Characterizing and modeling of an 88 MW grate-fired boiler burning wheat straw: Experience and lessons

Characterizing and modeling of an 88 MW grate-fired boiler burning wheat straw: Experience and lessons

Energy 41 (2012) 473e482 Contents lists available at SciVerse ScienceDirect Energy journal homepage: www.elsevier.com/locate/energy Characterizing ...

2MB Sizes 1 Downloads 54 Views

Energy 41 (2012) 473e482

Contents lists available at SciVerse ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Characterizing and modeling of an 88 MW grate-fired boiler burning wheat straw: Experience and lessons Chungen Yin a, *, Lasse Rosendahl a, Sønnik Clausen b, Søren L. Hvid c a

Department of Energy Technology, Aalborg University, 9220 Aalborg East, Denmark Risø DTU, Technical University of Denmark, 4000 Roskilde, Denmark c DONG Energy, Kraftværksvej 53, 7000 Fredericia, Denmark b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 8 December 2011 Received in revised form 3 February 2012 Accepted 23 February 2012 Available online 27 March 2012

Grate-firing is one of the main technologies currently used for biomass combustion for heat and power production. However, grate-firing is yet to be further developed, towards a better technology for biomass combustion, particularly towards higher efficiency, lower emissions, and better reliability and availability. To better understand grate-firing of biomass and to establish a reliable but relatively simple Computational Fluid Dynamics (CFD) modeling methodology for industrial applications, biomass combustion in a number of different grate boilers has been measured and modeled. As one of the case studies, modeling effort on an 88 MW grate-fired boiler burning wheat straw is presented in this paper. Different modeling issues and their expected impacts on CFD analysis of the kind of grate boilers are discussed. The modeling results are compared with in-flame measurements in the 88 MW boiler, which shows an acceptable agreement. The discrepancies are analyzed from different aspects. The lessons learned and experience gained from this and other case studies are summarized and discussed in detail, which can facilitate the modeling validation effort as well as improve grate-firing technology. Some of the addressed measures will be tested in a modern 500 kW grate boiler rig. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Biomass Grate-firing CFD Higher efficiency Lower emissions

1. Introduction The worldwide concern with global warming and the limited availability of fossil fuels has spurred interest in using biomass for energy production. To achieve the pressing near-term targets for significant increase of the share of renewable energy sources in energy system and reduction in CO2 emissions, co-firing biomass with fossil fuels in existing power plants offers an attractive option. Over the past decade, co-firing has been demonstrated in about 200 power plants worldwide, among which about 48% of the plants are equipped with pulverized fuel boilers, 24% with bubbling fluidized bed boilers, 19% with circulating fluidized bed boilers, and 9% with grate-fired boilers [1]. Grate-fired systems have greater fuel flexibility than pulverized fuel boilers and can be operated on 100% raw biomass. Grate-fired boilers are not as sensitive to fuel bed agglomeration as fluidized bed combustors. This is a great advantage when applied for biomass combustion, since biomass fuels often have low ash melting temperatures. However, grate-fired systems are yet to be further developed in order to achieve

* Corresponding author. Tel.: þ45 30622577; fax: þ45 98151411. E-mail address: [email protected] (C. Yin). 0360-5442/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. doi:10.1016/j.energy.2012.02.050

a better grate-firing technology, e.g., towards higher efficiencies, lower emissions, larger-scale applications, better reliability and availability. A series of comprehensive studies of biomass combustion in different full-scale grate-fired systems have been performed, in order to better understand grate-firing of biomass and to establish a reliable but relatively simple CFD modeling methodology for industrial applications. The latter is a very effective tool for design and optimization of grate-firing systems. As one of the case studies, biomass combustion in a 108 MW grate boiler at an indirect, parallel co-firing plant was investigated [2]. The 108 MW grate boiler fires 100% straw and the generated steam is integrated into the steam cycle of a 430 MWe ultra-supercritical utility boiler for heat and power production. In such a parallel, indirect co-firing installation, all the potential problems associated with straw combustion are restricted to the grate boiler and have no influence on the main utility boiler. In this paper, another case study, the modeling and measuring of an 88 MW grate-fired boiler burning straw at a direct co-firing plant, is presented. From this and other case studies, the lessons and experience that can facilitate the modeling validation effort are discussed and some measures that may improve grate-firing technology are also addressed.

474

C. Yin et al. / Energy 41 (2012) 473e482

2. The grate boiler and in-flame measuring The grate boiler was constructed in 1997, producing max. 34 kg/ s steam of 557  C and 226 bar. The maximum heat production of the boiler is 88 MW. The boiler is equipped with four water-cooled, vibrating grates between the two sidewalls of the furnace. The grates are located at the bottom of the combustion chamber, as sketched in Fig. 1(a). The grates have two main functions: lengthwise transport of the fuel from the inlet on the front wall to the ash discharging system on the opposite wall, and distribution of the primary air (PA) entering from the wind-boxes beneath the grates through small holes on the grates. A close view of the grates, including the water tubes, small holes and the space at the joint between neighboring grates, is given in Fig. 1(b). The profile of PA flux lengthwise along the grate is controlled in design by adjusting the air flow rate out of the different wind-boxes and using nonuniform distribution of small holes on the grate. Based on the PA flux, the grate may be divided lengthwise into four zones: pregrate, zone 1, zone 2, and zone 3, as indicated in Fig. 1(a). The pre-grate zone and zone 1 are lumped together and connected to the same windbox. Therefore, the PA flux in these two zones is mainly controlled by the non-uniform distribution of the small holes. By finely controlling the lengthwise PA flux, the designed fuel conversion rates along the grate could be achieved. For example, the majority of moisture (e.g., 85%) in the straw is expected to be released in the pre-grate zone and the remaining is released in zone 1. The opposite is applied to volatiles, i.e., most of the volatiles (e.g., 85%) released in zone 1 and the rest released in the pre-grate zone. Most of the char oxidation is assumed to occur in zone 1 and zone 2 (e.g., 35% and 55% in the two zones, respectively) whilst the rest is equally split onto the pre-grate zone and zone 3.

Since the grates are water-cooled, the PA is only confined to combustion requirement and the boiler is flexible with the use of an advanced secondary air (SA) system (including over-fire air, OFA). There are five groups of SA nozzles on the front wall (i.e., “T1-T4”, “F1-F2”, “F3-F4”, “KF”, and “OFA11-13”) and four groups of SA nozzles on the rear wall (“B1”, “B2”, “KB”, and “OFA21-23”) in this boiler, consisting of more than 200 individual air nozzles of different diameters and orientations. Some of the SA nozzles are indicated in Fig. 1(c). The staged SA jets of different momentums and orientations play a very important role in mixing, combustion and emissions from grate-fired boilers [3]. An 8-h measurement campaign is conducted mainly to measure the gas temperature and composition in the flame. The boiler operates comparatively steadily under a condition close to 100% load during the campaign. Table 1 summarizes the operating conditions averaged over the 8 h, which include data for fuel, air, water and steam. Here it has to be mentioned that the PA flow rate given in Table 1 is the corrected rate after accounting for the leakage air, which is believed to be mainly from the grate system. The lack of good sealing between the neighboring grates may serve as dilatation joints and the air leaked through them may help to control the temperatures of the local mechanical components in the lower part of the furnace. From the flue gas flow rate, O2 at the boiler exit, straw feeding rate and straw composition collected in the control room, the leakage air into this grate boiler is estimated to be about 20% of the total air supply as indicated in the control room. In this study, the leakage air is introduced through the PA stream. After this correction, the total PA flow rate is 12.16 kg/s and the total SA flow rate is 16.54 kg/s. The split ratio, PA/SA, is about 42/58, which is close to 40/60 preferred and often used in modern grate-fired boilers burning biomass. The low PA/SA ratio, compared

Fig. 1. Sketch of the water-cooled vibrating grate boiler, and close view of a straw inlet and grate (from the inside).

C. Yin et al. / Energy 41 (2012) 473e482

475

Table 1 The operating conditions averaged over the 8-h measurement campaign. 1) Straw: Feeding rate and the composition Feeding rate [kg/s] Moisture %wt (as received) 4.499

13.4

Ash C %wt (dry)

H

O

N

Volatiles

Fixed carbon

Lower heating value MJ/kg (as received)

4.5

5.9

41.9

0.7

79

16.5

14.9

47

2) Primary air (PA): the flow rate distribution along the lengthwise direction of the grate Four lengthwise zones Pre-grate Zone 1 Length [m] PA flow rate [kg/s] PA temperature [K]

0.915 2.143 563

1.403 3.287

3) Secondary air (SA): Flow rate distribution through different groups of SA nozzles Location On the front wall SA Nozzle-group Number of nozzles SA flowrate [kg/s] SA temperature [K]

T1 26 1.59 563

T2 28 1.59

T3 28 1.59

T4 26 1.59

F1-F2 29 1.17

4 5 6

ECO3 (counter flow) Furnace wall (from bottom to top) Roof, horizontal pass and 2nd pass walls Spray water 1 SH1 (counter flow) Spray water 2 SH2 SH3

284.3 337.5 390.4 337.5 415.2 337.5 436.6 449.5

337.5 390.4 420.3 337.5 454.7 337.5 449.5 470.0

to 80/20 in old units, secures a near char combustion stoichiometry condition in the fuel bed, avoids excessive NOx formation in the primary combustion zone, and leaves much more freedom to advanced SA supply in the freeboard. The gas temperature and composition in the flame are measured simultaneously with a fiber optic probe connected to a Fourier transform infrared (FTIR) spectrometer. The experimental setup is shown in Fig. 2, which comprises a water-cooled fibre-optic probe connected to an FTIR spectrometer, a purge generator, a blackbody calibration source and a computer for collecting, processing and storing spectra. Because of the strong background thermal radiation signal from the instrumentation, a two-point calibration procedure of the FTIR spectrometer is required, which needs two blackbody calibration sources. The fibre-optic probe was calibrated at 375  C with a portable blackbody source (as shown in Fig. 2) and at room temperature with a piece of cardboard. The cardboard at room temperature emits thermal radiation (including reflected ambient thermal radiation) like a blackbody. It is a fast and accurate

Zone 3

2.109 3.493

1.953 3.235

On the rear wall F3-F4 14 1.17

KF 4 0.62

OFA1 12 2.13

4) Water/stream flow path inside the boiler (from water in ECO3 to the main stream out of SH3) No. Heating surfaces Temperature [ C] Pressure [bar] Inlet Outlet Inlet Outlet 1 2 3

Zone 2

230 229.1 215.5 229.1 213.7 229.1 212.0 211.1

229.1 215.5 214.6 229.1 212.9 229.1 211.1 210.2

B1 10 1.17

B2 6 1.17

KB 5 0.62

OFA2 18 2.13

Mass flow [kg/s]

Heat flux [MW]

31.89 31.89 31.89 0.50 32.39 1.25 33.64 33.64

9.376 35.94 6.857 6.251 1.985 2.826

method to correct the measured FTIR emission spectra for effects due to self-emission of parts in the FTIR spectrometer and experimental setup. A close view of the probe head is also shown in Fig. 2. The tip of the protection tube in front of the optical mount and a beam stop defines the optical path. The beam stop, e indicated in Fig. 2, is water-cooled in order to stabilize the temperature, and reflections are reduced as the surface is grooved and coated with high-emissivity paint. The gas concentrations are determined by comparing the measured transmittance spectra with a spectroscopic database and validation measurements using the Hotgas facility. The gas temperature is found from the thermal radiation at the 2300 cm1 CO2 fundamental band. More details on the measurement technique can be found in [4]. For the gas temperature measurement, two values are estimated from the measured spectra, because the measuring volume is a slab of inhomogeneous gas at highly turbulent conditions with gas mixing over 250 mm path. One is the average temperature and the other is the highest temperature detected over the 250 mm path, labeled “Measured T”

Fig. 2. (i) FTIR measurement probe connected to an FTIR spectrometer, with the measurement distance, L, defined; (ii) A cross-sectional view of the probe head. The arrow illustrates the desired gas flow direction with minimum disturbance.

C. Yin et al. / Energy 41 (2012) 473e482

and “Measured T_max” in the results, respectively. At the measuring port, F (as indicated in Fig. 1), oxygen concentration, which can not be measured by the FTIR, is measured by the plant using a conventional multispecies gas analyzer.

reality they are expected to be better represented by parabolic profiles. In the role of a reliable CFD analysis, the correct total fluxes of mass, momentum, elements and heat into a combustor at different inlets largely overwhelm their detailed profiles (e.g., constant, linear or parabolic). This is also applicable to grate-fired boilers, in which the gas leaves the fuel bed into the freeboard at quite low velocities and then gets mixed in the freeboard with a large amount of high-momentum SA jets. In this study, an empirical model is used for the in-bed straw conversion. Based on the expected straw conversion rates as a function of the lengthwise position on the grate, the combustion gas released from the fuel bed into the freeboard can be calculated from the overall heat and mass balances of the three different streams (i.e., straw, preheated PA, and incident radiation heat flux). In this model, the straw volatiles are lumped into one single “species”, CH2.306O1.031N0.02, whose enthalpy of formulation is determined from the straw analysis data to be 193,904 kJ/kmol. Char in the fuel bed is assumed to be oxidized into CO. So, totally 5 species (i.e., H2O vapor, CH2.306O1.031N0.02, CO, O2 and N2) are considered in the gas released from the fuel bed into the freeboard. Temperature-dependent specific heats for all the gas species and the solid phase (e.g., char) are used to correctly calculate the gas flow velocity and temperature. The calculated lengthwise profiles at the fuel bed top are then used as the inlet condition for the freeboard CFD simulation. Fig. 3 shows the profiles of gas species, velocity and temperature at the fuel bed top when the coupled simulation is converged. Initially, the empirical in-bed conversion model starts with a guessed incident radiative heat flux onto the fuel bed, 4 MW, producing different profiles of the gas species, velocity and temperature. After quite some iterative switches between the in-bed conversion modeling and freeboard CFD simulation, the final net radiative heat transfer incident onto the fuel bed top is converged to 1.63 MW, which is used in the calculation of the profiles in Fig. 3. Integration of the profiles along the grate length is made to assure such profiles match correctly the desired total fluxes of mass, momentum, heat and elements into the freeboard. Here discussions about the volatile gases, char oxidation and soot particles released from the fuel bed are given in detail. In this study, all the volatile gases released during the in-bed devolatilization process are lumped into one single species, CH2.306O1.031N0.02. The influence of different representations of fuel volatiles in modeling of biomass combustion was investigated [14]. In one case, the straw volatiles were represented by a gas mixture consisting of a number of real gas species, based on the existing knowledge on pyrolysis and devolatilization process of straw particles. In the other case, the straw volatiles were lumped into one single “artificial” gas species. For the two different representations, the total fluxes of mass, different elements and heat are precisely same. When the same

3. Mathematical modeling 3.1. Overall modeling methodology Numerical simulations are also done for the condition given in Table 1. Modeling of straw combustion in a grate boiler involves two parts: modeling of straw conversion in the fuel bed and simulation of gas-phase reactions in the freeboard. The two processes are strongly coupled by the combustion gas leaving the fuel bed into the freeboard and the radiative heat flux emitted by the flame and furnace walls onto the fuel bed. The coupled modeling methodology needs to iteratively switch between the in-bed fuel conversion modeling and the freeboard computational fluid dynamics (CFD) simulation, until there is no substantial change in either the combustion gas leaving the fuel bed or the radiative heat flux incident onto the fuel bed. Such a method has been successfully used in fixed bed combustion and grate-firing modeling [2,5e12]. 3.2. In-bed fuel conversion modeling The in-bed fuel conversion modeling is to derive the profiles of temperature, composition and velocity of the gas leaving the fuel bed into the freeboard, based on the feeding rate and composition of the straw, the flux of the preheated PA from under the grates and the incident radiative heat flux at the top of the fuel bed. The derived profiles will be used as the grate inlet conditions for the freeboard CFD simulation. The in-bed fuel conversion can be modeled in different ways. For instance, a detailed in-bed conversion model, FLIC, was successfully developed and employed for conversion of waste [5] and straw [7,12] in grate-fired boilers; while empirical models for in-bed biomass conversion were also often used in numerical studies of grate-fired boilers [2,8,13]. The effect of in-bed fuel conversion models was investigated in the numerical study of a large-scale grate-fired boiler, in which the CFD results based on a detailed in-bed conversion model were compared with those based on an empirical model [2]. It was found that the influence of in-bed conversion models was virtually restricted to the vicinity of the fuel bed and the predicted flow pattern, gas species and temperature distribution in the majority of the freeboard volume did not show remarkable difference. This may be understood. For instance, in most of the large-scale combustion simulations in literature top-hat profiles of velocity, species and temperature are often used as the inlet conditions, although in

1500

1.5

1200

1.2

900

0.9

600

0.6

300

0.3

0.4

Gas temperature, Tg [K]

Species mass fraction, Y i [-]

0.45

0.35 0.3 0.25 0.2 0.15 0.1 0.05

0

0 0

1

2

3

4

5

6

Distance along the grate away from the straw inlet, x [m] O2

Volatiles

CO

H2O

7

0 0

1

2

3

4

5

6

7

Distance along the grate away from the straw inlet, x [m] Tg

Vy

Fig. 3. Lengthwise profiles of gas species, velocity and temperature at the fuel bed top when the coupled simulation is converged.

Upward velocity, Vy [m/s]

476

C. Yin et al. / Energy 41 (2012) 473e482

homogenous combustion mechanism and turbulence-chemistry interaction were used, the different representations of the straw volatiles were found to make no remarkable difference in the CFD results of a 500 kW straw/coal co-firing flame, both of which showed a good agreement with the experimental results [14]. The effect of such an issue was also investigated in a numerical study of oxy-fuel combustion of natural gas, in which the real natural gas consisting of 8 species was used in one case and lumped into one single species in the other case. They made no difference at all in the predicted combustion characteristics, as long as the same combustion mechanism and turbulence-chemistry were used and the lumped species was correctly defined in terms of mass, elements and heat [15]. Consequently, one can expect in the modeling of this 88 MW grate boiler the use of one single “artificial” species to represent the straw volatiles will be sufficient, especially when the two-step global mechanism combined with the finite rate/eddy-dissipation is used in the freeboard CFD simulation. About the in-bed char oxidation, the primary products are CO and CO2,

CðsÞ þ aO2 ¼ 2ð1  aÞCO þ ð2a  1ÞCO2

(1)

where the split ratio of CO/CO2 is CO/CO2 ¼ 2500 exp(6420/T) for the temperature range of 730e1170K [16], and ratios out of this temperature range can be calculated using one of the limiting temperatures [7]. From the plot of the split ratio of CO/CO2 as a function of temperature (shown in Fig. 4), one can see that CO well dominates over CO2 in the oxidation product at high temperatures. In addition, the content of char in the straw fired in this grate boiler, 16.5%wt, is relatively low. Therefore, it is assumed in this study that CO is the only product of the in-bed char oxidation, which is expected to have a minor effect on the freeboard CFD simulation. In this study, that soot or fly ash particles are not included in the freeboard CFD simulation. From the design and operation point of view, high particle concentration in freeboard will cause deposition on heating surfaces and also largely increase CO emissions. From practical point of view, the ash content in the straw is very low (<5%wt) and only about 10e20% of the total fuel ash will be entrained in the flue gas in such a grate boiler because of the low PA flow rate and the low gas velocity at the top of the fuel bed. Therefore, it may be acceptable to neglect soot and ash particles, at least for mixing, combustion and emissions simulation in the freeboard. This is partly justified by the site observations: no obviously visible solid particles are entrained into the freeboard.

12

CO/CO2

2500 exp( 6420 / T )

10

CO/CO 2 [-]

8 6 4 2 0 800

850

900

950

1000

1050

1100

1150

1200

Temperature, T [K] Fig. 4. The split ratio of CO/CO2 in the products of char oxidation, as a function of temperature.

477

The maximum size of particles that can be entrained into the freeboard along the grate may be estimated to be about a few hundred microns from the force balance,

1 ! ! ! v jð u  ! v Þ ¼ mp g C Ap rg j u  ! 2 D

(2)

! ! where CD, Ap, rg, u , ! v , mp, and g represent the drag coefficient, particle area normal to the drag force, gas density, gas undisturbed velocity at the point occupied by the center of mass of the particle, particle velocity, particle mass, and gravitational acceleration. Because of the low gas velocity at the top of the fuel bed, the amount and size of the particles entrained into the freeboard are expected to be small or negligible.

3.3. CFD simulation in the freeboard The CFD simulation in the freeboard is basically gas-phase combustion modeling. The inlet conditions at the grate and SA inlet conditions at different nozzles are defined by Fig. 3 and Table 1, respectively. The temperatures of different walls (including SH2 and SH3 panels) are estimated from the water/steam cycle in each section (see Table 1), and their emissivities are finely adjusted during the simulations to assure that the heat flux in each wall reported from the gas/flame side matches with the counterpart calculated from the water/steam side. The SH1 and ECO3 are both modeled as porous zones with known flow resistances and heat sinks. The CFD simulation is based on the commercial code, Ansys Fluent. The standard k-ε model is used for turbulence closure. Discrete Ordinates model is used for radiative heat transfer, in which the angular resolution is 2  2  8 ¼ 32 discrete directions. A two-step global reaction with CO as the intermediate species is used for combustion of the volatiles (CH2.306O1.031N0.02), and finite rate/eddy-dissipation model is employed to account for turbulence-chemistry interaction.

CH2:306 O1:031 N0:02 þ 0:561O2 ¼ CO þ 1:153H2 O þ 0:01N2

(3)

1 CO þ O2 ¼ CO2 2

(4)

d½CH2:306 O1:031 N0:02  kinetic reaction rates are, dt   8 210  d½CO CH2:306 O1:031 N0:02 0:2 ½O2 1:3 and ¼ 51012 $exp  Ru T dt   1:7108 0:25 12 ½CO½O2  ¼ 2:2410 $exp  , respectively, which is Ru T similar as those used in [14]. The homogeneous combustion in the freeboard is assumed to be mainly controlled by mixing. The refined eddy dissipation model constants, A ¼ 0.6 and B ¼ 0.5 as used in [2], are employed in this study to calculate the mixing rate. For the freeboard CFD simulation, a lot of efforts were also made to generate a fine, high-quality mesh. Mesh always plays a crucial role in CFD simulations: it largely affects the convergence and significantly influences the reliability of the results. Fig. 5 shows the computational domain (from the top of the fuel bed to the crosssection after ECO32), the overview of the final mesh for the key furnace section (with all SA nozzles), and the local close-up of the mesh. The mesh consists of 4,320,316 cells, all of which are hexahedrons. The quality specifications of the mesh are summarized in Table 2. This mesh not only adequately resolves the geometric details (including all the individual SA nozzles) and key flow physics in the boiler but also has been proven fine enough to obtain a grid-independent CFD solution. The

two

478

C. Yin et al. / Energy 41 (2012) 473e482

Fig. 5. The computational domain and the overview and local close-up of the final mesh used in the simulation.

4. Results and discussion 4.1. Validation efforts of the CFD analysis Fig. 6 shows some of the validation results, in which the CFD results are compared against the measurements at a few representative ports. The CFD results show an overall acceptable agreement with the measurements in both the gas temperature and main species concentration, considering uncertainties from the

boiler operation itself as well as those from the measuring and modeling efforts. Although the boiler operates comparatively steadily during the 8-h measurement campaign, there are still some variations in operation from time to time because of, e.g., the intermittent grate vibration at a certain frequency and amplitude, discontinuous biomass feeding, very common combustion instabilities inside the fuel bed. During the measurement campaign, the FTIR system needs to be moved from port to port. At each measuring port, the

Table 2 Summary of the quality of the mesh. Quality-type specification

The mesh of this grate boiler: 4,320,316 cells in total; all hexahedral Range

Average

Equi-angle skew, QEAS

[0, 0.69] among which 90.23% in [0, 0.25], 9.42% in [0.25, 0.5] and 0.35% in [0.5, 0.69] [1, 7.02] [1, 3.5] [1, 6.8] [0, 0.82]

0.150

Aspect ratio, QAR Diagonal ratio, QDR Edge ratio, QER Stretch, QS

1.624 1.257 1.637 0.251

On the quality-type specifications (by definition)

0  QEAS  1: 0 for equilateral element; 1 for a completely degenerate element. 0.4 (average) for high-quality 3D mesh. QAR  1: 1 for equilateral elements QDR  1: 1 for cubic hexahedrons QER  1. The higher the value, the less regularly shaped the element. 0  QS  1: 0 for equilateral elements; 1 for a completely degenerate elements

C. Yin et al. / Energy 41 (2012) 473e482

479

Fig. 6. CFD results vs. FTIR measurements at the measuring ports, E, F, G and H.

FTIR probe is inserted to different depths in the furnace to measure the local gas species and temperature. At each depth, the FTIR measurement lasts for about 20 min, and the results averaged over the period are indicated in Fig. 6. So what the FTIR system measures represent “instantaneous” results during the 8-h measurement campaign. Different “instantaneous” CFD simulations could be defined correspondingly, which may be better used to validate the measurements. For instance, at 1.85 m into the furnace at port F, the FTIR measurement is done for about 20 min. The operational conditions during the 20 min can be obtained and averaged as the input conditions for a CFD simulation. The converged CFD solutions can be used to compare with this FTIR measurement. In the similar way, a different CFD needs to be defined, when the FTIR probe is inserted to a different depth at a different port. Such a strategy is feasible, but may not be realistic in practice especially when the FTIR measurements are made at quite some locations. Moreover, there exists a certain lag in the combustion characteristics (e.g., temperature and emissions) in relation to the operational conditions. In this study, the operational conditions are averaged over the entire 8 h (as listed in Table 1) and used as the inputs of the CFD simulation. Other uncertainties with the operational conditions include the SA flow distribution through individual air nozzles, the leakage air flow rate and the way to account for the leakage air. For instance, there are five groups of SA nozzles on the front wall in this boiler. The group “T1-T4” includes 4 sub-groups, T1, T2, T3 and T4, each of which consists of 20e30 individual air nozzles, as shown in Fig. 1(c). In the plant, only the total air flow to the entire group “T1T4” is measured and controlled; while the air flow to each of the four sub-groups is unknown, not to mention the flowrate in each

individual nozzle. In this study, it is assumed the total flowrate is uniformly distributed through the different sub-groups and individual nozzles. In addition, there is a significant amount of leakage air into the boiler, estimated up to about 20% of the total flowrate indicated in the control room. In this study, the leakage air is introduced through the PA stream. Considering most of the leakage air may be from the very narrow spaces between the neighboring grates (rather than from the beneath the grates) and the draught effect generated by the leakage air stream inside the combustion chamber, it could be more reasonable to define the spaces between the grates as leakage air flow inlets. From the freeboard CFD simulation side, the uncertainties may have been minimized through the efforts on high-quality, gridindependent mesh, judicious consideration and selection of models, and perfect overall calibration results (e.g., overall heat, mass and species balance, and good agreement between the CFDpredicted heat flux at each wall section and the counterpart calculated from water/steam side). The uncertainties from the measuring technique itself may also play a minor role. The only uncertainty with the FTIR system is that the use of the HITEMP database might lead to a systematic too high H2O and CO2 estimate. The accuracy of the measured H2O concentration may be improved by replacing the HITEMP database with the measured spectra from hot gas cells. H2O reference spectra of better quality will also lead to more accurate CO2 and CO measurements because of the overlapping H2O lines. As a conclusion from the CFD calibration results and experimental validation, the CFD analysis reliably reproduces the key mixing and combustion characteristics in the grate boiler under the operational condition summarized in Table 1.

480

C. Yin et al. / Energy 41 (2012) 473e482

4.2. The CFD-predicted flow and combustion characteristics in the boiler Fig. 7 shows some more results from the CFD analysis in detail. On the vertical mid-plane between the two sidewalls, two relatively large recirculation zones can be identified, which makes an overall “S”-shaped flow pattern in the radiation pass, as shown in Fig. 7(a). From such a flow pattern, one may conclude that the entire furnace volume is reasonably utilized and the mixing in the freeboard is acceptably good. Fig. 7(b) shows that the contour of O2 concentration on the same mid-plane, which indicates an unnecessarily high O2 level in the corner bounded by the grate-end and the rear wall. It implies that the lengthwise PA distribution (especially close to the end of the grate) and the SA on the lower part of the rear wall could be improved based on the residency time of the fuel on the grate and the fuel reactivity. Fig. 7(c) and (d) shows the

gas temperature and O2 fraction on the horizontal cross-section where the measuring ports (E and F) are located. Large variations in the temperature and species can be observed. The results on the horizontal cross-section also indicate that the staggered SA jets on this section may be further improved in terms of jet momentum in order to increase the jet penetration and enhance the mixing. 4.3. Experience and lessons Besides the characterizing and modeling of this 88 MW gratefired boiler in a direct co-firing plant, quite some efforts have been made on other grate-fired boilers, e.g., the 108 MW grate boiler in an indirect, parallel co-firing plant [2]. Here the experience gained and lessons learned are briefly summarized, which could help to improve the modeling or even improve grate-firing technology for biomass combustion.

Fig. 7. (a) Velocity vector colored by gas temperature and (b) O2 mass fraction at the vertical mid-plane; and (c) gas temperature, and (d) mass fraction of O2 at the horizontal crosssection where the measuring ports, E and F, are located.

C. Yin et al. / Energy 41 (2012) 473e482

An improved fabric seal of the grate system is needed for a better grate-firing technology. Take this relatively new and modern 88 MW grate-fired boiler as an example, the leakage air, most of which is believed from the grate system, is estimated to be about 20%wt of the total air supply as indicated in the plant control room. Although the lack of good sealing between the neighboring grates can somehow serve as dilatation joints and the leakage air may also help to control the temperatures of the local mechanical components, an improved fabric seal of the grate system is still needed, which can not only result in a low excess air ratio but also contribute to a better control of the combustion process. A better control of different sections of the grate and gas-fuel mixing in the fuel bed on the grate will increase the char burnout and further improve the fuel flexibility. Generally speaking, the mixing inside the fuel bed in a grate boiler is very poor, which is partly responsible for relatively low efficiency and high emissions from grate-firing systems. Take bubbling fluidized bed combustors as an example, the superficial gas velocity in the fuel bed is typically about 2e3 m/s, which is sufficiently high to maintain the bed in fluidization state (with a high degree of mixing) but low enough to assure that little solid particles are entrained into the freeboard and exit the boiler as fly ash. Comparatively, the gas velocity in the fuel bed in grate boilers is much lower. For instance, the superficial gas velocity averaged over the grate in this 88 MW grate boiler is less than 1 m/s (as shown in Fig. 3). To enhance mixing in the fuel bed on the grate whilst still keep the PA/SA ratio at about 40/60, part of the flue gas may be recycled to the first and second sections of the grate, where the main fuel conversion processes are evaporation and devolatilization. The grate can also be divided into a few sections lengthwise, each of which can be independently controlled in terms of moving style (e.g., vibration frequency and amplitude). In this way, the lengthwise in-bed fuel conversion rates as expected in design can be better secured, or can be better adjusted in case of need. Both the enhanced in-bed mixing and well-controlled lengthwise fuel conversion favor a greater fuel flexibility. An improved on-line fuel and air monitoring system is required, based on which closed loop controls can be more effectively used to adjust the boiler operation to different levels of load, moisture, fuel quality and so on for an optimum performance. Take this 88 MW grate boiler as an example, the actual air flowrate through individual SA nozzles or even a group of SA nozzles is not known, which not only gives uncertainties to the modeling but also imposes difficulties onto the control of the combustion process. In addition, biomass fuels often have inconsistent quality, especially the large variations in moisture content, which needs an effective on-line fuel quality monitoring system in order to make appropriate adjustments in the operation conditions to achieve a good performance. In this 88 MW grate boiler, the operators grab at a fixed frequency some samples from the straw bales before they are transported to the straw rake, and then take the samples to lab for analysis. Such an offline, discontinuous analysis may not be enough for prompt combustion control and optimization. For instance, online monitoring of fuel moisture content by measuring the relative humidity of the flue gas in biomass-fired plants is recently discussed [17]. Further development of advanced staging secondary air system is required to lower CO emissions. Although the integration of advanced secondary air supply systems in grate-firing boilers represents one of the real breakthroughs in this technology in the past years, further development of secondary air systems is still needed to reduce CO emissions from grate boilers to the similar levels as pulverized fuel boilers or fluidized bed combustors. In addition, maintaining a certain fuel bed thickness and minimizing the entrainment of fine fuel particles into the freeboard can also contribute to the reduction of CO emissions from grate boilers.

481

Some of the above measures will be tested in a modern 500 kW grate-fired boiler rig, originally located in Aalborg University and currently moved to Babcock & Wilcox Vølund in Esbjerg. The planned tests are mainly for innovating in the grate system and improving the air supply, with the aim to fire more different biomass feedstocks (i.e., greater fuel flexibility), lower CO emissions and increase the boiler efficiency. 5. Conclusions Efforts on characterizing and modeling of an 88 MW grate-fired boiler burning wheat straw are reported. The CFD results show an acceptable agreement with the experimental characterization. Some modeling issues, for instance, in-bed fuel conversion, representation of volatile gases, char oxidation, soot or fine particle entrainment, and mesh, and their expected impacts on the CFD analysis of grate-fired systems are discussed. The flow and combustion characteristics in the boiler are also discussed on the basis of the CFD results. The overall mixing in the boiler is acceptable. However, the air supply, e.g., the lengthwise distribution of the primary air and the penetration or momentum of some secondary air jets, can be further improved and optimized. Lessons from characterizing and modeling of this 88 MW grate boiler as well as similar efforts on other full-scale boilers are summarized. Among others, improving fabric seal of the grate system, enhancing gas-fuel mixing inside the fuel bed whilst maintaining a certain fuel bed thickness, better controlling different sections of the grate lengthwise, improving on-line fuel and air characterization, and further development of advanced secondary air system are highlighted, in order to facilitate modeling validation effort as well as to improve grate-firing technology. Acknowledgments The research was financially supported by Grant PSO 4792, ‘‘Grate firing of biomass e Measurements, validation and demonstration’’. The first author would also like to thank Jørgen Hansen and Lars Busch from Babcox & Wilcox Vølund and Matthias Mandø, Jens Bo Holm-Nielsen from Aalborg University for their valuable discussions on grate-firing technology, particularly the discussions over new test plans on the 500 kW grate boiler rig which is targeted to make a breakthrough in this technology. References [1] IEA. Biomass co-firing database, Task 32 of the IEA Bioenergy Agreement. Available from: http://www.ieabcc.nl/database/cofiring.php; 2011 [accessed 02.12.11]. [2] Yin C, Rosendahl L, Kær SK, Clausen S, Hvid SL, Hille T. Mathematical modeling and experimental study of biomass combustion in a thermal 108 MW gratefired boiler. Energy Fuels 2008;22:1380e90. [3] Staiger B, Unterberger S, Berger R, Hein KRG. Development of an air staging technology to reduce NOx emissions in grate fired boilers. Energy 2005;30: 1429e38. [4] Bak J, Clausen S. FTIR emission spectroscopy methods and procedures for real time quantitative gas analysis in industrial environments. Meas Sci Technol 2002;13:150e6. [5] Yang YB, Goh YR, Zakaria R, Nasserzadeh V, Swithenbank J. Mathematical modeling of MSW incineration on a traveling bed. Waste Manage 2002;22: 369e80. [6] Kær SK. Numerical modeling of a straw-fired grate boiler. Fuel 2004;83: 1183e90. [7] Yang YB, Newman R, Sharifi V, Swithenbank J, Ariss J. Mathematical modeling of straw combustion in a 38 MWe power plant furnace and effect of operating conditions. Fuel 2007;86:129e42. [8] Albrecht BA, Zahirovic S, Bastiaans RJM, van Oijen JA, de Goey LPH. A premixed flamelet-PDF model for biomass combustion in a grate furnace. Energy Fuels 2008;12:1570e80. [9] Costa M, Dell’Isola M, Massarotti N. Numerical analysis of the thermo-fluiddynamic field in the combustion chamber of an incinerator plant. Energy 2009;34:2075e86.

482

C. Yin et al. / Energy 41 (2012) 473e482

[10] Porteiro J, Collazo J, Patino D, Granada E, Gonzalez J, Miguez JL. Numerical modeling of a biomass pellet domestic boiler. Energy Fuels 2009;23:1067e75. [11] Venturini P, Borello D, Iossa C, Lentini D, Rispoli F. Modeling of multiphase combustion and deposit formation in a biomass-fed furnace. Energy 2010;35: 3008e21. [12] Yu Z, Ma X, Liao Y. Mathematical modeling of combustion in a grate-fired boiler burning straw and effect of operating conditions under air- and oxygen-enriched atmospheres. Renew Energy 2010;35:895e903. [13] Göerner K, Klasen T. Modelling, simulation and validation of the solid biomass combustion in different plants. Prog Comput Fluid Dyn 2006;6:225e34.

[14] Yin C, Kær SK, Rosendahl L, Hvid SL. Co-firing straw with coal in a swirlstabilized dual-feed burner: modeling and experimental validation. Bioresour Technol 2010;101:4169e78. [15] Yin C, Rosendahl L, Kær SK. Chemistry and radiation in oxy-fuel combustion: a computational fluid dynamics modeling study. Fuel 2011;90:2519e29. [16] Froment GF, Bischoff KB. Chemical reactor analysis and design. New York: Wiley; 1979. [17] Hermansson S, Lind F, Thunman H. On-line monitoring of fuel moisturecontent in biomass-fired furnaces by measuring relative humidity of the flue gases. Chem Eng Res Des 2011;89:2470e6.