Article pubs.acs.org/EF
Mechanistic Model for Ash Deposit Formation in Biomass Suspension Firing. Part 1: Model Verification by Use of Entrained Flow Reactor Experiments Stine Broholm Hansen,† Peter Arendt Jensen,*,† Flemming Jappe Frandsen,† Bo Sander,‡ and Peter Glarborg† †
Department of Chemical and Biochemical Engineering, Technical University of Denmark, Building 229, DK-2800 Lyngby, Denmark DONG Energy Thermal Power, Kraftværksvej 53, DK-7000 Fredericia, Denmark
‡
ABSTRACT: Two models for deposit formation in suspension firing of biomass have been developed. Both models describe deposit buildup by diffusion and subsequent condensation of vapors, thermophoresis of aerosols, convective diffusion of small particles, impaction of large particles, and reaction. The models differ in the description of the sticking probability of impacted particles: model #1 employs a reference viscosity in the description of the sticking probability, while model #2 combines impaction of viscoelastic particles on a solid surface with particle capture by a viscous surface. Both models were used to describe the deposit formation rates and deposit chemistry observed in a series of entrained flow reactor (EFR) experiments using straw and wood as fuels. It was found that model #1 was not able to describe the observed influence of temperature on the deposit buildup rates, predicting a much stronger influence of this parameter. Model #2 was able to provide a reasonable description of the influence of temperature on the deposit buildup rates observed in the EFR experiments. A parametric study was conducted to examine the influence of some physical parameters, including ash concentration, viscosity of ash and deposits, surface tension, Young’s modulus, and porosity. On the basis of this model evaluation, where a wide range of temperatures (700−1000 °C) and fuels (straw and wood) were applied, model #2 can be regarded as a promising tool for the description of deposit formation from biomass ashes. ⎧ μref μ > μref ⎪ p=⎨ μ ⎪ 1 μ≤μ ⎩ ref
1. INTRODUCTION In order to reduce CO2 emissions and the dependence on fossil fuels, complete substitution of coal with biomass in power stations by the year 2050 is targeted in Denmark,1 where combustion is the main thermal conversion process for heat and power production.2 Initially, pure biomass (primarily straw) was utilized only in grate-fired boilers, but since 2001 suspensionfired boilers have also been utilizing pure straw and/or wood, since combustion in suspension firing has a higher electrical efficiency (46−48%) than grate-fired boilers (25−30%) and therefore is an attractive option.3 However, the use of biomass in pulverized fuel combustion plants is a challenge because of the chemical and physical nature of biomass ash, which may lead to increased deposit formation and subsequent corrosion in such boilers. Deposits are formed by at least three different mechanisms, which apply to different physical fractions of the fly ash:4 (1) condensation on heat transfer surfaces, especially by gas-phase species such as KCl and K2SO4; (2) thermophoresis of aerosol particles (10 μm) on heat transfer surfaces. One major challenge in ash deposition modeling is to determine the particle stickiness upon impaction with a heat transfer surface. Walsh et al.5 examined ash deposit buildup and suggested the following simple relationship between the viscosity (μ) and sticking probability (p) of a particle: © XXXX American Chemical Society
(1)
Particles with viscosities lower than a reference viscosity (μref) were assumed to stick. Walsh et al.5 set the reference viscosity equal to 8 Pa·s. The model has been applied in several models of coal ash deposition, where the value of the reference viscosity was set equal to, e.g., 104 Pa·s6−8 or 105 Pa·s.9,10 Srinivasachar et al.11 showed that the critical viscosity of a deposit buildup is dependent on the particle kinetic energy, i.e., on the mass/size and velocity of the particles (see Figure 1). On the basis of these results, Srinivasachar et al.11 suggested that the reference viscosity in coal combustion should lie between 104 and 108 Pa·s because of the particle sizes and velocities used in these systems. The reference viscosity determined in one combustion system is thus not generally applicable,12 and μref can therefore be regarded as a fitting parameter during modeling13 since no correlation between particle size and velocity and the reference viscosity has been suggested at present. An alternative approach to describe the sticking probability has been employed when modeling deposit buildup during biomass combustion in grate-fired or circulating fluidized bed boilers, where the ash contains a significant fraction of alkali, possibly as KCl and K2SO4. These salt-forming components influence the Received: July 14, 2016 Revised: October 7, 2016
A
DOI: 10.1021/acs.energyfuels.6b01659 Energy Fuels XXXX, XXX, XXX−XXX
Article
Energy & Fuels
Once a layer of deposited ash (and/or salts) has formed on a heat transfer surface, the surface may capture incoming particles that are otherwise nonsticky. During deposit buildup, the deposit surface temperature increases, leading to a lower viscosity of the deposit surface. When a particle impacts this surface, the kinetic energy of the particle will be lost through resistance of viscous drag as the particle (partly) penetrates the viscous surface. Raask40 describes the critical velocity for surface capture as us,crit =
sticking behavior of the ash and deposits by formation of a melt. It was assumed that particles are nonsticky when they contain less than 10−15% melt and completely sticky when the melt fraction exceeds 70%.14−17 The simple models mentioned above do not consider particle size or velocity in their sticking criteria and thus may not be widely applicable. Models considering particle or surface properties as well as the size and velocity of the particles to determine sticking or rebound have been developed for impact of elastic−plastic materials,18−22 purely elastic materials,23−26 viscoelastic materials,27,28 and molten materials (droplets).29−36 Some have been applied to ash deposit formation models.20,27−30,37 The models generally describe the impact by an energy balance, considering the kinetic energy of the particles, the energy lost by deformation of the particle or surface, the contact area established, and the adhesion energy of the contact area. The material properties of the ashes from suspension firing of straw and straw/wood mixtures were analyzed by Nordgren et al.,38 who found that fly ashes and deposits contain significant amounts of amorphous phases.38 Amorphous materials behave like a glass (elastic material) at low temperatures (below the glass transition temperature), like a rubber (viscoelastic material) at intermediate temperatures, and as a viscous liquid at high temperatures.39 For viscoelastic particles impacting a solid surface, a critical velocity for rebound or capture can be calculated using eq 2,24 which is based on the JKR adhesion theory:25
(3)
2. EXPERIMENTS The combustion experiments were conducted in an EFR that was designed to simulate the environment of a suspension-fired boiler.42 The experiments reviewed for this modeling approach were conducted in four different series of experiments42−46 over the years 2002−2009. Figure 2 shows a schematic drawing of the experimental setup, which consisted of a reactor and a bottom chamber.42 In the bottom chamber, the main part of the flue gas was led through a 4 cm × 8 cm duct toward a deposit probe. In the duct, a propane burner was mounted to ensure a flue gas temperature of approximately 800 °C just before the probe. The probe was a Ø = 1 cm stainless steel probe with a controlled surface temperature in the range 450−600 °C.42,44,46 Four different straws (A−D) were fired in 19 different experiments (S1−S19), and four woods (bark, pine, beech, and waste wood (WW)) were fired in five experiments (W1−W5). Proximate and ultimate analyses are provided in Table 1. Straws C and D had ash compositions similar to a typical Danish straw,47 whereas straws A and B had lower contents of K and Cl. The wood ashes showed significant variation in ash content and composition. The pine ash had a composition closest to the composition of the wood pellets used in a Danish power plant examined in previous work.42,48,49 The key input parameters and the experimental results are shown in Table 2.
⎛ γ ⎞5/6 ⎜ ⎟ ⎝ rp ⎠ (ρp3 E2)1/6
γ 2 sin α
where μs is the viscosity of the deposit surface, g is the acceleration due to gravity, and α is the angle of impact. The model has been found to apply to surfaces with viscosities well above the molten range (≫10 Pa·s).40 Deposit formation models are found in three levels according to their complexity: (i) empirical indices models, (ii) mechanistic models, and (iii) CFD models.41 Most of the models proposed during the past decade(s) are CFD models, which strongly focus on flow dynamics but often treat the ash chemistry and physics in a rather rough way. These have been reviewed by Weber et al.13 and are concluded to be indicative at best because of tuning of key parameters (e.g., the reference viscosity) to obtain a model fit.13 In mechanistic models, calculations of the complicated combustion process and fluid dynamics are simplified. Such models have been used to assess the ash deposition tendency and prediction of ash behavior. Mechanistic models of deposit formation are mainly found in relatively old models of coal ash deposition (e.g., by Baxter4) but have also been developed for grate firing of straw by Zhou et al.16 The aim of this work was to develop a mechanistic model for ash deposit buildup on a single cylinder during suspension firing of straw or wood. The model predictions have been compared to measurements made in an entrained flow reactor (EFR) with a deposition probe mounted. In these measurements, various straws and woods were combusted and deposit formation was quantified at a range of probe and flue gas temperatures.
Figure 1. Variation of the reference viscosity of deposition as a function of the kinetic energy of a particle (Ke) relative to the kinetic energy of small particles (Ke,0). Results from experiments with soda-lime-glass spheres are shown. Adapted from Srinivasachar et al.11
u p,crit =
ρp g 2r 3μs
(2)
In eq 2, γ is the surface tension, rp and ρp are particle radius and density, respectively, and E is the Young’s (elastic) modulus of the particle. In order to describe the elastic modulus of a viscoelastic solid, the time (and temperature) history of the solid must be followed. In a computational fluid dynamics (CFD) model, Losurdo and co-workers27,28 present four models for describing the Young’s modulus from the time−temperature history of the particle, the time duration of the impact, and the viscosity of the particle. B
DOI: 10.1021/acs.energyfuels.6b01659 Energy Fuels XXXX, XXX, XXX−XXX
Article
Energy & Fuels
reactor and are “lost” primarily as bottom ash. The ash mass balance is expressed as 1 = X vap + Xcoarse,fly + Xbottom/lost
(4)
where Xvap is the vaporized fraction of the fuel ash (forming gases and aerosols), which is estimated in the model; Xcoarse,fly is the fraction of the ash found as coarse ash, which follows the flue gas; and Xbottom/lost is the lost fraction of the coarse ash. In the experiments, the fly ash concentration was measured by collection of ash particles in the sample flow in a cyclone and a filter. These collected ash samples was weighed and analyzed in terms of water-soluble content (K, Na, Cl, and SO4). The watersoluble content can be regarded as originally present as gases and aerosols in the flue gas. From these data, the amount of insoluble ash collected was determined and used to calculate Xcoarse,fly. The description of gas and aerosol concentrations occurs in three consecutive steps in the model: (1) release of volatile elements, (2) gas-phase reactions, and (3) aerosol formation. In recent work, a model for estimation of the release of volatile elements during suspension firing of biomass was developed.49 In this model it is assumed that the fuels experience full release of S and Cl to the gas phase, while the release of K is described by a function of the (Ca + Mg)/Si and K/Si molar ratios:
Figure 2. Sketch of the entrained flow reactor used for the deposition experiments.
3. THE MODEL In this section, the model for prediction of ash deposit buildup is outlined. However, prior to a calculation of deposit formation, a proper description of the ash found in the flue gas is needed. The model thus consist of two parts: an ash formation model and a deposit buildup model. 3.1. Ash Formation Model. During the combustion, the inorganic matter (ash) in a fuel particle is divided into two parts: a vaporized fraction and a coarse fraction (see Figure 3). The vaporized part (K, Cl, and S) may then react and nucleate, forming gaseous components and aerosols. The coarse fraction of the ash can be described by a particle size distribution and by the chemistry of the particles. Part of the coarse ash will be found as large particles that are unable to follow the gas through the
fK‐release
⎧ (Ca + Mg) for > 0.5 ⎪ 0.9 Si ⎪ ⎪ ⎪ (Ca + Mg) K = ⎨ 0.9 for < 0.5 and >1 Si Si ⎪ ⎪ (Ca + Mg) K K ⎪ < 0.5 and
