WP2: Development and Validation of Scale-up Tools (TUD)

A steady-state process model, a dynamic process model, and three-dimensional models for reactor simulation were developed and validated by experimental data from comprehensive test campaigns in WP1. All models from WP2 were scaled-up to 20 MWth and applied to support the upscaling activities planned in WP3 to WP6.

Steady-state process model

An existing in-house steady state process model [1] was advanced to account for the physical effects inside circulating fluidized bed reactors regarding the sorbent-gas and sorbent-sorbent interactions using the software ASPEN PLUSTM with customized FORTRAN routines. In the first place, the process model (Figure 7) is capable to predict mass and energy balances of the Calcium Carbonate Looping (CCL) system. Particles attrition effects and consequently overall reaction rates in both reactors were considered. The carbonator and the calciner are modelled by detailed circulating fluidized bed (CFB) sub-models (a, b in Figure 1 left). Each of the reactor sub-models take into account the CFB hydrodynamics, as well as reaction kinetics for carbonation and calcination [1, 2]. Both cyclones are modelled according to Muschelknautz [3], based on the dimension of those cyclones applied in the pilot plant. Figure 1 (right) depicts the simulated efficiencies of carbonator (ECarb) and calciner (ECalc) that are in good accordance with experimental data obtained during four different operating points in tests from WP1.

Figure 1: Simplified flowsheet of the steady-state process model (left) and comparison of carbonator and calciner efficiencies (right)

Dynamic process model

A dynamic process model was developed using the software ASPEN PLUS Dynamics/Custom modeler. The model includes a detailed reactor model for the carbonator that includes several sub-models to calculate among other values e.g. the entrainment rate, hydrodynamics, particle distribution and CO2 absorption rate depending on the history of limestone. Furthermore, the model calculates a heat and mass balance of the carbonator in association with custom-built sub models that consider hydrodynamics and heat transfer among phases (Figure 2 left). Hence the model allows the prediction of the reactor temperature over time as it is exemplary depicted in Figure 2 (right) during start-up operation. The detailed thermodynamic model allows the calculation of the reactor temperature for different operating conditions. Hence realistic start-up, shut-down and load change scenarios using burner power or pre-heated air can be investigated and studied with the model. The sorbent distribution is described by a two-zone model according Kunii and Levenspiel that assumes perfect sorbent mixing in each zone. Additionally, a sub-model for considering the effects of steam on the CO2 absorption rate was included. Appropriate loop controllers can be used to set up specified behavior of single CCL plant components (e.g. screw-conveyer, electrical pre-heaters etc.). The dynamic process model was validated with experimental data during start-up, shut-down and load change showing satisfactory agreement during all operating phases of the CCL plant. After thorough model validation with experimental data from 1 MWth pilot plant the model was scaled-up to 20 MWth, see WP3.

Figure 2: Thermodynamic heat streams of dynamic carbonator process model (left) and simulated carbonator temperature profile during start-up operation (right)

3D CFD models

3-D Computational Fluid Dynamics (CFD) tools were developed and validated for the simulation of the gas-solid flows inside the carbonator and calciner reactors. The models were developed using three different numerical approaches that can be divided in Euler-Lagrange and Euler-Euler (Figure 3). In the latter mentioned method, the particulate phase is treated as a fluid with closure equations according the theory of granular flows to describe the solid properties. In the former case the particles are tracked in the course of time with individual properties such as diameter, velocities, densities, sorbent conversion etc. Furthermore, the Euler-Lagrange models can be distinguished by their collision methodology, either deterministic or stochastic based collision formulation. The coarse grain discrete element method model that is based on a deterministic collision algorithm was developed by TUD while the stochastic based collision models were developed by GE. The Euler-Euler model was incorporated by CERTH/CPERI. In all models appropriate reaction rates were applied either retrieved from the recent literature or small scale experimental batch reactor tests. Apart from this, a new version of the innovative sub-grid EMMS (Energy Minimization Multi-Scale) method was applied to the Euler-Euler and Euler-Lagrange models for considering a realistic description of momentum exchange between phases. Due to that, the particle distribution in the CFB unit was simulated more realistically in comparison to conventional drag models. Additionally, a routine was implemented to maintain a realistic particle size distribution for inert ash and reactive limestone particles during the transient simulation. For modeling of the appropriate drag, reaction rates and particle size distribution were integrated by means of custom built models in the C programming language into the ANSYS platform using User Defined Functions (UDFs).

Figure 3: Numerical methods applied in Task 2.3

The developed Barracuda approach based reactor models couple first principle thermochemical models (particle physical properties, particle chemistry, and gas-particle reaction kinetics) with scalable hydrodynamic models, to provide a robust basis for describing performance at various scales. The modelling of reactors considers mass and heat transfer between gas and particle species, multiple simultaneous reactions, and heat loss to the environment. Reactor modelling also considers the entire circulation loop, capturing particle entrainment flux and allowing the tracking of individual particles, both position and associated properties in time. Figure 12 depicts on the left hand side two instantaneous contour plots of the solids distribution using Gidaspow and EMMS drag models. It can be seen that the EMMS model gives a more pronounced dense zone with a thinner lean zone in comparison to the conventional Gidaspow model. Additionally the pressure profile is directly affected by the particle distribution, as it is depicted on the right hand side of Figure 4.

Figure 4: CFD modeling results of 1 MWth scale

Design improvements

Potential design improvements were investigated on the experiences from 1 MWth pilot testing, validated process and 3D models respectively. The experiences from pilot testing allow identifying optimization potentials, especially in geometry of the carbonator calciner, cyclones, loops seals and devices for coupling between the two fluidized bed reactors for different operating conditions. Furthermore, the CFD analysis in terms of CO2 absorption efficiency favored a modified carbonator fat bottom design, that yielded a higher capture efficiency than the original design. Design improvements were jointly defined in a separate workshop. TUD integrated the proposed design modifications for the 1 MWth plant into an existing scaled cold flow model and tested the performance under cold conditions (Figure 5). Experiments consisted of four separate configurations: each fluidized bed (CFB600=carbonator, CFB400=calciner) internally recirculating on its own, known as stand-alone operation, and two configurations in which solids circulation occurred between fluidized beds, known as coupled operation.  By adjustment of the J-valve and cone valve, it was possible to control the amount of solids between reactors and achieve stable operation. The removal of LS4.5 from the reactor configuration was also investigated. In this case, the cone valve’s transfer capacity improved so much that the J-valve became the limiter of solids transfer between the two reactors. However, the configuration was much more vulnerable to gas bypass, especially at start-up, because fluidization air from CFB 600 could flow directly up the standpipe into the cyclone of CFB 400.

Figure 5: Scheme of Cold Flow Model (left) and adaption with FBHE loop seal (right)

Conclusions

The developed models in WP2 are strong tools that can be used for the process scale-up. The developed steady-state process model successfully validated by the experimental data in 1 MWth scale allows to define the design heat and mass balance in 20 MWth scale in WP3 and to assess the full-scale implementation in WPs4-6.The developed 3D CDF models allow to evaluate the design based on the process modelling. Detailed investigations of reactor geometries and its effect on the performance are possible.

References

[1] J. Ströhle, M. Junk, J. Kremer, A. Galloy, and B. Epple, “Carbonate looping experiments in a 1MWth pilot plant and model validation,” Fuel, vol. 127, pp. 13-22, 2014.
[2]  I. Martínez, G. Grasa, R. Murillo, B. Arias, and J. C. Abanades, “Kinetics of Calcination of Partially Carbonated Particles in a Ca-Looping System for CO2 Capture,” Energy & Fuels, vol. 26, pp. 1432–1440, 2012.
[3]  E. Muschelknautz, “Die berechnung von zyklonabscheidern für gase,” Chemie Ingenieur Technik, vol. 44, pp. 63-71, 1972.