Insights into radiation property prediction for numerical simulation of pulverized coal/biomass oxyfuel combustion

Abstract

The radiation property of flue gas in pulverized coal/biomass oxyfuel combustion distinguishes obviously from that in air combustion. Moreover, the particle emissivity and scattering factor can vary during its thermal conversion. Both phenomena challenge the accurate prediction of radiative heat transfer in coal/biomass oxyfuel combustion. As one solution to surmount this challenge, a new efficient exponential wide band model + particle radiation property model is proposed in this work. For validation, the proposed model is compared with other parallel models, and their predictions are compared against experimental data. It is found that the efficient exponential wide band model + particle radiation property model can give reliable predictions at different operating conditions. The maximum relative errors between the prediction and the experimental surface incident radiation are within 7% at the burner outlet and 2.5% at the peak point. Moreover, the iteration time consumptions are 0.55 and 0.87 s for the 0.5 MW case, and 35.45 and 35.66 s for the 600 MW case with the parallel radiation model and the proposed model. After validation, the feasibility and characteristics of the coal/biomass oxyfuel co-combustion process in a 600 MW tangentially fired boiler are predicted. It is found that the boiler can run stably and the CO₂ mass fraction in the discharged flue gas can be around 90%.

Graphical Abstract

Open in new tabDownload slide

1. Introduction

Climate change due to anthropogenic greenhouse gas (GHG) emissions adversely affects the ecology and environment, so reducing CO₂ emissions is currently a primary priority [1]. In this scenario, many efforts have been put into carbon capture and sequestration (CCS) during coal combustion which provides the largest source of CO₂ emission [2]. Therefore, oxyfuel combustion of blended coal/biomass is an efficient and feasible CCS technology. By mixing oxygen and the recirculated flue gas, the gas composition in oxyfuel combustion differs significantly from that produced by conventional combustion technologies [3, 4]. This difference can further lead to distinct flue gas radiation properties. However, most of the current radiation models are only well established for air–fuel combustion and have proven to be unsuitable for oxyfuel combustion [5, 6]. In addition, for pulverized coal/biomass oxyfuel combustion, the particle radiation property can also vary during its thermal conversion. This variation in particle composition during combustion can also affect radiation [7–9].

So far, many models for predicting the gas and particle radiation properties have been proposed by researchers all over the world. The line-by-line (LBL) model and the narrow-band model (NBM) can provide accurate solutions for the gas radiation properties [10]. However, the high computational intensity makes them unsuitable for application in the simulation of industrial-scale facilities. To improve the efficiency of LBL model, Xie [11] presented a machine learning-based LBL absorption coefficient calculation method for CO₂ in the applications of atmospheric remote sensing. It was found that the model was compact, accurate, and efficient. Compared with the LBL model and the band models, the global model attempts to calculate the radiation absorption coefficient directly and is computationally efficient [10]. Therefore, the weighted sum of gray gases (WSGG) model first proposed by Hottel [12] is widely used. With the WSGG model, the gas radiation absorption coefficient is easily calculated according to the molar ratio of H₂O and CO₂. Based on the original WSGG model, Cai [13] proposed a new WSGG model to predict the oxyfuel combustion of semi-coke. It was found that the proposed model was more accurate when calculating the radiation characteristic parameters of the gas mixture for the semi-coke combustion. Besides the WSGG model, Wang [14] compared and refined the various full-spectrum k-distribution and spectral line WSGG models for nonhomogeneous media. It was found that the equivalent spectral line WSGG schemes were generally somewhat less accurate than their full-spectrum k-distribution counterparts due to their low-order spectral integration scheme. Compared with LBL and NBM methods, the EWBM first developed by Edwards [15, 16] is relatively less computationally intensive with a slight loss of accuracy. Compared with the WSGG model, the EWBM can readily calculate the radiation properties for gas mixtures with multiple radiative species at arbitrary molar fractions. To further improve the EWBM accuracy and efficiency, Yan [17] modified the original band intensity in EWBM using the LBL method based on the HITEMP 2010 database, and the table look-up method was introduced to develop the efficient EWBM (E-EWBM). Yin [18] also successfully applied E-EWBM to CFD simulations about oxyfuel combustion. It was found that the E-EWBM prediction is similar to the WSGG prediction and mentioned that the accuracy of the E-EWBM may be further improved by refining some of the model parameters. Besides the effects of the gas phase on radiation, the radiation in the flame zone can also be dominated by the char and ash particles. Most coal/biomass combustion simulations simplify the particle emissivity by taking an empirical constant of 0.7–0.9 [8, 19–21]. The particle scattering factor is simplified by choosing the empirical constant of 0.6 or 0.9 [22, 23]. However, the particle radiation properties can actually vary with the conversion of particles during combustion [24]. In terms of CFD simulations of oxyfuel combustion of pulverized fuel, Nakod [25] compared the gray and nongray radiation models for oxyfuel combustion of coal. To consider the variation of coal particle radiation property during combustion, the particle emissivity and particle scattering factor of coal, char, and ash are given different values. Huynh [26] studied the effect of particle radiation on the NO prediction in a pilot-scale facility for both air and oxyfuel combustion of coal. The WSGG model is used to predict the gas absorption coefficient while the linear equation is used to predict the variations of coal particle emissivity and particle scattering factor during combustion. Buckius and Hwang [27] developed a model to calculate the extinction/absorption coefficients and the asymmetry factor for polydispersions and applied the model to calculate the radiation properties of coal according to the particle size distribution. Cai et al. [28] studied the radiation property for gas–particle mixture of pulverized coal flames. A new radiation model that can account for nongray gas–solid mixtures is developed for pulverized coal flames. The new models are validated against a pulverized coal ignition flame. Wang et al. [29] developed a full-spectrum correlated k-distribution look-up table method for radiative transfer in nonhomogeneous participating media with gas–particle mixtures. The particle radiation property was treated based on Buckius and Hwang’s work.

From the literature review, it is known that research on the gas and particle radiation properties in numerical simulation of pulverized coal/biomass oxyfuel combustion is still not sufficient [30, 31]. As one contribution to this field, a new efficient exponential wide band model+particle radiation property model (E-EWBM+PRPM) is proposed in this work to calculate both the gas and particle radiation properties in oxyfuel combustion of pulverized coal/biomass. The E-EWBM model [17] is used to predict the absorption coefficient of gas mixtures including H₂O, CO₂, CO, and CH₄. The coal particle emissivity and particle scattering factor are predicted by PRPM. The results of the simulations using E-EWBM+PRPM are validated against experimental data from a 0.5 MW aerodynamically air-staged burner. After validation, the coal/biomass oxyfuel co-combustion in a 600 MW tangentially fired boiler is predicted.

2. Materials and methods

The commercial software Fluent is used to do the numerical simulation and Visual Studio C++ software is used to compile the user-defined function (UDF) about the E-EWBM+PRPM model. The gas–particle flow is simulated numerically using the Euler–Lagrange method, and the velocity–pressure coupling is solved by the semi-implicit method for the pressure-linked equation (SIMPLE) algorithm. The second-order upwind scheme is employed to discretize the convection terms of the equations. The realizable k–ε turbulence model which can be well applied to swirl combustion is adopted [32]. The radiative transfer equation (RTE) is solved using the discrete ordinates (DO) model in the simulation [33]. The gas and particle radiation properties in the DO RTE are predicted with E-EWBM+PRPM proposed in this work. The discrete phase model is used to track coal/biomass particles with the Lagrange approach. The particle-turbulence dispersion is considered by the discrete random walk model with 10 tries. A total number of 100 particle streams are tracked for each simulation. The residuals of the energy and radiation equations are controlled to below 1e−6, and residuals of the other equations are controlled to below 1e−3.

For flue gas radiation property, the E-EWBM is used to calculate the effective gas absorption coefficient, $αr$⁠, in this work [17]. The Planck mean absorption coefficient $αp$ and the total emissivity $εt$ are calculated by E-EWBM according to Equations (1) and (2) [34, 35]. The effective absorption coefficient, $αr$⁠, of the gaseous phase is calculated with the E-EWBM model using Equation (3). Although $αp$ is not directly used in the effective absorption coefficient calculation, it is one of the most important radiation parameters and is used to adjust the reference E-EWBM parameters:

where $αη$ denotes the absorption coefficient contributed by the nth band; $I¯bΔη,n$ is the mean block intensity of the nth band.

where $βΔη,m$ is the mth block transmissivity; it is the products of all the band transmissivities that belong to this block; $ηL,m$ and $ηU,m$ are the lower and upper limits of block m, respectively, and they are determined according to the band-center wavenumber and the calculated nth bandwidth of species i; L is the domain-based path length; $F(η/T)$ is the fractional function of blackbody radiation and can be calculated by the following:

where ξ denotes C₂η/T; C₂ is the Planck second radiation constant; and T is the actual gas temperature.

For the particle radiation property, the variations of particle emissivity and particle scattering factor during the coal conversion process are calculated with PRPM according to the fractions of unburned char and ash as determined by Equations (5) and (6) [24]:

where U_C and U_VM,C are the fractions of unburned char and unburnt combustibles, respectively. The initial values of U_C and U_VM,C are the mass fractions of fixed carbon and volatile of coal and biomass.

The effective absorption coefficient used in the DO RTE can be directly calculated with the E-EWBM model. The equivalent particle absorption coefficient and the equivalent particle scattering factor used in the DO RTE can be calculated with Equations (7) and (8). The proposed gas and particle radiation property model, E-EWBM+PRPM, is then compiled to couple with ANSYS Fluent by UDF:

where ε_pn, f_pn, and A_pn are the emissivity, scattering factor, and projected area of particle n. The summation is over N particles in volume V.

3. Model validation

To fully validate the numerical methods in this work, the model predictions are compared against two sets of reported experimental data in terms of the surface incident radiation (SIR) from oxyfuel combustion of coal (Case I) and oxyfuel combustion of blended coal/biomass (Case II).

3.1 Description of reported experiment condition

The reported experiments were carried out in a 0.5 MW aerodynamically air-staged burner [36]. The device was operated by mixing O₂ with flue gas at different proportions to represent different actual operating conditions. A detailed structure of the burner can be found in the literature [36].

The primary air conveys the pulverized coal (or blended coal/biomass) to the burner from the inner inlet. The secondary air enters the burner from the outer ring inlet with a swirl velocity. The swirl number (S) of the secondary flow is 0.6, calculated by Equation (9) [37]. When the flow rate and the swirl number are known, the swirl velocity can then be calculated according to Equation (9). The temperatures of the primary and secondary streams are kept at 343 and 543 K, respectively. The O₂ mole fraction in the primary stream is kept at 21%, while the O₂ concentration in the secondary stream is adjusted according to the recycled flue gas flow rate. The convective heat transfer coefficient of the refractory is assumed to be 14 W/(m²∙K), and the ambient temperature is 300 K, and the wall emissivity is set as 0.85 [38–40].

The analyses of coal and biomass for Case I and Case II are shown in Table 1 [36]. The details of the inlet mass flow rate according to the recycle ratios (RR) are presented in Table 2 (Case I) and Table 3 (Case II) [36].

where $vtva$ is the ratio of tangential velocity to axial velocity of secondary air; $τ0$ and $τ0$ denote the radius of inner and outer vortex generators, respectively. S is then defined as the ratio of the axial flux of angular momentum to the axial flux of momentum.

For Case I, the pyrolysis rate is calculated by two competing rate models, Equations. (10), according to the literature [41], and the corresponding reaction kinetic parameters are given in Table 4.

where α_I denotes the initial volatile mass fraction in coal; α_II is set to 1; m_V(t) is the volatile yield up to time t; m_P is the mass of coal particle; R denotes the gas constant; T is the particle temperature. For the blended coal/biomass combustion, the pyrolysis rate is obtained by the single rate model according to literature [42], and the corresponding kinetic parameters are given in Table 5. Note that the reaction kinetics for the coal combustion and the blended coal/biomass combustion can be different, so different pyrolysis models are chosen according to the literature.

The gas-phase turbulence–chemical interaction is calculated with the Finite Rate/Eddy Dissipation Model (FR/EDM). With FR/EDM, the Arrhenius rate and the vortex dissipation reaction rate are limited by chemical kinetics and turbulent mixing, respectively, and the net reaction rate is the lower value of the two rates. This work considers a two-step reaction mechanism with acceptable accuracy in predicting O₂, CO₂, and H₂O concentrations [42]. The kinetic rates of reactions R1 for the first-step coal volatile combustion, R2 for the first-step biomass volatile combustion, and R3 for the second-step volatile combustion are calculated by Equations (11) and (12) [42].

where, k_r is the rate constant calculated by the Arrhenius equation, and the corresponding kinetic parameters are listed in Table 6.

Following pyrolysis and volatile combustion, the residual char reacts with surrounding O₂ and CO₂ shown by reactions (R4) and (R5). The char combustion rate is modeled with a multiple-surface-reaction model in FLUENT [42], and the kinetic parameters of particle surface reactions are listed in Table 7.

3.2 Comparison with experimental data

To balance the simulation economy and the simulation accuracy, the mesh independence analysis is done. Three meshes with 12 205, 22 285, and 33 699 quadrilateral cells are generated, and the comparisons of temperatures against the radial coordinate at 0.35 m from the burner are shown in Fig. 1. Although there are differences between the predictions from the three meshes, the results generated from the latter two meshes are close. Thereby, the second mesh with 22 285 cells is chosen for the following calculations, as shown in Fig. 2.

Then, simulation results with four different gas–particle radiation methods including WSGG+PRPM, WSGG+Constant (particle radiation property), E-EWBM+PRPM, and E-EWBM+Constant (particle radiation property) are compared against the experimental data for Case I, as shown in Fig. 3. The WSGG model used in the Fluent software was built by Smith [43]. When the constant particle emissivity and particle scattering factor are used, the values are set to 0.9 as recommended by literature [20]. Besides the SIR distribution comparison, the predicted adiabatic flame temperature of Case RR65%, Case RR72%, and Case RR75% are also compared with experiment data. The adiabatic flame temperatures for the three cases are about 2350, 2050, and 1900 K. In comparison, the calculated adiabatic flame temperatures with E-EWBM+PRPM for the three cases are 2325, 2023, and 1933 K, close to the reported data. In addition, the iteration time consumptions for Case RR65% with WSGGM+PRPM and E-EWBM+PRPM are counted to be 0.55 and 0.87 s, respectively.

From Fig. 3, it can be seen that the simulations using WSGGM + Constant and the E-EWBM+Constant methods are inclined to overestimate SIR when considering the constant particle radiation property. When considering the variation of particle radiation property in radiative heat transfer, both simulations using WSGGM+PRPM and the E-EWBM+PRPM can well predict the peak SIR for different flue gas RR. The prediction errors of the WSGGM+PRPM and the E-EWBM+PRPM for the oxyfuel combustion cases are summarized in Table 8. For Case Air, the prediction errors for the SIR from E-EWBM+PRPM and WSGGM+PRPM are quite close to each other. This is because the WSGG model was originally designed for conventional air–fuel combustion. At positions far from the burner, E-EWBM+PRPM can also give better SIR prediction. This is because, in the near burner zone, the concentrated and high-temperature particle radiation dominates, leading to similar SIR peak prediction with WSGG+PRPM and E-EWBM+PRPM. At positions far from the burner, the particle concentration is lowered and the gas radiation becomes evident. It can be seen from Panel (d) that the predictions from the WSGG+PRPM model are close to the predictions from the E-EWBM+PRPM model. This is because the WSGG model was originally designed for conventional air–fuel combustion. It can then be concluded that considering the particle radiation variation with the variation of coal composition is essential. The E-EWBM+PRPM can be more accurate than WSGG+PRPM when predicting the gas radiation absorption coefficient for oxyfuel combustion.

Finally, simulation results with E-EWBM + PRPM are further compared against the experimental data for Case II, as shown in Fig. 4. It can be seen that simulations using E-EWBM+PRPM can predict well the peak SIR for different biomass blending ratios and different flue gas RR. At the burner outlet where the axial coordinate is close to 0 m, the prediction errors for Panels (a), (b), and (c) are 0.61%, 0.39%, and 0.20%, respectively. The prediction errors for the peak SIR of Panels (a), (b), and (c) are 0.58%, 1.41%, and 0.59%, respectively. At 3.2 m far from the burner outlet, the prediction errors for Panels (a), (b), and (c) are about 16.67%, 5.88%, and17.86%, respectively. The SIR values are underestimated. This is mainly because the constant wall thermal boundary condition leads to lower gas temperature at locations far from the burner. In addition, the flue gas thermophysical property can also influence the gas temperature and further influence the SIR distribution.

4. Oxyfuel combustion of coal/biomass in a 600 MW boiler

After model validation, the feasibility and property of oxyfuel combustion of blended coal/biomass in a 600 MW tangentially fired boiler are numerically studied. The effects of the biomass blending ratio and the O₂ molar fraction in the oxidant mixture are analyzed.

4.1 Description of the 600 MW tangentially fired boiler

The geometry and mesh of the 600 MW tangentially fired boiler are shown in Fig. 5a [44]. The specific arrangement of the burners at the four corners is shown in Fig. 5b. In Fig. 5c, the diameter of the tangent circle formed by the injection of the primary and secondary burner nozzles at the No. 1 and No. 3 corners is ϕ1458 mm. The diameter of the formed tangent circle at the No. 2 and No. 4 corners is ϕ1882 mm. The rotation direction of the tangent circle formed by the primary air, secondary air, and oil wind is counterclockwise. The rotation direction of the tangent circle formed by the overfire air (OFA) is clockwise. To ensure the stability of the numerical calculation results, the grid of the generated burner area is encrypted, as shown in Fig. 5d. The analyses of coal and biomass used in this section are shown in Table 1.

4.2 Analyses of the 600 MW tangentially fired boiler

Orthogonally combining different biomass blending ratios and different O₂ mass fractions, a total of 16 tests are simulated in this section, as listed in Table 9. The velocity field, temperature field, and incident radiation field of the 16 tests are illustrated in Figs 6–8. The CO₂ mass fractions in the outlet flue gas of the 16 tests are summarized in Table 10.

From Fig. 6, it can be seen that the swirl velocity is generated in the burner zone for all the tests, which enhances the mixing of gas and particle fuels. The velocity near the furnace wall is high, which can denudate the deposited ash and slag and enhance the convection and radiation heat transfer between the flame and the water wall. The velocity magnitude at the furnace outlet is greatly lowered by the platen superheater and reheater, which can mitigate the erosion of fuel particles on the heat exchangers. The biomass blending ratio has a slight effect on the velocity field while increasing the O₂ fraction in the oxidant can significantly lower the velocity magnitude in the furnace. This is because increasing the O₂ fraction can directly decrease the oxidant flow rate.

From Fig. 7, it can be seen that tangentially arranged torches are formed stably near the burner for all the 16 tests. The gas temperatures are higher near the water wall, which can enhance the heat transfer in the furnace. The gas temperature becomes homogeneous at the furnace outlet, which can ensure that the superheaters are heated homogeneously. Increasing the biomass bending ratio, the temperature field varies slightly. The temperature in the furnace can be significantly increased by increasing the O₂ fraction in the oxidant. This is because when the O₂ fraction is increased, the combustion rate in the furnace can be increased directly, and the heat release rate can then be increased.

From Fig. 8, it can be seen that the SIR mainly concentrates on the water wall in the furnace and the platen superheater at the furnace outlet. The comb-like SIR distribution is mainly caused by the layered distribution of the torch from the tangentially fired burner. Increasing the biomass blending ratio, SIR varies slightly. Increasing the O₂ fraction in the oxidant, SIR can be significantly increased. This is because the gas temperature can be obviously increased by the O₂ fraction increment.

From Table 9, it can also be seen that the CO₂ mass fraction in the outlet flue gas of all the tests is high, which can greatly lower the CO₂ compression energy intensity. Note that the CO₂ mass fraction is not measured in the dry flue gas. The CO₂ mass fraction in the outlet flue gas gradually decreases with the O₂ mass fraction increment, because the CO₂ mass fraction in the inlet oxidant is decreased. The CO₂ mass fraction gradually decreases with the biomass blending ratio increment, because the carbon content in biomass is lower than that in the coal.

6. Conclusions

A new E-EWBM+PRPM model is proposed in this work to predict the radiation property of gas and particles in pulverized coal/biomass oxyfuel combustion. The E-EWBM+PRPM is validated against experimental data, and compared with other parallel radiation models including the E-EWBM+Constant, the WSGG+Constant, and the WSGG+PRPM. Finally, the coal/biomass oxyfuel co-combustion process in a 600 MW tangentially fired boiler is predicted. With this work done, the main findings can be drawn as follows:

1. The E-EWBM + PRPM model is advanced in predicting the gas and particle radiation properties for coal/biomass oxyfuel combustion. With the E-EWBM+PRPM model, the relative errors between the prediction and the experimental SIR are within 7% at the burner outlet and 2.5% at the peak point for different operation conditions.

2. The iteration time consumptions for the 0.5 MW case with WSGGM+PRPM and E-EWBM+PRPM are 0.55 and 0.87 s, respectively, while the iteration time consumptions for the 600 MW case with WSGGM+PRPM and E-EWBM+PRPM are 35.45 and 35.66 s, respectively.

3. For the 600 MW boiler, it is found that the biomass blending ratio has slight effects on the velocity, temperature, and incident radiation in the furnace. Increasing the O₂ fraction in the oxidant can significantly decrease the gas velocity while increasing the gas temperature and SIR in the furnace.

4. The CO₂ mass fraction in the discharged flue gas can be around 90% for the oxyfuel combustion of coal/biomass in a 600 MW tangentially fired boiler. The CO₂ mass fraction decreases gradually with the increments of the biomass blending ratio in the fuel and O₂ fraction in the oxidant.

Author contributions

Linbo Yan (Conceptualization [Equal], Methodology [Equal]), Yang Liu (Validation [Equal], Visualization [Equal]), Kexin Li (Software [Equal], Writing—original draft [Equal]), Cong Geng (Investigation [Equal]), and Boshu He (Conceptualization [Equal], Writing—review & editing [Equal])

Conflict of interest statement

None declared.

Funding

This article received financial supports from the State Key Laboratory of Intelligent Construction and Healthy Operation and Maintenance of Deep Underground Engineering, China University of Mining & Technology, Xuzhou (SKLGDUEK2216) and National Natural Science Foundation of China (12102456).

Data availability

The data that support the finds of this study are available from the corresponding author upon reasonable request.

References