循环流化床反应器内流动和反应过程与时空多尺度结构紧密耦合，是典型的非线性非平衡复杂系统。此类多相复杂系统，远离平衡态，单纯依靠细化网格并不能捕捉真实物理过程，必须建立合理的介尺度模型。在过去的数十年中，有大量的应用欧拉-欧拉模型模拟流化床的研究工作发表。已有研究表明，气固相间曳力对模拟循环流化床内流动行为非常重要，必须考虑介尺度非均匀结构对曳力的影响。EMMS模型通过稳定性条件分析不同尺度间的联系，揭示产生结构的主要机理，可准确捕捉循环流化床内的流动特征，并且在较粗网格上可以达到网格无关，因而可应用于工业规模反应器的模拟计算。然而，在宽颗粒筛分时，如采用多流体模拟，每增加一个颗粒相的拟流体，欧拉-欧拉模型都需要增加一组方程，计算时间和资源要求较多，且耦合收敛难度增加。欧拉-拉格朗日模型对离散相采用牛顿第二定律获得位移、速度等信息，直接采用离散方法模拟固相，更适合宽筛分条件下的研究。然而，如果直接追踪每一个颗粒的信息，则由于频繁和大量的颗粒碰撞搜索等因素，计算规模受限。采用粗粒化方法，既降低了计算粒子的规模，同时又保留了追踪离散颗粒的特征，较适合于大规模宽粒径分布反应器的研究。粗粒化离散模拟方法仍处于发展阶段，目前已发表的大规模反应器研究的文献主要集中在Geldart B类体系的模拟。对于A类颗粒体系，介尺度结构对曳力的影响，计算的网格相关性，以及粗粒化参数的选取等因素还需要进一步研究。针对上述问题，本论文开展了流化床反应器从二维到三维全循环的模拟研究，并将其结果与欧拉-欧拉方法的研究结果进行对比，指出粗粒化欧拉-拉格朗日模型仍需考虑介尺度结构对曳力的影响，同时对粗粒化参数、颗粒碰撞参数等关键模型参数进行了系统分析，为大规模流化床反应器流动和反应的模型和参数选取提供了基础。论文第一章介绍了流态化、颗粒分类以及模型分类等基础信息，并介绍了研究的内容和目标。论文第二章介绍了粗粒化欧拉-拉格朗日模型之一的Dense Discrete Phase Model (DDPM)的控制方程、曳力系数和粗粒化参数。论文第三章介绍了DDPM在二维循环流化床提升管中的模拟应用，研究了曳力模型和网格尺寸对Geldart A类和B类颗粒体系的影响。研究表明，与欧拉-欧拉模型研究结果类似，对于Geldart A类颗粒体系，EMMS曳力模型可以在较粗网格下获得网格无关结果，并且模拟结果与实验结果吻合较好，而采用均匀曳力模型的模拟结果则与实验结果存在较大差距；对于Geldart B类颗粒体系，EMMS模型和均匀曳力模型都可以得到与网格近似无关的结果，且结果与实验吻合较好。在此基础上，论文第四章将DDPM模型应用于中试规模循环流化床提升管的三维模拟，并对模型中粗粒化参数的选取进行了研究。结果表明，在较粗的网格下，EMMS曳力模型的结果比均匀曳力模型的结果更接近实验值。对于DDPM模型的关键参数：粗粒化比率 (dcl/dp),本研究将其与介尺度结构-团聚物的直径（由EMMS求解得到）关联起来。结果表明，采用团聚物直径范围内的粗粒化比率(dcl/dp= 65~125) 可得到与实验吻合的结果 。基于此，将耦合EMMS曳力模型并根据EMMS模型得到粗粒化比率的DDPM模型称为基于EMMS的DDPM模型。另外，本章还进行了欧拉-欧拉模型和粗粒化欧拉-拉格朗日模型模拟结果的比较，研究表明，两种模型都需要考虑介尺度结构的影响。论文第五章将DDPM模型应用于中试规模循环流化床提升管的三维全循环模拟，并对曳力系数、颗粒作用的关键参数（颗粒 - 颗粒恢复系数和颗粒 - 壁恢复系数）和颗粒粒径分布进行了系统研究。研究结果进一步证明，基于EMMS的DDPM模拟，可以获得合理的全循环流动结果。此外，颗粒- 颗粒恢复系数 (epp)和颗粒-壁面恢复系数 (epw) 对流体力学的研究表明，epw 在理想和非理想范围内对流体动力学行为没有显著影响，而epw在非理想范围内，即 epp= 0.1~0.9，模拟结果与实验吻合较好，当 epp= 1.0 时，模拟结果不合理。对粒径分布的影响表明，对颗粒粒径分布较窄的情况，采用多粒径分布和采用平均粒径的模拟都可以给出合理的轴向分布、颗粒径向速度和浓度分布，但采用多粒径分布的模拟结果可更细致刻画不同大小颗粒在床内运动规律的不同。在此基础上，论文第六章将基于EMMS的DDPM模型应用于工业规模循环流化床燃烧反应的模拟。研究结果成功复现了炉膛内复杂的流动、传热和反应特性，为循环流化床锅炉的设计、优化以及污染物脱除规律的研究奠定了基础。第七章总结了本论文获得的主要成果，并对基于EMMS的DDPM模型的应用前景以及进一步开展研究的方向进行了讨论。;The flow and reaction processes in a circulating fluidized bed reactor (CFB) are closely coupled with the spatio-temporal multi-scale structure, which is a typical nonlinear non-equilibrium complex system. This kind of multiphase complex system is far from equilibrium state, and it is impossible to capture the real physical process simply by refining computational mesh, thus requiring a reasonable meso-scale model. In the past few decades, a large number of studies have been devoted to simulations of fluidized beds by using Eulerian-Eulerian models, showing that the drag force between gas and solid is very important, and the influence of meso-scale structure must be considered. The energy-minimization multi-scale (EMMS) model reveals the main mechanism of structure formation by analyzing the relationship between different scales through stability conditions. It can accurately capture the flow and reaction characteristics in circulating fluidized beds, and can achieve grid-independent prediction on coarse grids, so it can be applied to the simulation of industrial scale reactors.However, in the case of wide particle size distributions, if multi-fluid simulation is used, the Eulerian-Eulerian model needs to add a set of equations for each quasi-fluid of a particle phase, which requires more computational time and resources, and makes the coupling convergence more difficult. The Eulerian-Lagrangian model treats the discrete phase by Newton's second law to obtain the information of displacement, velocity etc., and the use of discrete method to simulate solid phase directly is more suitable for the study of wide particle size distributions. However, if the information of each particle is tracked directly, the computational scale is limited due to the frequent and large number of particle collision searches. Coarsening method not only reduces the size of calculated particles, but also retains the characteristics of discrete particles tracking. Thus it is more suitable for large-scale reactor with wide particle size distribution.The coarse-grained discrete simulation method is still in development, and the published literature on simulations of large-scale reactors mainly focuses on the Geldart B system. For Geldart A system, the influence of mesoscale structure on drag force, grid dependency and the selection of coarsening parameters need further study. In consideration of the above problems, in this thesis, the simulations of fluidized bed reactors from two-dimensional to three-dimensional full loop scale were carried out, and the results were further compared with the results of the Eulerian-Eulerian method. The researches revealed that the influence of mesoscale structure on drag force should also be considered in the coarse-grained Eulerian-Lagrangian model. The key model parameters, such as coarse graining ratio, particle collisions were analyzed systematically, which provided the basis for the selection of model and parameter for the simulation of flow and reactions in large-scale fluidized bed reactors.In Chapter 1, the basic information of fluidization, particle classification and model classification are introduced, and the research contents and objectives are also introduced.In Chapter 2, the governing equations, drag coefficients and coarsening parameters of Dense Discrete Phase Model (DDPM), one of the coarse-grained Eulerian-Lagrangian models, are introduced.In Chapter 3, simulation with DDPM in two-dimensional circulating fluidized bed riser is introduced. The effects of drag model and mesh size on Geldart A and B particle systems are studied. Similar to the results of the Eulerian-Eulerian model, the DDPM with the EMMS drag can obtain mesh-independent results for the Geldart A granular system with coarse grids, and the simulation results are in good agreement with the experimental results. However, there exist a large gap between the simulation results by using the conventional uniform drag and the experimental results. For Geldart B granular systems, the simulations using both the EMMS and uniform drag models can reach grid-independent results, and the results are in good agreement with the experiment.On this basis of the previous chapter, in Chapter 4, 3D DDPM simulations are carried out on the riser of a pilot-scale circulating fluidized bed. The selection of coarsening parameters in the model is studied. The simulation results using the EMMS drag are closer to the experimental than those of the uniform drag in coarser grids. Coarse-graining ratio (dcl/dp), a key parameter of the DDPM model, is correlated with mesoscale structure-the diameter of clusters (obtained using EMMS). The results show that, the simulations done with the coarse-graining ratio (dcl/dp= 65-125) within the range of the cluster diameters can give reasonable results with the experimental data. Based on this, the DDPM model which is coupled with the EMMS drag and with coarse-graining ratio calculated by the EMMS model is called the EMMS based DDPM model. In addition, the simulation results of Eulerian-Eulerian model and coarse-grained Eulerian-Lagrangian model are compared, and the results show that the effect of mesoscale structures should be considered in both models.In Chapter 5, 3D DDPM simulations are carried out on the full-loop of a pilot-scale circulating fluidized bed, and the effects of the drag coefficient, the key parameters of particle interaction (the particle-particle restitution coefficient and particle-wall restitution coefficient) and particle size distribution are systematically studied. The results further proved that the EMMS based DDPM simulations can obtain reasonable results in a full-loop CFB. In addition, the results showed that the particle-wall restitution coefficient (epw) has little effect on the hydrodynamic behavior; for the particle-particle restitution coefficient , when it is in the non-ideal range, i.e. epp= 0.1~0.9, the simulation results agree well with the experimental data. For the case of narrow particle size distribution, reasonable axial distribution, radial velocity and concentration distribution can be obtained by using both particle size distribution and average particle size, though the simulation with considering of particle size distribution allows detailed description of the trajectories of different particles.On this basis, in Chapter 6, the EMMS based DDPM model is applied to the simulation of combustion reaction in an industrial scale circulating fluidized bed. The complex flow, heat transfer and reaction characteristics in the furnace were successfully captured, thus providing a foundation for the design, optimization and pollutant removal of a CFB boiler.Chapter 7 summarizes the main achievements of this paper, and discusses the application prospects and further research directions of the EMMS based DDPM model.