Reduction of Fine Particle Emissions from Wood Combustion with

Jul 22, 2009 - Research Centre of Finland, Fine Particles, P.O. Box 1000,. FI-02044 VTT, Espoo, .... valves, the flows of the hot and the cold water w...
0 downloads 0 Views 272KB Size
Environ. Sci. Technol. 2009, 43, 6269–6274

In this study, we designed and built a condensing heat exchanger capable of simultaneous fine particle emission reduction and waste heat recovery. The deposition mechanisms inside the heat exchanger prototype were maximized using a computer model which was later compared to actual measurements. The main deposition mechanisms were diffusio- and thermophoresis which have previously been examined in similar conditions only separately. The obtained removal efficiency in the experiments was measured in the total number concentration and ranged between 26 and 40% for the given pellet stove and the heat exchanger. Size distributions and number concentrations were measured with a TSI Fast mobility particle sizer (FMPS). The computer model predicts that there exists a specific upper limit for thermo- and diffusiophoretic deposition for each temperature and water vapor concentration in the flue gas.

that the particle size, number concentration, surface area, and chemical composition have greater relevance in terms of health effects than the particle mass concentration (4). The common aerosol size distribution from wood combustion peaks at 50-400 nm with relatively high number concentrations (5). In power plants, particle number concentrations are typically 1 order of magnitude smaller due longer residence times and the resulting agglomeration. The exposure levels, and thus the health risk, is further elevated as the majority of residential wood combustion occurs in devices that lack both emission control technology and optimized heat recovery. Condensing heat exchangers (CHX’s) can be optimized for simultaneous particle collection and waste heat recovery. If one condenses the water vapor from the flue gas, this will improve the heat recovery and reduce the slagging and the fouling of the heat exchanger by generating a constant water film on the heat exchanging surface which carries away any deposited particles. In pure wood combustion, there is no corrosion problem since the chlorine and sulfur content of wood fuel is low (6) and consequently only small amounts of corrosive alkali chlorides, HCl and SO2 are present in the flue gas. Fouling of the CHX might be affected by the used fuel type and combustion device. Furthermore, in order to make a CHX suitable for batch combustion, a flue gas blower must be employed. Particle deposition in heat exchangers has been studied previously mainly focusing either on diffusion (7) or on thermophoresis (8, 9) and there have been claims that by using these mechanisms simultaneously, then the deposition efficiency could be further enhanced (10). In these studies, the deposition efficiency has been found to vary from 25 to 95% depending clearly on the conditions, especially the temperature gradient and the condensation flux. In this work, we modeled, designed and applied an optimized tube bundle CHX for a commercial pellet stove. A computer model was developed and used to design the CHX and later the prototype was tested with nine combinations of two investigated variables. The investigated characteristics were the reduction in number concentration of particles and heat transfer in the CHX. The model was used to simulate some feasible alternatives for simultaneous heat recovery and particle reduction.

Introduction

Materials and Methods

Reduction of Fine Particle Emissions from Wood Combustion with Optimized Condensing Heat Exchangers ¨ HN,† VALTTERI SUONMAA,† ARTO GRO A R I A U V I N E N , ‡ K A R I E . J . L E H T I N E N , §,| A N D J O R M A J O K I N I E M I * ,†,‡ Department of Environmental Science, Fine Particle and Aerosol Technology Laboratory, University of Kuopio, P.O. Box 1627, FI-70211 Kuopio, Finland, VTT Technical Research Centre of Finland, Fine Particles, P.O. Box 1000, FI-02044 VTT, Espoo, Finland, Finnish Meteorological Institute, Kuopio Unit, P.O. Box 1627, FI-70211 Kuopio, Finland, and Department Physics, University of Kuopio, P.O. Box 1627, FI-70211 Kuopio, Finland

Received January 5, 2009. Revised manuscript received June 15, 2009. Accepted July 14, 2009.

Wood combustion generates fine particles which contribute significantly to the emissions from the energy sector. For example, in the emission inventory conducted by the Finnish Environment Institute, the energy sector accounts for of 72% of the PM2.5 fine particle emission in Finland, with residential wood combustion representing the largest single source with a share of 25% of the total direct PM2.5 emissions (1). As the use of wood-based fuels is predicted to increase by encouraging renewable energy use in general, the emissions can also be expected to grow in the future if no limiting measures are taken. The epidemiological correlation between elevated ambient fine particle concentrations and adverse health effects has been examined in many studies (2-4). There is evidence * Corresponding author phone: +358 40 505 0668; fax: +358 17 16 3229; E-mail: [email protected]. † Department of Environmental Science, University of Kuopio. ‡ VTT Technical Research Centre of Finland. § Finnish Meteorological Institute. | Department Physics, University of Kuopio. 10.1021/es8035225 CCC: $40.75

Published on Web 07/22/2009

 2009 American Chemical Society

Experiments. For the CHX type, we selected a simple one pass counter flow shell-and-tube heat exchanger. For example, compared to cross-flow geometries, the used heat exchanger has reduced heat transfer properties but is more efficient in terms of particle removal. Inside the tubes, the

TABLE 1. The Tested Experimental Settings Arranged by the Volume Flow (Q′) of Cooling Water and Average H2O Concentration (CH2O (g)) experimental setting

Q′cooling (m3/h)

CH2O (g) (vol-%)

L1 L2 L3 M1 M2 M3 H1 H2 H3

0.124 0.124 0.124 0.360 0.360 0.360 0.580 0.571 0.571

5.6 16.1 28.2 5.4 16.3 28.6 6.2 16.0 28.3

VOL. 43, NO. 16, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6269

TABLE 2. Particle Statistics at the Inlet (Before) and the Outlet (After) of the CHX Combined with Collection Efficiencies, Measured with the FMPSa N(tot) before (× 1014) 3

L1 L2 L3 M1 M2 M3 H1 H2 H3

N(tot) after (× 1014)

(no./Nm )

σ

(no./Nm3)

σ

CMD before (nm)

CMD after (nm)

GSD before

GSD after

ε

2.99 2.99 2.99 2.99 2.80 2.80 2.99 2.80 2.80

0.402 0.402 0.402 0.402 0.398 0.398 0.402 0.398 0.398

2.18 1.96 2.23 1.97 1.84 1.73 1.92 1.77 1.67

0.346 0.269 0.385 0.364 0.257 0.258 0.270 0.250 0.267

39.4 39.3 39.7 39.7 39.4 39.8 39.2 39.1 38.7

43.2 44.3 44.3 42.9 44.1 44.3 44.0 44.5 44.1

1.5 1.5 1.5 1.5 1.6 1.6 1.6 1.6 1.6

1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5

27% 35% 26% 34% 34% 38% 36% 37% 40%

a ε ) Collection efficiency (including agglomeration), N(tot) ) particle number concentration, σ ) standard deviation of particle number concentration.

flue gas flow was designed to be laminar so that the temperature- and the H2O vapor pressure gradients span the entire radius of the tube. Laminar flow also prevents the mixing of aerosol particles which would reduce the deposition. The experimental setup is schematically described in the Supporting Information (SI). The combustion device was a Wodtke pellet stove with a nominal output of 8.0 kW. One significant characteristic of the used stove is its relatively high air-to-fuel ratio (3:5). We used a commercial pellet with a calorific value of 20.25 MJ/kg, 6.9% water content and 0.34% ash content as a fuel. Detailed descriptions of the stove and the commercial pellet are given by Sippula et al. (11). There were two variables in our experiment; the volumetric H2O vapor concentration of the flue gas and the volume flow of the coolant water. Changes in temperature and H2O vapor concentration of the flue gas can both be represented using volume flow of the cooling water. The maximum H2O vapor concentration was limited by the FTIR (fourier transform infrared spectroscopy) analyzer restrictions. The experimental settings are listed inTable 1. The H2O vapor content of the flue gas was altered with two steam generators (Veit GmbH) with a maximum steam output of 6 kg/h each. The steam flow to the stack was first controlled with a needle valve and then heated with a tube furnace filled with aluminum oxide pellets. Due to the brief residence time of steam inside the furnace, it was not possible to control the inlet temperature of the moistened air. This reduced the range in which removal efficiencies could be compared in our model. The cooling water was taken from the municipal water supply system. By means of pressure control and needle valves, the flows of the hot and the cold water were mixed to the desired temperature. The temperatures of the water were measured at the inlet and at the outlet of the CHX with k-type thermocouples. The water flowed in the opposite direction to the flue gas flow to achieve maximal heat recovery. The flue gas velocity in the stack was measured in duplicate at the beginning and at the end of the campaign with SF6 marker (12). The measured average velocity in the stack varied from 0.46 to 0.65 m/s, resulting in velocities from 1.92 to 2.74 m/s inside CHX tubes. The corresponding Reynolds number is approximately 900. In the modeling part a computational value was used for the flue gas velocity in each data point. The H2O, O2, and CO2 concentrations of the flue gas were analyzed both before and after the heat exchanger with a Gasmet FTIR gas analyzer. The sampling for the FTIR was equidistant from the CHX inlet and outlet. In addition, the temperatures of the flue gas at the inlet and at the outlet were measured in a similar way with k-type thermocouples. Particle sampling was undertaken immediately before and approximately 1 m (undisturbed length) after the CHX, which 6270

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 43, NO. 16, 2009

FIGURE 1. Size distributions before and after the CHX with collection efficiency according to the particle size for the data point with lowest modeled total collection efficiency (L1, see Table 1). Error bars represent standard deviation of the individual size channels.

FIGURE 2. Size distributions before and after the CHX with collection efficiency according to the particle size for the data point with highest modeled total collection efficiency (H3, see Table 1). Error bars represent standard deviation of the individual size channels. corresponds to 5.5 stack diameters undisturbed after the CHX. The samples were diluted immediately after extraction with a porous tube diluter followed by an ejector diluter (13). The dilution ratio was calculated from the CO2 concentration values measured both from inside the stack and in the sample line. The dilution air for the porous tube diluters was heated to 70 °C to prevent condensation. After dilution, the aerosol was analyzed with a TSI fast mobility particle sizer (FMPS model 3091). The sampling source was changed with a three-way valve at the same rate as the FTIR sampling source. Each experimental setting, we took three parallel and continuous 20 min samples. During this time, several FTIR and FMPS spectra at a 5 s averaging interval were collected with the sampling source remaining the same for approximately 3-5 min. The results were calculated to volume concentrations under NTP-conditions (Nm3; 0 °C, 101.325 Pa) with zero humidity and O2-reduction to 6% of residual oxygen. The CO2-data used in the dilution ratio (DR) calculation was found to contain some delay and noise due to the analyzers,

TABLE 3. The Modeled and Measured Results of Temperatures (T), H2O Vapor Concentrations (CH2O (g)) and Collection Efficiencies (ε, Includes Agglomeration) from Different Experimental Settings before and after CHX set-up

measured/modeled

Tgas in (°C)

Tgas out (°C)

TH2O (l) out (°C)

TH2O (l) in (°C)

CH2O (g) before (vol-%)

CH2O (g) after (vol-%)

ε (%)

H1

meas. mod. meas. mod. meas. mod. meas. mod. meas. mod. meas. mod. meas. mod. meas. mod. meas. mod.

164.3 164.3 170.4 170.4 175.2 175.2 145.1 145.1 150.9 150.9 164.3 164.3 148 148 156.3 156.3 137.3 137.3

29 28 30.4 31 48.9 47 24.8 27 27.4 32 57.4 55 29.1 33 34 37 60.3 61

14.6 14.6 17.1 17.1 25.4 25.4 18.1 18.1 21.3 21.3 37.7 37.7 23 23 30.1 30.1 52.9 52.9

11 10.9 10.9 11.3 10.5 10.2 11.8 11.5 9.6 9.6 10.5 10.7 11.7 11.5 9.5 11.8 10.4 10.9

6.2 6.2 5.4 5.4 5.6 5.6 16 16 16.3 16.3 16.1 16.1 28.3 28.3 28.6 28.6 28.2 28.2

2 2 2.1 2.2 3.6 4.2 2.5 2.2 2.6 2.5 8 7.8 3.9 3.2 4.9 3.9 13.8 12.5

36 34.6 34 33.4 27 28.4 36.7 37.8 34.3 36.1 34.6 29.9 40.4 43.3 38.2 43.1 25.6 32.6

M1 L1 H2 M2 L2 H3 M3 L3

resulting in a significant error in the real time DR. The issue was solved by denoising the data (6th degree Daubechy wavelet decomposition used as a low-pass filter (14)) of the used CO2 analyzers. The arithmetic means and standard deviations of the particle size data (FMPS) were calculated with the statistical bootstrapping method (15). All individual size spectra during one setup were considered to be individual samples, resulting in over 200 for each size channel and total concentration. The mean and standard deviation were resampled 5000 times, and medians of these distributions were used as the mean and the standard deviation of the mean in the presented results. Modeling. For modeling, a composite Eulerian model was constructed where the heat exchanger is divided into computational cells. The model consists of two parts describing the thermo- and the particle- dynamics inside the CHX. In the particle dynamics part the equations used have all been individually validated in previous studies (8, 16, 17) but there are no simultaneous measurements for these mechanisms in heat exchangers. In order to calculate the temperature- and condensation profiles of the fluids in the heat exchanger, a system of, nonlinear equations describing heat- and mass transfer was devised (presented in the SI). The system of equations is solved numerically in each computational cell using Matlab’s Trust-Region Dogleg method. For simplicity, the latent heat released from condensation is presumed to transfer immediately to the cooling water and the effects of the condensed water film and fouling are neglected. The results of the thermodynamic model are used for the calculation of the particle dynamics. The fraction of particles

that is deposited is calculated in each cell. The deposition mechanisms included in this model are Brownian diffusion, thermo-, and diffusiophoresis. The particle number concentration is reduced also by agglomeration. All of the deposition mechanisms are presumed to be additive without any interactions. In the model, the measured aerosol size distribution is used as an input. A moving sectional approach is applied so that a log-normal size distribution with 32 sections is fitted in the measurements and the changes in the number concentration are calculated for each size bin. Moving size bins are used to take into account the effect of agglomeration without having to move particles between the size bins. In the calculation of the reduction of the aerosol number concentration, the particles are presumed to be evenly distributed at the inlet of the exchanger. Since the gas flow inside the heat exchanger tubes is laminar, the distance the particles have traveled toward the tube wall defines how many particles in the volume have been deposited. Particles are assumed to adhere when they encounter the tube wall. The equation used for thermophoretic velocity (m/s) is (16):

Vtherm )

-KTυ∇Tr , Tp

(1)

where the radial temperature gradient is computed from the following relation:

FIGURE 3. Modeled (left) and measured (right) reduction in aerosol number concentration with respect to flue gas humidity and temperature difference across the heat exchanger. VOL. 43, NO. 16, 2009 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6271

∇Tr )

Nuhot(Thot - Twall) . dinner

(2)

Here, T is temperature (K), d diameter (m), Nu the Nusselt Number (dimensionless), υ kinematic viscosity (m2/s), and KT thermophoretic coefficient (∼0.55). Subscript “hot” refers to flue gas, “wall” to CHX tube wall, “inner” to inside of CHX tube, “r” to radial and “p” to particle. The same definitions are used in all equations. Diffusiophoretic velocity (m/s) is given as (18) Vdiph ) VStephan-Vdiffusion )

1 xst√Mst + xnc√Mnc

(

)

RThot mcond , M AinnerP



st

(3)

ξ)

0.8191e-3.657µ + 0.0975e-22.3µ + 0.0325e-57µ, µ g 0.0312 1 - 2.56µ2/3 + 1.2µ + 0.177µ4/3, µ < 0.0312

,

(4)

where the diffusion coefficient D and the dimensionless deposition parameter µ are given as µ)

D)

tD , dinner 2 2

( )

kbThotCc . 3πηdp

(5)

(6)

Here t is time (s), kb the Boltzmann constant (m2 · kg · s-2 · K-1), Cc the Cunningham slip correction factor (dimensionless) and η dynamic viscosity (Pa · s). As the particles in the heat exchanger are in the solid phase, the effect of agglomeration was included in the model. For the treatment of agglomeration, we assumed that the particle size can be represented with the self-preserving size distribution. The measured geometric standard deviation

FIGURE 4. Deposition efficiency (mass-%) as a function of initial gas water concentration (5-45 vol-%) and temperature (400 and 800 °C). The temperature of cooling water was increased from 10 to 25 °C while gas temperature decreased to ∼15 °C. Water vapor concentration after CHX was