COLLECTION OF SUBMICRON PARTICLES IN ELECTROSTATIC PRECIPITATORS: INFLUENCE OF EHD AGITATION AND OF PARTICLES DISINTEGRATION

Transcription

1 COLLECTION OF SUBMICRON PARTICLES IN ELECTROSTATIC PRECIPITATORS: INFLUENCE OF EHD AGITATION AND OF PARTICLES DISINTEGRATION PIERRE ATTEN Key words: Electrostatic precipitation, Fine particles, Electrohydrodynamics (EHD), EHD turbulence, Electrical disintegration of aggregates. The paper focuses on two particular phenomena which contribute to the poor collection efficiency of fine particles (from ~.1 µm to ~ 2 µm) by electrostatic precipitators (ESPs). Firstly, the influence of turbulence on collection rate is considered as well as the fact that the charged fine particles are expected to give the main contribution in the generation of small eddies in ESPs. Experiments on collection efficiency η of cigarette smoke show that η decreases as the dust concentration is increased, due to the subsequent increase in eddy diffusivity. The second phenomenon observed in the study is the deagglomeration of agglomerates in the ESP. An agglomeration process occurs, which is more and more marked as the smoke concentration is raised. Once charged in the ESP, some of the agglomerates disintegrate under the Coulomb repulsion, which leads to an apparent collection efficiency strongly depending on the particle size. 1. INTRODUCTION Industrial one-stage electrostatic precipitators (ESPs) typically consist in a series of grounded parallel plates through which the gases to be treated are flowing. Equally spaced ionising electrodes raised at high negative potential are located midway between the plates. The collection of particles involves the ionisation of the gas, the charging of particles, their migration towards the collecting electrodes and the removal of the dust layers from the collection surfaces. Most industrial ESPs treating very large fluxes of flue gases have a collection efficiency in terms of mass higher than 99 or 99.9 % [1]. However the non-captured part consists mainly of fine particles as can be seen on Fig. 1 showing the proportion of particles issuing from the ESP. These fine particles are hazardous for health and more stringent regulations concerning air pollution control make it necessary to substantially increase their removal. G2Elab, CNRS, Univ. Joseph Fourier and INP Grenoble, BP 166, 3842 Grenoble Cedex 9, France pierre.atten@grenoble.cnrs.fr Rev. Roum. Sci. Techn. Électrotechn. et Énerg., 55, 2, p , Bucarest, 21

2 162 Pierre Atten 2 Fig. 1 Measurement on industrial sites of penetration versus the particle diameter (from [2]). Fig. 2 Normalised size distributions in mass and in number of ashes collected in the last field of the ESP of a coal-fired power plant. Some phenomena which contribute to make the collection efficiency η f so low are examined here. The first reason for the low collection rate of fine particles is their small drift velocity. The second factor which influences η f is the degree of turbulence of the gas flow. This is well known but the origin of the turbulence is not so clear and authors who take into account the electrical forces usually consider the action of the field on the ionic space charge only. Here, we consider the effect of the space charge of fine particles. The third phenomenon which might play a role in the collection of fine particles is the disintegration of agglomerates. 2. PARTICLES CHARGING AND DRIFT VELOCITY The particles in flue gases of coal-fired power plants have a size extending over four decades, from typically 2 nm to 2 µm [3]. The size distribution is characterised by several modes depending on the coal properties. Generally there is a mode of coarse particles [3] which originate from the ash mineral inclusions in the coal. Most often another mode has d p values in the range of micrometers as can be seen on Fig. 2 relative to the ashes collected in the last field of the ESP of a coal-fired power plant in France. The submicron particles in the flue gas have a size distribution centred at about.1 µm and they are dendritic clusters consisting of partly sintered primary ultra-fine particles of diameter ~ 2 3 nm [4]; the ultrafine primary particles are generated by complex chemical processes [4]. The low efficiency of ESPs for fine particles (with d p ranging from.1 to 2 µm) is mainly due to the rather low drift velocity w E of these particles with respect to the gas. Two mechanisms are usually distinguished for particle charging: field charging giving a saturation charge q p,sat E d p 2 [5], E being the field, and diffusion charging leading to a charge value q p d p and having a logarithmic dependence on time [6]. In the diameter range.1 to 2 µm, the two mechanisms

3 3 Collection of submicron particles in electrostatic precipitators 163 interplay and their sum gives an upper bound for q p (Fig. 3). In practice, in this size range, as a first approximation many authors retain the empirical formula of Cochet [7]. qp (in elementary charges e ) qsat qdiffusion qsat + qdiffusion qtotal(cochet) Fig. 3 Charge of spherical particles (in elementary charges e) as a function of their diameter (ε r = 4, T = 43 K (13 C), E = 4 kv/cm, t charging = 1 s). Fig. 4 Drift velocity of spherical particles as a function of their diameter (ε r = 4, T = 43 K (13 C), E = 4 kv/cm, t charging = 1 s). Under a field E the particles with charge q p experience the force F e = q p E which tends to move them toward the collecting electrodes. The drift velocity with respect to the gas is determined by the viscous force F v compensating for F e ; as the molecular mean free path λ g is not much smaller than d p, the friction force has to be divided by the Cunningham factor Cu [6]. The particle drift velocity w E has a non trivial dependence on d p as can be seen on Fig. 4. In typical conditions for ESPs, w E for fine particles (d p from.1 to 2 µm) ranges between 4 and about 1 cm/s. 3. INFLUENCE OF GAS TURBULENCE The flow of the flue gas in the ducts is always turbulent (Reynolds number Re ~ 1 4 > Re crit 2,). But the turbulent rate measured in ESPs is higher than in the pure flow itself [8]. This is not surprising because the electrical forces acting on the space charge can generate a strong agitation of the gas. The fine particles convected by the mean flow are very sensitive to the velocity fluctuations u of the gas. Indeed, for particles with w E < 1 cm/s, the fluctuations u which are high enough can entrain them in the direction opposite to the electric force at some places. This intricate problem can be simplified by using the concept of eddy diffusion of the particles [9] characterised by the coefficient D t. The mixing property of turbulence results in a tendency to make the distribution

4 164 Pierre Atten 4 of particles concentration more uniform in the z direction (normal to the plates). The Deutsch model considering a uniform distribution c in the z direction is equivalent to an infinite value of the z-component of turbulent diffusivity (Fig. 5). The model developed by Leonard et al. [9], based on finite and uniform value of D t showed that the z distribution of particles concentration c becomes more and more uniform as D t is increased (see the qualitative illustration of Fig. 5). The flux of collected particles is proportional to the product c(x,z=d) E(x, z = d), where c(x, z = d) is the particles concentration on the plate. From Fig. 5 it is clear that higher turbulence intensity will result in lower collection efficiency. In practice, the pertinent parameter controlling the particles collection rate is the electric Peclet number Pe [9] defined by: we( dp) d Pe ( d p ) =. (1) D Pe reflects the effect of eddy diffusivity which tends to counteract the action of the field inducing the drift velocity w E and driving the particles towards the collecting plates. The influence of Pe on collection rate is qualitatively illustrated by Fig. 6. t Collection efficiency η Electric Peclet number Pe Fig. 5 Concentration of charged particles between the symmetry plane and the collecting plate (at distance x from the inlet) for different D t values. Fig. 6 Typical variations of the fractional collection efficiency η as a function of the electrical Peclet number Pe. 4. TURBULENCE GENERATION BY FINE PARTICLES Generally an intense movement of the fluid is induced by the action of the field E on the space charge of density ρ resulting from corona effect in gases and from ions injection by one of two plane parallel electrodes in liquids [1]. The primary electrohydrodynamic (EHD) phenomenon in the latter case of injected space charge is the strong positive coupling between fluctuations u of the fluid

5 5 Collection of submicron particles in electrostatic precipitators 165 velocity field u and perturbations of the space charge density ρ [11]. This coupling leads to an instability when the action of the electric force of density ρe overcomes the viscous damping, i.e. above a critical value T c of the EHD parameter T [11]: εvappl T = > Tc. (2) K µ Here ε denotes the fluid permittivity, V appl the applied voltage, K the mobility of charge carriers and µ the dynamic viscosity. For strong injection of ions in air at ambient temperature, (V appl ) c 4 kv but for charged fine particles of much lower mobility, (V appl ) c ~ a few hundreds of volts [11, 12]. It is therefore clear that an important density of charged fine particles can induce instabilities and, in practice, can create eddies of small size. The amplitude of the induced velocity fluctuations is of the order of (ε /ρ f ) 1/2 E where ρ f is the fluid density and E the mean field value [1, 11]. The second basic non dimensional EHD number M is defined by: ε / ρf M =. (3) K For M << 1, the amplirude of fluctuations of fluid velocity u are much smaller than the drift velocity of charge carriers w E = KE; then the trajectories of the charge carriers are only very slightly perturbed by the fluid motion which has no significant influence on the distribution of ρ (case of ions in air: w E ~ 1 m/s >> electric wind u ~ 1 m/s). For M >~ 3 (u > KE), the charge carriers can be entrained by the fluid flow that they contribute to create and there is a significant increase of mean current density [1, 11]. The two parameters T and M are also pertinent for more complex EHD flows as obtained in asymmetric electrode configurations [12]. In the typical conditions of ESPs, the mobility parameter M relative to the particles, depends on the particles size [12]. For fine particles charged under a typical mean field E = 5 kv/cm, M > 3 [12]; therefore a turbulent mixing can be expected, which partly redistributes ρ. The above considerations are very general and concern problems with only one type of charge carriers. In ESPs both ions and charged dust particles contribute to the space charge and have distributions which are inter-related. For sustained corona discharges in gases, from Kaptzov hypothesis, the field E wire on the ionising wire(s) takes the value it has at inception voltage V th. For a clean gas (no dust), the global charge density ρ% total affecting the field distribution is due to ions only: ( ~ ρ ) ~ ( ~ total = ρions = ρions ) ( Vappl Vth ), clean (4) ( ρ ~ ions = global charge density due to ions). For a dusty gas, as the field E wire remains the same whatever the dust concentration c at the ESP inlet, we have:

6 166 Pierre Atten 6 ( ρ total ) = ρ ions +ρ p = = ( ρions ) dusty % % % cst %, (5) ρ ~ p being the global space charge density of particles. The total current I also has two components, I ion and I p ( I = I ion + I p ); the current densities write: j ions = K ρ E, j = K ρ E, (6) ions ions where K ions, ρ ions and K p, ρ p are the ions and particles mobilities and charge densities respectively. As the ratio K p /K ions ~ 1-3, for a dusty gas we have: p ( ), I I ions ρ% ions ρ% ions ρ% p (7) and I should decrease with the dust concentration c. In practice the relative current decrease gives an estimate of the mean space charge of charged particles: ρ% p ρ% ions I I 1. (8) For fine particles, except just around the wires, the particles drift velocity is lower than the gas velocity so that the distribution of ρ p is controlled mainly by the flow which, itself, is determined by the total space charge distribution ρ ion + ρ p. When ρ p becomes comparable to ρ ion a marked change in the mean gas flow pattern is expected along with a significant rate of turbulence induced by ρ p [13]. p p 5. INFLUENCE OF PARTICLES CONCENTRATION ON COLLECTION EFFICIENCY 5.1. EXPERIMENTAL SET-UP The study of fine particles collection was carried out using a laboratory ESP characterised by a duct of width 2d = 9 cm and height h = 29 cm. The test cell of total length L = 1 cm was divided in two sections of 5 cm in length. Initially a set of 9 equidistant ionising wire electrodes was used to create the space charge in the first section. In the second section, a plate was introduced midway and could be polarized to create a uniform field of collection E 2 in the two half-ducts. The efficiency of particles collection was determined from concentrations measurements with and without applied voltage, using an optical counter (TOPAS LAP32) on scales ranging from.31 µm to 2 µm. The study was performed with cigarette smoke filling a balloon of about 2.5 m in diameter (volume ~ 8 m 3 ). Starting with the empty balloon, the air pushed by a fan flows through a paper filter, then through a box where some cigarettes are burning and progressively fills the balloon. Once the balloon is full, the accumulated smoke is directed to the test cell. As the optical counter has a lower

7 7 Collection of submicron particles in electrostatic precipitators 167 bound for the particles diameter of.31 µm, French cigarettes (trademark: Gauloises) were used which generate particles of not too small size (99% of particles with d p < about 1.5 µm). Fig. 7 shows the size distributions for different numbers of burnt cigarettes. For 8 and 16 burnt cigarettes, the curves exhibit a maximum for d p.4 µm and.55 µm respectively. The increase of mean size of particles results from an agglomeration of smaller particles cigarette 2 cigarettes 4 cigarettes 8 cigarettes 16 cigarettes 1 1 Vappl = Vappl = 24 kv Vappl = 26 kv Concentration (cm -3 ) 1 1 Concentration (cm -3 ) Fig. 7 Size distributions (in number) of particles of "Gauloises" cigarette smoke for different numbers of burnt cigarettes. Fig. 8 Size distributions of smoke particles (16 burnt cigarettes) for two applied voltages (9 ionising wires) RESULTS With the 9 ionising wires (φ = 1 mm) perpendicular to the horizontal air flow of mean velocity U =.4 m/s (E 2 = in the 2 nd section), the measurements were performed by counting the particles for successive sequences with and without applied voltage (Fig. 8). The fractional efficiency of collection of particles η f as a function of their size d p was determined from the distributions as shown in Fig. 8. The major interesting fact visible in Fig. 9 is the clear tendency of η f to decrease as c is increased. Note that the influence of c is most marked in the range.31 to.5 µm. The low η f values obtained for 16 burnt cigarettes partly arise from the disintegration of charged agglomerates under the influence of Coulomb repulsion. The efficiency decrease with the particles concentration is presumably due to the increase in secondary flow and turbulence intensities. The results appear to be consistent with those obtained in [13]. In particular a marked decrease in discharge current was obtained (Fig. 1). For the highest smoke concentrations, the space charge due to the charged particles represents 3 to 4% of the total space charge.

8 168 Pierre Atten 8 Collection efficiency η f 1 8 (%) cigarette 2 cigarettes 4 cigarettes 8 cigarettes 16 cigarettes Ratio I / I Vappl = 24 kv.2 Vappl = 26 kv Particles concentration c (mg/m 3 ) Fig. 9 Fractional efficiency of collection versus particles diameter for various smoke concentrations c (9 ionising wires, V appl = 26 kv). Fig. 1 Ratio of the currents I for a η f flow of smoke and I for filtered air (U =.4 m/s) as a function of c (9 ionising wires). 6. PARTICLES DEAGGLOMERATION In order to have η f measurements in conditions comparable to those of ref. [13], the 9 wires were replaced by a unique ionising wire (φ =.5 mm) located at x = 21 cm from the inlet. The study of size distributions was performed for U = =.4 m/s and V appl = 23.5 kv. The curves of Fig. 11 clearly show that, in the range.31 to.5 µm, the fractional collection efficiency η f tends to decrease when the particles concentration c is increased; conversely it tends to increase for d p > ~.6 µm. An interesting observation is that for high concentrations (8 and 16 burnt cigarettes), negative values of η f are obtained, which means that, at the outlet, there are more very fine particles than at the inlet of the test cell. Fig. 7 clearly indicates that the mean size of the particles increases as the smoke concentration is increased. This implies a noticeable agglomeration of the fine particles in suspension. Now, when an aggregate captures electric charges, these charges distribute over the different elements constituting the aggregate which then experiences a Coulomb repulsion so that the aggregate can disintegrate. This process of deagglomeration of aggregates explains why there are so many small particles issuing from the ESP and also why the relative number of the biggest particles at the outlet decreases when the smoke concentration is increased (this corresponds to an increase in effective fractional collection efficiency η f ). The process of deagglomeration of aggregates was also present when working with 9 ionising wires ( 5.2); but the addition of the influences of the successive wires partly damped the effect exhibited by the unique wire experiment.

9 9 Collection of submicron particles in electrostatic precipitators 169 Collection efficiency ηf (%) cigarette 4 cigarettes 16 cigarettes Collection efficiency ηf2 (%) E 2 = 5 V/cm 1 cig. 4 cigs. 8 cigs. 16 cigs Fig. 11 Fractional collection efficiency η f (d p ) in the first section with a unique ionising wire for 3 different smoke concentrations (V appl = 23.5 kv). Fig. 12 Fractional collection efficiency η f2 (d p ) in the second section of the test cell for different c values (applied uniform field E 2 =.5 kv/cm). In order to eliminate the influence of deagglomeration of aggregates when determining η f, we performed measurements in the second section of the test cell where the field is uniform, where there is no particle charging (no ionic space charge) so that, presumably, no disintegration of aggregates occurs. A voltage V appl = = 2 kv was applied permanently on the ionising wire in the first section of the ESP. In the second ESP section a field E 2 was applied and the efficiency η f2 of the second section was determined from the size distributions without and with field E 2. The results of Fig. 12 (E 2 =.5 kv/cm, U =.4 m/s) confirm that the efficiency decreases with the smoke concentration c for the whole diameter range in which the measurements are significant. The difference of these results with those relative to the first section of the test cell ( 5.2) is clear and can be ascribed to the absence of disintegration of aggregates in the second section. The decrease of η f2 as c is increased is consistent with the picture of the turbulence intensity increasing with the mean charge density associated to the charged fine particles. 7. CONCLUSIONS This study performed on cigarette smoke shows that there is an important effect of agglomeration of fine particles when the smoke concentration is high enough. This agglomeration is presumably due to collisions induced by the Brownian motion. The rather unexpected phenomenon observed is the deagglomeration accompanying the charging of the aggregates by the ions created by corona discharges. Such a disintegration process implies rather limited adhesion forces between the particles constituting the aggregates. The presented results

10 17 Pierre Atten 1 suggest that this process might play a role in electrostatic precipitation and might partly explain the difficulties in collecting the fine particles. By taking into account the influence of disintegration of aggregates occurring in the fist section of the ESP, the presented measurements show that the fractional collection efficiency η f of the particles decreases as their concentration c is increased. These results provide an indirect proof of the increase of the turbulence rate and give a further support to the conjecture [12] that an important source of turbulence in ESPs is "intrinsic", through the action of the space charge associated with the fine particles. Received on 3 February, 29 REFERENCES 1. K.R. Parker, Electrostatic precipitation, Chapman & Hall, N. Plaks, Improving collection of toxic fine particles in ESPs, Proceedings VI ICESP, Budapest, June 1996, pp Z. Chengfeng, Y. Qiang, S. Junming, Characteristics of particulate matter from emissions of four typical coal-fired power plants in China, Fuel-processing-technology, 86, 7, pp , M. Thellefsen Nielsen, H. Livbjerg, Formation and emission of fine particles from two coalfired power plants, Combust. Sci. & Techno., 174, pp , M. Pauthenier, M. Moreau-Hanot, La charge des particules sphériques dans un champ ionisé, J. Phys. et Radium, 86, 3, pp , H.J. White, Industrial electrostatic precipitation, Wesley Publishing Company Inc., R. Cochet, Lois de charge des fines particules (submicroniques). Etudes théoriques Contrôles récents. Spectres de particule, Proc. Colloque Intern. n 12 "La physique des forces électrostatiques et leurs applications", CNRS, Paris, 1961, pp G.L. Leonard, M. Mitchner, S.A. Self, An experimental study of the electrohydrodynamic flow in electrostatic precipitators, J. Fluid Mech., 127, pp , G. Leonard, M. Mitchner, S.A. Self, Particle transport in electrostatic precipitators, Atmosph. environment, 14, pp , J.-C. Lacroix, P. Atten, E. Hopfinger, Electroconvection in a dielectric liquid layer subjected to unipolar injection, J. Fluid Mech., 69, pp , P. Atten, Electrohydrodynamic instability and motion induced by injected space charge in insulating liquids, IEEE Trans. Diel. & Electr. Insul., DEI3, pp. 1 17, P. Atten, F.M.J. McCluskey, A.C. Lahjomri, The electrohydrodynamic origin of turbulence in electrostatic precipitators, IEEE Trans. Ind. Appl., IA-23, 4, pp , J. Podlinski, A. Niewulis, J. Mizeraczyk, P. Atten, ESP performance for various dust densities, J. Electrostatics, 66, pp , 28.