Large eddy simulation of a 660 MW utility boiler under variable load conditions

Abstract

Large eddy simulation (LES) has become a promising tool for pulverized coal combustion with the development of computational fluid dynamics (CFD) technologies in recent years. LES can better capture the unsteady features and turbulent structures of coal jet flame than Reynolds averaged Navier Stokes (RANS). The coal-fired power plants in China are now required to be operated in a wide load range and quickly respond to the electric grid. The boiler performance of variable loads should be evaluated in terms of flow, heat transfer, and combustion processes. In this paper, LES was applied to simulate a 660 MW ultra-supercritical boiler under BMCR (boiler maximum continue rate), 75%THA-100, and 50%THA-100 conditions. The predicted gas velocities agree well with the thermal calculation and the temperature error is less than 130 K. The simulation results show that the operation load has significant effects on the boiler performance. It is also proved that LES can provide guidance for the design and operation of advanced coal-fired boilers.

Graphical abstract

Keywords

large eddy simulation / ultra-supercritical boiler / operation load

1 Introduction

The electricity generated by coal-fired power plants makes up of more than 70% of the total electricity production in China [1]. In the past decades, the availability hours of generating equipment have dropped to less than 4000 h per year. This means that most of the power plants are operated at low loads. According to the Paris Agreement, China has pledged to decrease the carbon emission per unit GDP by 40%–45% and increase the ratio of renewable energy consumption to 15% in 2020 compared with 2005. However, the renewable energy such as wind and solar power is highly unstable so that the operation safety of grid is threatened. Therefore, coal-fired power plants are required to adjust the load dynamically to offset the volatility of the renewable energy. The transition to a more flexible power system needs to thoroughly understand the boiler performance under wide load conditions. Computational fluid dynamics (CFD) is a useful tool to study the flow, heat transfer, and combustion processes in utility boilers. CFD can be classified into three levels which are Reynolds averaged Navier Stokes (RANS) computations, large eddy simulations (LES), and direct numerical simulations (DNS) [2]. RANS techniques are developed to solve the mean values of all quantities, which require turbulent models for unclosed terms. DNS solves the full instantaneous Navier-Stokes equations without any model for turbulent motions. Luo et al. investigated the pulverized coal jet flame in a heated co-flow by means of DNS [3,4]. However, it is only limited to academic problems due to tremendous computational costs. LES explicitly calculates the turbulent large eddies and models the small-scale eddies. Chen and Ghoniem found that LES better captured the pulverized coal jet flame and provided insights into the flame stabilization mechanism [5].

Many researchers investigated the effects of operating conditions on tangentially coal-fired boilers due to their widespread application. Belosevic et al. investigated the influence of turning off the burners, the air/fuel ratio, and the boiler load [6]. The outlet average flue gas temperature and carbon-dioxide content decreased with the boiler load reduction from the full load. Turning off the burners to achieve load decrease would change the overall flow and temperature field. Belosevic et al. further studied the flame geometry and position in order to optimize the operation parameters of a utility boiler [7]. It was recommended to supply more fuel through the upper- than the lower level of the main burners to avoid excessively low position of the flame. Al-Abbas and Naser explained the effects of boiler loads on the combustion processes [8]. When the operation load is at full load, the coal reaction with oxidizers can be improved due to the increased residence time in the combustion zone. When the operation load is above the full load, the higher aerodynamics and oxygen concentrations lead to the improved mixing and reaction conditions. Tian et al. found that the CO concentration dramatically decreased with the operation load changing from 100% to 50% and the NO_x concentration at the outlet was not sensitive to the operation load [9]. The simulation results were validated by the experimental measurements. However, Belošević et al. found a negative correlation between the NO_x concentration and the boiler load [10]. Dong et al. studied the wall temperature distributions in an ultra-supercritical boiler [11,12]. The wall temperatures in full load condition are much higher than those in the 75% and 50% load conditions. Liu et al. investigated the effects of flue gas recirculation injection position on the temperature distribution, NO_x emission, and gas temperature deviation under boiler maximum continuous rating (BMCR), 75% turbine heat acceptance (THA), and 50%THA load conditions [13]. The boilers investigated in previous studies had different arrangements of burners and heating panels were operated under various conditions. Thus, some of conclusions seem to be contradictory and are lack of generality.

In this paper, a 660 MW ultra-supercritical boiler was simulated by LES under a wide range of operation load conditions. Besides, the particle phase was tracked by a simplified direct quadrature method of moments (DQMOM) in the Eulerian framework [14]. Moreover, the simulation results were compared with the thermal calculation and they were found to be in good agreement. Furthermore, the effects of operation load on the flow, heat transfer, and combustion processes were discussed. This study is helpful to the design and operation of utility boilers.

2 Utility boiler investigated

The 660 MW ultra-supercritical tangentially fired boiler, which was built in Suqian, Jiangsu province, China, underwent a commissioning test at the end of 2018. It adopted a tower-type furnace and a double-reheat steam cycle in order to improve the efficiency. The steam parameters of 31 MPa/600°C/620°C/620°C were first realized in this boiler. Figure 1 shows the structure of the boiler and the arrangements of burners and heating panels.

There are six layers of pulverized coal burner at each corner. Two layers of over-fire air with three inlets each are installed over the main burner zone. An imaginary circle which rotates clockwise is formed in the center of the furnace as depicted in Fig. 2. The deflected air between two adjacent coal inlets deviates from the direction of primary air and is closer to the wall. The recirculation flue gas is injected into the ash hopper to adjust the gas temperature.

Table 1 lists the designed coal used in the thermal calculation. The coal is ground into fine particles with R90 in the range of 18%–20%. The uniformity index of coal particles is between 1.0 and 1.1. Table 2 presents the operating parameters of the 100%, 75% and 50% loads which are simulated in this study.

3 Numerical models

A LES tool named Arches was used to simulate the utility boiler. Arches solves the conserved quantities (mass, momentum, energy, and scalar) in a turbulent flow field by the massively parallel computing method. It has been applied to many industrial problems including solid fuel combustion and gasification [15–17].

3.1 Governing equations of gas phase

The gas conservation equations are written in a finite volume form including mass balance, momentum balance, mixture fraction balance, and energy balance equations.

(1){\displaystyle \frac{\partial \overline{\rho }}{\partial t}}\mathrm{+}{\displaystyle \frac{\partial }{\partial x_i}}\left(\overline{\rho }{\tilde{u}}_i\right)\mathrm{=}{\overline{S}}_m,

(2){\displaystyle \frac{\partial \left(\bar{\rho }{\tilde{u}}_i\right)}{\partial t}}+{\displaystyle \frac{\partial \bar{\rho }{\tilde{u}}_i{\tilde{u}}_j}{\partial x_j}}={\displaystyle \frac{\partial }{\partial x_j}}[{\bar{\tau }}_{ij}-\bar{\rho }(\widetilde{u_i{u}_j}-{\tilde{u}}_i{\tilde{u}}_j\left)\right]-{\displaystyle \frac{\partial \bar{p}}{\partial x_i}}+\bar{\rho }{g}_i+{\bar{S}}_{u_i},

(3){\displaystyle \frac{\partial (\bar{\rho }\tilde{h})}{\partial t}}+{\displaystyle \frac{\partial }{\partial x_i}}\left(\bar{\rho }{\tilde{u}}_i\tilde{h}\right)={\displaystyle \frac{\partial }{\partial x_i}}[\lambda {\displaystyle \frac{\partial \bar{h}}{\partial x_i}}-\bar{\rho }(\widetilde{u_{i}h}-{\tilde{u}}_i\tilde{h}\left)\right]-{\displaystyle \frac{\partial {\bar{q}}_i}{\partial x_i}}+{\bar{S}}_h,

(4){\displaystyle \frac{\partial (\bar{\rho }\tilde{\eta })}{\partial t}}+{\displaystyle \frac{\partial \bar{\rho }{\tilde{u}}_i\tilde{\eta }}{\partial x_i}}={\displaystyle \frac{\partial }{\partial x_i}}[\bar{\rho }D{\displaystyle \frac{\partial \tilde{\eta }}{\partial x_i}}-\bar{\rho }(\widetilde{u_i\eta }-{\tilde{u}}_i\tilde{\eta }\left)\right]+{\bar{S}}_{\eta },

where S represents the source term, the bar and tilde over variables denote filtered and Farve-filtered operators respectively, andh is the mixture fraction used to determine the thermochemical state of the flow field.

3.2 Simplified DQMOM model

A simplified DQMOM model is used to track the particle phase in the Eulerian framework. The particle phase can be represented by a number density function (f) which describes the number of particles per volume and varies with space (\overrightarrow{x}), time (t), and internal coordinates (\overrightarrow{\xi }). Seven internal coordinates are used to represent the intrinsic characteristics of coal particles, which are raw coal mass, char mass, particle diameter, particle enthalpy, and three components of particle velocities [18]. DQMOM solves the transport equations of weights (ω_α) and weighted abscissas ({\zeta }_{n\alpha }={\omega }_{\alpha }{\langle {\xi }_n\rangle }_{\alpha }) as expressed in Eqs. (5) and (6).

(5){\displaystyle \frac{\partial }{\partial t}}({\omega }_{\alpha })+{\displaystyle \frac{\partial }{\partial x_i}}({\langle u_i\rangle }_{\alpha }{\omega }_{\alpha })-{\displaystyle \frac{\partial }{\partial x_i}}({\mathrm{\Gamma }}_{x_i,\alpha }{\displaystyle \frac{\partial }{\partial x_i}}({\omega }_{\alpha }))=a_{\alpha },

(6){\displaystyle \frac{\partial }{\partial t}}({\varsigma }_{n\alpha })+{\displaystyle \frac{\partial }{\partial x_i}}({\langle u_i\rangle }_{\alpha }{\varsigma }_{n\alpha })-{\displaystyle \frac{\partial }{\partial x_i}}({\mathrm{\Gamma }}_{x_i,\alpha }{\displaystyle \frac{\partial }{\partial x_i}}({\varsigma }_{n\alpha }))=b_{n\alpha },

where a_α and b_nα are the unknown source terms which are solved by a set of linear equations. Solving the linear equations can be avoided by assuming no birth and death of particles and no dispersion terms between different particle groups. Thus, the source terms can be obtained without solving the linear equations. Detailed explanation can be found in Ref. [14].

(7)a_{\alpha }=\mathrm{0},

(8)b_{n\alpha }={\omega }_{\alpha }{G}_{n,\alpha }.

3.3 Coal models

As for pulverized coal combustion, the particles are highly dispersed so that any particle-particle interactions are small enough to be omitted. Only drag force and gravity are included in the single particle velocity model and other forces can be neglected for micro-sized particles [18]. The devolatilization equations combines the single rate reaction model and the distributed activation energy model. The model parameters are defined by comparing with the CPD data [19]. Three reactions given by Eqs. (9) – (11) are used to model the char oxidation process [18]. An ash deposition with the wall heat transfer model is also incorporated into the code [20]. To predict the NO_x distribution, a semi-empirical NO_x model is applied to quantify NO homogeneous reduction by the concentration of CO+ H₂ [21].The NO_x model was validated by air-staged combustion experiments in an electric heated down-fired furnace.

(9)\mathrm{C}(s)+\mathrm{0.5}{\mathrm{O}}_{\mathrm{2}}\to \mathrm{CO,}

(10)\mathrm{C}(s)+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\to \mathrm{CO}+{\mathrm{H}}_{\mathrm{2}},

(11)\mathrm{C}(s)+{\mathrm{CO}}_{\mathrm{2}}\to \mathrm{2}\mathrm{CO}\mathrm{.}

3.4 Simulation setup

In consideration of the computational costs, only a portion of the furnace (from the ash hopper to the high-temperature super-heater) was chosen as the simulation domain as demonstrated in Fig. 3 where the X axis refers to the vertical direction along the boiler height while Y, and Z axes denote the horizontal directions. The uniform cubic meshes with a length of 0.1 m were used across the simulation domain. The total mesh number was nearly 24 million. Although the mesh resolution was not high enough for near-wall flows, LES could provide more reliable results because the large eddies dominated the flow field in the boiler. Three cases with operation loads of BMCR, 75%THA-100, and 50%THA-100 were simulated and the load decrease was achieved by turning off burners. The excess air ratios were 1.15, 1.17, and 1.26 respectively. Each case was run for 168 h on 1421 processors. The total physical time was more than 20 s which was long enough to reach a steady-state. Five particle sizes were used to represent the continuous particle size distribution. Table 3 gives the initial particle size distribution set in the simulations. The particle sizes are defined by the Gaussian quadrature theory to ensure conserved moments [22].

4 Results and discussion

Figure 4 shows that the mass-averaged gas temperature at the inlet of the low-temperature super-heater changes over time. The gas temperature reaches a steady-state after 10 s. In the following analysis, the variables are all averaged over a time period of 3 s. To validate the numerical models, the simulation results are compared with the thermal calculation under different operation load conditions. The vertical gas velocity distributions along the furnace height scaled by the furnace width are presented in Fig. 5. They increase with the operation load because the total air flow rate increases. It is found that the LES results agree well with those of thermal calculation which are denoted by points. The gas temperature plays an important role in the heat transfer process. Figure 6 exhibits the gas temperature profiles of different operation loads. In the main burner zone, the gas temperatures increase dramatically and the distributions are quite similar. However, the gas temperatures at the 50% load are higher than those at the 100% and 75% loads. The reason for this is that the excess air ratio of the 50% load is larger as tabulated in Table 4. The gas temperature keeps rising at the OFA burnout zone except for the 50% load. The pulverized coal is fully burnt off along with more heat released from combustion when extra air is injected into the furnace through the OFA burners. The maximum error of gas temperatures between LES and thermal calculation is 127 K.

Figure 7 displays the tangential velocity profiles along the centerlines of the D layer scaled by the furnace width. The differences of the velocity profiles among three operation loads are relatively small. The diameter of the real velocity circle is much larger than that of the imaginary circle. Figure 8 shows the gas temperature profiles on the same cross-section as Fig. 7. These curves take on the distribution of “saddle shape.” The high-temperature zones are very close to the furnace wall. The diameter of the temperature circle is even larger than that of the velocity circle.

Figure 9(a) shows the species distributions along the furnace height under different operation load conditions. O₂ is fully consumed in the main burner zone while its mass fraction increases to 4% over the OFA zone. When the operation load decreases, the CO mass fractions decrease and CO₂ mass fractions increase. This is because the excess air ratio at the main burner zone is larger for a lower operation load. Figure 9(b) presents the NO profiles in the furnace. The NO distributions are quite similar for BMCR and 75%THA-100 cases. When the operation load is as low as 50%THA-100, the NO mass fractions are relatively high in the main burner zone and decrease to a low level in the OFA zone. The design value of NO_x concentration ranges from 93 ppm to 142 ppm which is consistent with LES results. These species distributions will be further validated by experimental data.

Figure 10 shows that the wall heat flux is extremely high near the burner port while it is uniformly distributed above the main combustion zone. The area where the wall heat flux is larger than 300 kW/m² shrinks when the operation load decreases from 100% to 50%. Figure 11 quantitatively presents the time-average wall heat flux distributed along the furnace height. The distribution of the 50% load is lower than those of the 75% and 100% loads in the main combustion zone. The wall heat flux of BMCR peaks in the OFA burnout zone while it is not obvious for the other two cases.

5 Conclusions

LES has been applied to study the effects of the operation load on the flow, heat transfer, and combustion processes of a 660 MW ultra-supercritical boiler. Three operating conditions including BMCR, 75%THA-100, and 50%THA-100 have been simulated. The predicted gas velocities agree well with the thermal calculation and the maximum temperature error is less than 127 K. The gas velocity distributions on the horizontal cross-section are similar among three operation loads while the vertical gas velocities distributed along the furnace height decline with the decrease of operation load. The gas temperatures are little affected by the operation load in the main burner zone. However, the area where the wall heat flux is larger than 300 kW/m³ shrinks along with the operation load declining. As for the species distributions, CO mass fractions decrease and CO₂ mass fractions increase when the operation load decreases. The NO concentration of the 50% load is higher in the main combustion zone and lower in the OFA zone compared to the 75% and 100% loads. The species distributions should be further validated by the experimental data. The simulation results show that the operation load has significant effects on the boiler performance. It is proved that LES can provide guidance for the design and operation of advanced coal-fired boilers.

References

[1] China Electricity Council. Annual Statistics of China Power Industry 2018. 2019–6-10

[2] Poinsot T, Veynante D. Theoretical and numerical combustion. 2nd ed. USA: RT Edwards, Inc., 2005

[3] Luo K, Wang H, Fan J, Yi F. Direct numerical simulation of pulverized coal combustion in a hot vitiated co-flow. Energy & Fuels, 2012, 26(10): 6128–6136

[4] Luo K, Bai Y, Jin T, Qiu K, Fan J. Direct numerical simulation study on the stabilization mechanism of a turbulent lifted pulverized coal jet flame in a heated coflow. Energy & Fuels, 2017, 31(8): 8742–8757

[5] Chen L, Ghoniem A F. Simulation of oxy-coal combustion in a 100 kW_th test facility using RANS and LES: a validation study. Energy & Fuels, 2012, 26(8): 4783–4798

[6] Belosevic S, Sijercic M, Tucakovic D, Crnomarkovic N. A numerical study of a utility boiler tangentially-fired furnace under different operating conditions. Fuel, 2008, 87(15–16): 3331–3338

[7] Belosevic S, Sijercic M, Crnomarkovic N, Stankovic B, Tucakovic D. Numerical prediction of pulverized coal flame in utility boiler furnaces. Energy & Fuels, 2009, 23(11): 5401–5412

[8] Al-Abbas A H, Naser J. Computational fluid dynamic modelling of a 550 MW tangentially-fired furnace under different operating conditions. Procedia Engineering, 2013, 56: 387–392

[9] Tian D, Zhong L, Tan P, Ma L, Fang Q, Zhang C, Zhang D, Chen G. Influence of vertical burner tilt angle on the gas temperature deviation in a 700 MW low NO_x tangentially fired pulverised-coal boiler. Fuel Processing Technology, 2015, 138: 616–628

[10] Belošević S, Tomanović I, Crnomarković N, Milićević A, Tucaković D. Numerical study of pulverized coal-fired utility boiler over a wide range of operating conditions for in-furnace SO₂/NO_x reduction. Applied Thermal Engineering, 2016, 94: 657–669

[11] Dong J, Zhou T, Wu X, Zhang J, Fan H, Zhang Z. Coupled heat transfer simulation of the spiral water wall in a double reheat ultra-supercritical boiler. Journal of Thermal Science, 2018, 27(6): 592–601

[12] Dong J, Fan H, Wu X, Zhou T, Zhang J, Zhang Z. Study on the effect of flame offset on water wall tube temperature in 600°C and 700°C ultra-supercritical boiler. Combustion Science and Technology, 2019, 191(3): 472–490

[13] Liu H, Zhang L, Li Q, Zhu H, Deng L, Liu Y, Che D. Effect of FGR position on the characteristics of combustion, emission and flue gas temperature deviation in a 1000 MW tower-type double-reheat boiler with deep-air-staging. Fuel, 2019, 246: 285–294

[14] Shen H, Wu Y, Xu K, Smith P J, Zhang H. Eulerian LES simulation of coal jet flame with a simplified DQMOM model. Fuel, 2018, 216: 475–483

[15] Sundén B, Faghri M. Transport Phenomena in Fires. Britain: WIT Press, 2008

[16] Pedel J, Thornock J N, Smith P J. Ignition of co-axial turbulent diffusion oxy-coal jet flames: experiments and simulations collaboration. Combustion and Flame, 2013, 160(6): 1112–1128

[17] Pedel J, Thornock J N, Smith S T, Smith P J. Large eddy simulation of polydisperse particles in turbulent coaxial jets using the direct quadrature method of moments. International Journal of Multiphase Flow, 2014, 63: 23–38

[18] Smoot L D, Smith P J. Coal Combustion and Gasification. New York: Plenum Press, 1985

[19] Schroeder B B, Smith S T, Smith P J, Fletcher T H, Packard A, Frenklach M, Hegde A, Li W, Oreluk J. Scale-bridging model development for coal particle devolatilization. arXiv, 2016:1609.00871

[20] Zhou M, Parra-Álvarez J C, Smith P J, Isaac B J, Thornock J N, Wang Y, Smith S T. Large-eddy simulation of ash deposition in a large-scale laboratory furnace. Proceedings of the Combustion Institute, 2019, 37(4): 4409–4418

[21] Zhang Z, Wu Y, Chen D, Shen H, Li Z, Cai N, Zhou M, Smith S T, Thornock J N, Isaac B J. A semi-empirical NO_x model for LES in pulverized coal air-staged combustion. Fuel, 2019, 241: 402–409

[22] Shen H, Wu Y, Zhou M, Smith S T, Zhang H, Yue G. Identification of the initial particle size distribution for coal combustion simulations. AIChE Journal. American Institute of Chemical Engineers, 2019, 65: e16610

RIGHTS & PERMISSIONS

Higher Education Press