aiming at presenting a new model valid at all levels of oxygen. Among other things, the availability of such data allows the reliable identification of the kinetic constants. for the various types of catalysts. and for all operating conditions. Herein. we will present such a general model. and we will describe the process of parameter identification with reference to a specific set of data. Then. a validation of the identified model has been performed over a different set of experimental data.

1. INIRODUCTION The emission of nitrogen oxides due to the operation of thermal power plants has been increasingly controlled in the last few years. Together with sulfur oxides, nitrogen oxides (NOJ are not only harmful since they cause the acid rain, but also because their reaction with volatile organic compounds increases the level of ozone (see e.g. Boer et aI., 1990). About 95% of the molar concentration of nitrogen oxides is due to nitrogen monoxides (NO). Many techniques have been developed for NO x abatement (see Bosch and Janssen, 1987). They can be basically divided into two groups. Primary techniques (or combustion control) are based on low NOx burners, staged combustion, reburning technology. and gas recirculation. These methods control the NOx generation and possibly destruction at combustion level. Secondary techniques are based on flue gas treatment. Of course, one can combine both primary and secondary techniques in order to increase efficiency and decrease costs. In this paper we consider a secondary technique based on the so-called Selective Catalytic Reduction (SCR) where ammonia is injected into a catalytic reactor in order to reduce the nitrogen oxides in the flue gas to nitrogen. The research activity led to a previous paper (Belli et al .• 1996). where a mathematical model of the SCR process based on first principles and suitable constitutive equations was developed. The model proposed there, and the related simulator. has been used to design the control system and to verify the effectiveness of the designed reactor in relation with

o Corresponding author: [email protected]


With this model, the effects of the operating variables, in particular oxygen concentration and temperature, on the nitrogen oxide abatement is investigated. The paper is organized as follows. A list of symbols is introduced in Sect. 2. The mathematical model derivation and discussion is the subject of Sect. 3, whereas the experimental facility is described in Sect. 4. The identification procedure of the kinetic constants is outlined in Sect. 5. The influence of operating variables in steady state and dynamic conditions are analyzed in Sect.6. 2. UST OF SYMBOLS C NH3 (x, z, t) concentration of ammonia in the gas phase in the porous medium CNO(x, z, t) concentration of nitrogen oxide in the gas phase in the porous medium C02(x, z, t) concentration of oxygen in the gas phase in the porous medium C NH3.b(Z, t) concentration of ammonia in the gas phase in the channel CNO.b(Z, t) concentration of ammonia in the gas phase in the channel C02.b(Z, t) concentration of ammonia in the gas phase in the channel eN (x, z, t) fractional coverage of the surface with ammonia eo(x, z, t) fractional coverage of the surface with oxygen concentration of ammonia at the inlet C concentration of nitrogen oxide at the inlet concentration of oxygen at the inlet concentration of ammonia at the outlet CNO.ouI concentration of nitrogen oxide at the outlet CNH3.out concentration of oxygen at the outlet C0 2.ouI total mass flow in the channel Wg gas-solid mass transfer coefficient of NH3 h.i.NH3 d mass transfer coefficient of NO gas-soli h.i.NO d mass transfer coefficient of O 2 gas-soli h.i.02 effective intraporous diffusivity of NH3 DNH3 e intraporous diffusivity of NO effectiv DNO effective intraporous diffusivity of O 2 D02 NH3 adsorption constant kadsNH3 NH3 desorption constant k.Je.NH3 intrinsic kinetic rate constant O 2 adsorption constant kads02 (j) perimeter of the channel of the catalyst adsorption capacity area of the channel section Ag density of flue gas pg void fraction of the catalyst

Generally, the reactor presents a honey-comb structure in form of monolith and its constitutive materials (acting as catalyst element) are V205 and Ti0 2 plus possibly W0 3. Many different types of reaction mechanism have been proposed in the literature to describe the above mentioned global reaction. Most authors as Inomata et al. (1980), Durnestic (1992) and Pinoy and Hosten (1993) assume an Eley-Rideal mechanism. That is, after ammonia adsorption by the catalyst, a reaction between adsorbed ammonia and nitrogen oxide occurs. This reaction mechanism has been modified (pinoy and Hosten, 1995) to describe the reoxidation of the catalyst which has been reduced after the reaction of nitrogen oxide with the adsorbed ammonia. More precisely, the global reaction can be described in three subsequent main steps: • On the sites activated by oxygen (oxygen sites, Cf_O) the adsorption of gaseous ammonia allows the yield of sites activated by ammonia (ammonia sites, Cf_ONH3), NH3 +u_0+ ++u_ ONH3 • Ammonia sites react with NO in gaseous phase and after the reaction they are inactivated (no active site, Cf_OH) , u _ONH3 + NO ++ N2 + H 20+u _OH • The non active sites are then activated to oxygen sites by the adsorption of oxygen, 4u _OH +02 ++ 4u _0+2H20 A surface element (~S) of the catalyst contains a fraction (eN) of ammonia sites, a fraction (eo) of oxygen sites. Obviously, l-e N-e O is the fraction of non active sites. In according to the chemical physical theory of absorption of Langmuir, the fluxes of O 2 and NH3 adsorbed and desorbed per unit length on the catalyst thickness are

lads .m =kads.m C02(I-B N -Bo) 1 ads.NH3 =k ads .NH3 C NH3 BO

=kdts.02Bo 1 dts.NH3 = kdts.NH3BN

1 dts.02

Assuming that the oxygen desorption is negligible k.Je..02=0, so that in the sequel the flux of desorbed oxygen is set to zero. In agreement with the theory of Eley-Rideal, the reaction rate of NO and the ammonia sites (per unit length of the catalyst thickness) is proportional to the nitrogen oxide concentration and to the fraction of ammonia sites:

3. MODE L OF TIlE SCR PROCESS The coefficients kads.NH3 , kads.02, k.t.s.NH3 and k,...., are assumed to be related to the catalyst temperature, in the us range between 573-673 K, in the form of the Arrheni O the and k factor cy frequen The E;/RT). ( k=koexp law: activation energy Ea of each k..!..NH3, kads.02, k.Je..NH3 and

The SCR reaction consists in the reduction of NO to water and nitrogen in presence of oxygen by means of ammonia as reducing agent according to the global reaction


knac will be determined in the sequel from experimental data via an identification procedure. The reactor is modelled as a single channel surrounded by the catalyst. The inlet variables are oxygen, ammonia and nitrogen oxide concentration and the mass flow rate of the flue gas and its temperature at the entry of the catalyst; the outlet variables are the concentration of the reagents at the end of the channel. The reagents flow trough the channel in a forced convective motion and are diffused from the channel to the porous wall (external diffusion). Then they are diffused in the porous medium of the channel wall (internal diffusion). The mass conservation equation of Oz, NH3 and NO in the channel are considered taking into account the storage term, the convective term and their fluxes toward the walls of the catalyst. The hydrodynamic variables (total mass flow rate and pressure) are assumed uniform with respect to z. This results in Ag

iC 0 [C ----a= -wg Ol -;; 02 •b

A iC NH3•b g


02 •b )


-hdifl.02ltJLC02.b -C 02 (z,0,t)

!!.-[C Ol Pg

NH3 b •



in the channel and the surrounding catalyst has been added, see (Belli and Ferretti, 1995) for details.






CNHJ (z,O,t)]

0 (D iCNHJ) ciCNH3 it - - 8.t efl .NH 3 8.t- -J ads .NH 3 +J des .NH 3 efl.NO




J reaz

J ads.NH3



J des .NH3


ct>T = J ads.NH3 -

J des.NH3


---+z D



CNO•out C 02.out

Porous medium

Remark An important issue is whether the so-called "entrance effects" must be taken into account or can be neglected (as in (Belli et al., 1996» . Since the catalyst used in the experimental rig is relatively short, we were induced to incorporate such effects into the model and its simulator, see (Fittipaldi and SpineUi, 1998) for details. On this subject, it is worthy pointing out that our extensive simulations show that for long catalysts (as those used in normal plants) such effects are indeed negligible. As is well known, the entrance effects paly a major role in the heat and mass transfer coefficients. In our case, such dependence is particularly important for the mass transfer coefficients. For, we have made reference to the analogy with the heat transfer relationships, as in the Graetz-Nusselt problem. _

The experimental facility was designed for testing the SCR catalyst performances (DeNOx activity, S02lS03 conversion, pressure drops) on the scale of the monolith. The experimental facility is fed with a flue gas side stream coming out from the ENEL 160 MWe oil-fired unit #2 of Livorno power station. The test rig treats a gas flow rate up to 300 nm31h at a temperature ranges from 280 to 420°C. The flue gas is taken upstream the air/gas recovery heat exchanger of the boiler by a forced draft fan and then routed to the test facility by a 6" diameter 100 m length pipeline. Before entering the catalytic reactor the gas is controlled in temperature by means of an electric heater and then added with a 10% volume ammonia-nitrogen mixture. The mixture is prepared by adding the carrier nitrogen with the gaseous ammonia coming from the dry liquid NH3 plant storage (75 kg). In order to have an accurate control of the


J des .02


COIlvective flow


Finally, the mass conservation equations of the fractional coverage of the catalytic surface concerning the sites activated by NH3 and the sites activated by O 2 are

ct>d =4J ads.02 -

CNH External

The binary diffusion coefficients of 02, NH3 and NO in the flue gas are obtained from Reid et al., (1987). The effective intraporous diffusivities of the three reagents in connection with the external diffusion are evaluated from the catalyst morphological properties according to Cunningham and Geankoplis (1968) using a modified form of the Wakao-Smith random pore model (1962) for SCR catalyst.

02 02 ) 0- ( D iC1 ciC -J ads.02 +-J it - --8.t efl .02 8.t 4 des.02



Fig. 1. Geometry of a channel of the catalyst

As for the diffusion of Oz, NH3 and NO in the catalytic medium, we also take into account their fluxes toward and from the catalyst surface; correspondingly we obtain:

C--a- = 8.t0(D

x. DChannel


- hdifl .NH3W[CNHJ.b -

CNHl , CNO •in CO2 •ln

J reaz

As already mentioned, the kinetic relationships are strongly dependent on the catalyst temperature so that a suitable one dimensional model of the temperature field


reaction temperature the catalyst specimen is installed inside a thermostatic cell equipped with three thermocouples placed along the reactor axis. To avoid the risk of sulfuric acid condensation during the plant start-up and shut-down the circuit is operated with the hot air taken downstream the air/gas heater. The plant is equipped with the following main measurements: flue gas and ammonia mixture flow rate, permanent gas analysis (NOx, S02, C02, 02, CO), S03 concentration, catalyst pressure drops. The ammonia plant mass balance is verified by the determination of the ammonia slip from the catalytic reactor.

The available data is the result of an extensive campaign of measurements. Precisely, each set of data is constituted by the measurements of the operating variables: • temperature of flue gas; • inlet and outlet concentrations of ammonia; • inlet and outlet concentration of nitrogen oxides; • inlet concentration of oxygen; • area velocity; in stationary conditions (i e at steady state). Each set refers to a fixed temperature, ranging from 280 to 420°C. The experiments can be grouped into two subclasses: those experiments at low oxygen concentration, and those at high oxygen concentration. In total, we had more than 94 sets of data, organised as follows: for model estimation: 30 sets of data at high oxygen 52 sets of data at low oxygen


for validation purpose: 12 sets of data (both at high and low oxygen) STACI<

Fig. 2. Sketch of the experimental facility

5. IDENTIFICATION OF THE K1NEfIC CONSTANTS The monolith catalyst, taken as reference, is Frauenthal 3005500347. It is based on V20 S as active element. The chemical, physical and geometrical-morphological characteristics of this catalysts are known. The availability of experimental data with high and low oxygen concentration in flue gas and the possibility of simulate with the model both these conditions have allowed to obtain a "process oriented" method of identification for the kinetic constants. The kinetic constants relative to ammonia adsorption, ammonia desorption, oxygen adsorption and reaction have all been expressed in terms of the Arrhenius law

Notice that a single set of data refers to a certain temperature. The considered temperatures are: 693, 668.4,667,663.2,633.25 and 633.6 K for high oxygen, and 587.5, 625.5, 658.6 and 652 K for low oxygen. For each temperature, the measurements were repeated with different values of the operating variables. The number of repeated trials were ranging between 3 to 16. The primary unknown of the identification procedure are k tuis.NH3 , ktks.NH3 , k reaz , k tuis.02 ' It is advisable to group these unknowns into two subgroups, namely {ktuis.02 } and {ktuis•NH3 , k tks.NH3 , kreaz }. Obviously, all these quantities (in fact functions of the temperature) appear jointly into the model. However, the high oxygen data have been used to identify {ktuis•NH3 , k tks•NH3 , k reaz } whereas the low oxygen data allow the estimation of {ktuis.02 } ' Obviously, in principle, the value of ktuis.02 cannot be neglected in the identification of {ktuis•NH3 , ktks.NH3 , kreaz } since it appears in the model. This is why, we have organised the identification procedure as follows. i) High oxygen data

i.l) ktuis,02' In high oxygen conditions, as already said,

the concentrations of NO at the outlet of the reactor do not depend upon the oxygen concentration. Therefore the value given to ktuis,02 is immaterial, provided that it is taken high enough. So, we give to ktuis,02 a preliminary value for identification purposes only. The correct estimate of ktuis,02 twill be determined in the sequel from the low oxygen data (point ii.l). i.2) Identification of k tuis,NH3 , k tks.NH3 , k reaz as function of the temperature. Each set of experimental data refers to a given temperature. From each of these sets, we estimated, temperature for temperature, the values of k tuis•NH3 , k tks.NH3 , k reaz (having taken the value of k tuis•02 for granted as said at i.l). As for the estimation method, a simple direct optimisation search method ("Grid search") was used.


(2) (3)



ii) Low oxygen data

Fig. 3. This is very satisfactory if compared with thf measurement instruments precision. 6. INFLUENCE OF OPERATIVE CONDmONS

identification of kads•02 as function of the temperature. We have used the low oxygen data for the estimation of kads•02. To this purpose, we have made reference to the model described in Section 3 with kads•NH3 , kdes•NH3 ,kreaz as above obtained at point i.2). Notice that this is the function we will use in the final model, for both low and high oxygen concentrations. With such an estimate, we have again tested the high oxygen data, with practically no variation with respect to the prediction obtained by the model worked out at point i) . ii.l)

6.1 Steady state considerations

Many papers dealing with the catalytic behaviour in the removal of NOx of vanadia, titania-vanadia, titania· tungsta, and titania-tungsta-vanadia catalysts are available in the literature. However, investigations ovet commercial SCR catalyst in the form of monolith5 addressing the effects of the operating variables, in particular of oxygen, on the efficiency (Conversion Factor) of the SCR process and on the catalyst design parameter (length) are few. A main use of our model is to assess the efficiency 01 the SCR for pollutant abatement. Fig. 4 refers to the case when the ratio between the molar concentration 01 ammonia and nitrogen oxide if 0.9. The diagram shows the fraction abated nitrogen oxide versus the oxygen volume percent in the flue gas, with the temperature as a parameter. As can be seen, there is a clear increase in NO conversion up to 0.5+ 1% v/vas the percent 01 oxygen increases. For high value of oxygen, there is a saturation effect. This phenomenon confirms the investigations of (Inomata et aI., 1980; Svachula et al.,1993).

Table 1. Estimates parameters for de Nox kinetics over commercial SCR honeycomb catalyst (Frauenthal 30055(0347).



k ads ,NH3


EA 8900

6.410 7




k tks ,NH3 k reaz

6.1 10

k ads ,02



In this way, the estimation of kads•NH3 , kdes,NH3 , kreaz and kads,02 as function of the temperature is completed. From

NOoo ...

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

such function, one can determine the secondary unknowns, i.e. frequency factors and the activation energies of each function. To this purpose, one can apply linear regression to expression (1)-(4) taken in a logarithmic form. The final values of kinetic constants are reported in Table 1.




0.9 0 .8 0 .7 0.6


0 .4 0 .3 0.2






02 (%volJ

Fig. 4. NOx conversion versus oxygen concentration for different temperature values at a=O.9 .


0 .5

T =300,325,350,375,4001<



The threshold over which the oxygen does not influence the reaction can not be univocally determined since it depends on the value of the temperature. Without altering the other operative conditions, a higher temperature results into a higher threshold.

0 .1

6.2 Dynamic considerations

0 0

0.1 0 .2 0 .3 0 .4 0.5 0.6 0.7 0 .8 0 .9


As far as the dynamic behaviour is concerned, with reference to the SRC process with low level of oxygen in the flue gas, it has been verified by simulation that small step variations of ammonia mass flow rate present a transient of NO concentration at the outlet of the reactor slower than the one in the case with high level of oxygen. Large step variations corresponding to a ammonia inlet concentration from 0 to 450 ppm and from 450 ppm to 0, starting from a steady state condition with low level of oxygen concentration, are shown respectively in Fig. 5 and 6. The simulations reported in these figures refer


Fig.3. Parity plot. The model eventually obtained has been tested on the 9 validation sets of data (besides the sets of data used for estimation). The validation experiments are performed in a wide temperature range (583+626 K), with various oxygen concentration and area-velocity. The validation performance was assessed on the basis of the prediction error of the output nitrogen oxide concentration, whose value was about 7.5 % only, see also the parity table of


to standard operating conditions (CNO = 500 ppm , AV = 10 NmIh , T =350 C). NO... ~.---------------------~


reported). The model is an useful tool for reactor design since it allows the study of the effect of the operatin! conditions and the geometry of the reactor on tht abatement of NO and on the anunonia slip Furthermore, the model is the starting point for tht design of suitable control system of the SCR





~~~===-========~0~2=O~ . 5°~ ~==



100 ·5





Paper supported by MURST project Identification aM Control of Industrial Processes and CNR - Cestia (Milano).

120 145 170 195 220


Fig. 5. Outlet concentration of Nitrogen Oxide for an ammonia step concentration from 0 to 450 ppp.

REFERENCES Anderson S., L.Gabrielsson, P.L.T. Odenbrand (1994). Reducing Nox in diesel exhaust by SCR technique: experiments and simulations. AlChE Journal, 40. Boer F. P., L.L. Hegedus, T.R. Gouker, K.P. Zak (1990). Controlling power plant Nox emission. Chemtech 1990,20, pp. 312-319. Bosch H. and F. Janssen (1987). Catalytic reduction 01 nitrogen oxides. Areview of the fundamentals and technology. Catalysis today. Belli P., S. Bittanti, P. Bolzem, A. De Marco, A. Ferretti, S. Malloggi, W. Prandoni (1996). A control system for nitrogen oxides pollution abatement by SCR (Selective Catalytic Reduction). Proc. 13th Triennial Word Congress IFAC. Belli P., A. Ferretti (1995). Modellistica e controllo di un impianto di riduzione selettiva catalitica (SCR) degli ossidi di azoto prodotti da una centrale termo elettrica, Thesis, Politecnico di Milano (in Italian). Cunningham RS., C.J.Geankopolis (1968). Effects of different structures of porous solids on diffusion of gases in the transition region, I&C Fund., vol.7, p.535. Durnesic J.A., B.S.Clausen, E.Tomquist, N.Y.Topsoe (1992). Temperature-programmed desorption !Reaction and in situ spectroscopic studies of vanadialtitania for catalytic reduction of nitric oxide, Journal of Catalysis, vol. 135, p. 246. Fittipaldi G., L. Spinelli (1998). Modello dinamico e identificazione dei parametri della cinetica chimica di un reattore catalitico selettivo (SCR) per la riduzione degli ossidi di azoto prodotti da centrali terrnoelettriche), Final dissertation, Politecnico di Milano (in Italian). Inomata M., A.Miyarnoto, T.Ul, K.Kobayiashi, Y.Murakami (1982). Activities of VzOs IAlZ03 catalyst for the reaction of NO and NH3 in the presence of Oz, Ind. Eng.Chem.Prod.Res Dev, vo121 , pgg.424-428. Pinoy L.J., B.G.Massart, L.H.Hosten, The kinetics of DeNOx on a hydrated and unhydrated VzOsWOyriO z catalyst, Env. Catalysis, p.215, 1985 Reid R, 1. Prausnitz, B. Poling (1987). The properties of gases and liquids. IV Ed.. Mc Graw Hill N.Wakao, J.M.Smith (1962). Diffusion in catalyst pellets, Chem. Eng Sci., vol. 17, pg.825.

NO... ~.---------------------~


02=0.1% 300 200 100

o~----~------~------~ ·5





Fig. 6. Outlet concentration of Nitrogen Oxide for an ammonia step concentration from 450 to O. The simulations show that the dynamics response of the process is slower for the step increase than for the step decrease (- 100 s versus 10-15 s). In fact the transient following anunonia injection increase is associated with the build up of NH3 coverage: this process results from a competition between ammonia adsorption and NO reduction, the latter occwring at the expense of adsorbed NH3. On the other hand NH3 adsorption is absent after anunonia injection decrease: in this case depletion of pre-adsorbed NH3 results primarily from its surface reaction with NO. This behaviour is also apparent at low oxygen concentration, where the characteristic times are increased for the limitation to adsorption phenomena due to oxygen. The figures indicate also that the transients of the outlet nitrogen oxide concentration associated to high and low oxygen concentration are different.

7. CONCLUSIONS In this paper, a dynamic model of an SCR reactor has been presented. The identification of the kinetic constants from experimental data is described. A distinctive feature of this model, in comparison with our previous model (Belli et al, 1996) is that the reoxidation phenomenon of the catalytic surface is taken into account. In this way, the model is valid for low oxygen concentration in the flue gas as well. The model can be used for both stationary and dynamic simulations. In all cases, it provides good fitting of experimental data (here only the static fitting is