Characterization and Early Detection of Bed Agglomeration during the

120

Energy & Fuels 2006, 20, 120-132

Characterization and Early Detection of Bed Agglomeration during the Fluidized Bed Combustion of Olive Husk Fabrizio Scala* and Riccardo Chirone Istituto di Ricerche sulla Combustione-Consiglio Nazionale delle Ricerche, Piazzale Vincenzo Tecchio, 80, 80125 Napoli, Italy ReceiVed July 27, 2005. ReVised Manuscript ReceiVed October 14, 2005

The fluidized bed combustion of a biomass residue (olive husk) common in the Mediterranean area was investigated in a bench-scale reactor. The focus of the study was the high propensity of this fuel to have bed agglomeration problems during combustion as a consequence of the high potassium content of the ash. Temperature and pressure profiles in the bed were followed as a function of time during steady combustion tests at different operating conditions. Bed defluidization characteristic times were measured and correlated to the fuel ash buildup on the bed sand particles. In addition, a diagnostic tool based on the measurement of the dynamic pressure signal inside the bed was tested for its ability to predict bed agglomeration. On the basis of SEM/EDX analysis of agglomerate samples discharged from the bed after defluidization had occurred, the mechanisms of fuel ash-bed particle interaction and agglomerate formation are discussed.

Introduction Since the early eighties, fluidized bed combustion (FBC) of low-grade coals (mostly lignite1-4 and in one case anthracite5) was reported to result in bed agglomeration problems. In the most severe cases, agglomeration resulted in the total bed defluidization and combustor shutdown.1,3 These problems were recognized to be connected to the large amounts of alkali metals (sodium and potassium) included in the fuel ash. At the typical FBC operating temperatures, alkali metals were found to build up on the inert bed-particle surface (typically silica sand or limestone) where low-melting-point eutectics with silica and calcium were formed, as soon as the alkali concentration reached a large enough value. Melting of material at the surface of the inert particles enhanced their stickiness, leading to the attachment of an increasing number of bed particles together and to the formation of large agglomerates. Alkali form and content in the fuel ash, the nature of the inert bed material, the combustion time and the temperature were all found to be key param* Corresponding author. Phone: +39 081 7682969. Fax: +39 081 5936936. E-mail: [email protected]. (1) Goblirsch, G.; Vander Molen, R. H.; Wilson, K.; Hajicek, D. Atmospheric Fluidized Bed Combustion Testing of North Dakota Lignite. In Proceedings of the 6th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1980; pp 850-862. (2) Rice, R. L.; Shang, J. Y.; Ayers, W. J. Fluidized-Bed Combustion of North Dakota Lignite. In Proceedings of the 6th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers; New York, 1980; pp. 863-871. (3) Goblirsch, G. M.; Benson, S. A.; Hajicek, D. R.; Cooper, J. L. Sulfur Control and Bed Material Agglomeration Experience in Low-Rank Coal AFBC Testing. In Proceedings of the 7th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1982; pp 1107-1120. (4) Bobman, M. H.; Hajicek, D. R.; Zobeck, B. J. A Study of Bed Agglomeration Resulting from the AFBC of Low-Rank Coals. In Proceedings of the 8th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1985; pp. 13991407. (5) Rozelle, P. L.; Scaroni, A. W. Bed Agglomeration and Ash Disposal for Anthracite-Fired Fluidized-Bed Combustors. In Proceedings of the 9th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1987; pp 284-291.

eters in the agglomeration process.1-4,6-7 Manzoori and Agarwal8-10 studied the FBC of high-sodium lignite and shed new light on the agglomeration mechanism. They found that, independent of the fuel particle size and temperature, sodium vaporization was limited under FBC conditions while inorganic elements in the ash tended to form a molten matrix on the char surface. This ash transferred to the surface of the inert bed particles most likely as a result of random collisions in the bed. No evidence was found to suggest that ash deposition was caused or initiated by chemical reaction or by vapor condensation. Interestingly, the bed temperature was not found to significantly affect the composition of the ash coating, but it was found to alter its physical properties (e.g., viscosity). The rate of ash deposition on the bed particles was typically constant during the combustion runs, dependent only on the fuel ash alkali content and the bed temperature. Moreover, a critical thickness of the ash coating was necessary to start the agglomeration process. In the last two decades, the declining conventional energy supplies and pressing environmental constraints have directed a growing interest to the exploitation of biomass resources as a renewable and CO2-neutral energy source. The attractiveness of biomass has, in turn, prompted research on technologies suitable for burning this class of fuels. Among the others, FBC was indicated as one of the most promising ones because of its fuel flexibility, high combustion efficiency, and low environmental impact. However, agglomeration/defluidization problems were often reported,11-18 even at operating temperatures as low (6) Rizeq, R. G.; Shadman, F. Alkali-Induced Agglomeration of Solid Particles in Coal Combustors and Gasifiers. Chem. Eng. Commun. 1989, 81, 83-96. (7) Ataku¨l, H.; Ekinci, E. Agglomeration of Turkish Lignites in FluidisedBed Combustion. J. Inst. Energy 1989, 62, 56-61. (8) Manzoori, A. R.; Agarwal, P. K. The Fate of Organically Bound Inorganic Elements and Sodium Chloride during Fluidized Bed Combustion of High Sodium, High Sulphur Low Rank Coals. Fuel 1992, 71, 513-522. (9) Manzoori, A. R.; Agarwal, P. K. The Role of Inorganic Matter in Coal in the Formation of Agglomerates in Circulating Fluid Bed Combustors. Fuel 1993, 72, 1069-1075. (10) Manzoori, A. R.; Agarwal, P. K. Agglomeration and Defluidization under Simulated Circulating Fluidized-Bed Combustion Conditions. Fuel 1994, 73, 563-568.

10.1021/ef050236u CCC: $33.50 © 2006 American Chemical Society Published on Web 11/15/2005

Fluidized Bed Combustion of OliVe Husk

as 650 °C,15 especially when crops cultivated with a high degree of fertilization were burned, as these residues generally contain higher amounts of alkali in their ash. Operational experience showed that bed defluidization was accompanied by a sharp drop in the differential pressure across the bed, as a result of channeling through the bed by fluidizing air.13 In addition, a substantial decrease of the in-bed temperatures (because of the lack of combustion in the bed) and a significant increase of the temperature in the lower freeboard section (because of the biomass combustion on the top of the agglomerated bed) were observed upon bed defluidization making the operation uncontrollable.13,15 Despite considerable research efforts devoted to this subject, the mechanisms of biomass ash-bed material interaction leading to bed agglomeration are still not well understood. Ghaly et al.19,20 simulated the high-temperature interaction between straw ash and either silica or alumina sand in a muffle furnace. They found that at temperatures above 850 °C the straw ash interacted with silica sand forming a melt because of the presence of a low-temperature eutectic in the potassium-silica system. On the other hand, at the same temperatures alumina sand was not found to interact significantly with the straw ash. Nevertheless, at such temperatures the straw ash melted and covered the alumina particles with a sticky layer. Latva-Somppi et al.21 and Valmari et al.22 studied the mechanisms of ash deposition on quartz sand particles in full-scale fluidized bed boilers firing different biomass fuels. Results suggested that ash deposition was dominated by two parallel mechanisms: (1) the adhesion (11) Albertson, D. M. Design Characteristics of a 12 MW AFB Power Station Firing Agricultural Wastes. In Proceedings of the 10th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1989; pp 647-652. (12) Davies, C. E.; Dawson, S. G. B.; Fieldes, R. B. An Investigation of Thermal Agglomeration in Fluidized Beds. In Fluidization VIsProceedings of the 6th International Conference on Fluidization; Engineering Foundation: New York, 1989; pp 555-562. (13) Salour, D.; Jenkins, B. M.; Vafaei, M.; Kayhanian, M. Control of In-Bed Agglomeration by Fuel Blending in a Pilot Scale Straw and Wood Fueled AFBC. Biomass Bioenergy 1993, 4, 117-133. (14) Nordin, A. Optimization of Sulphur Retention in Ash when Cocombusting High Sulfur Fuels and Biomass Fuels in a Small Pilot Scale Fluidized Bed. Fuel 1995, 74, 615-622. (15) Grubor, B. D.; Oka, S. N.; Ilic, M. S.; Dakic, D. V.; Arsic, B. T. Biomass FBC CombustionsBed Agglomeration Problems. In Proceedings of the 13th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1995; pp 515-522. (16) Bapat, D. W.; Kulkarni, S. V.; Bhandarkar, V. P. Design and Operating Experience on Fluidised Bed Boiler Burning Biomass Fuels with High Alkali Ash. In Proceedings of the 14th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1997; pp 165-174. (17) Wu, S.; Sellakumar, K. M.; Chelian, P. K.; Bleice, C.; Shaw, I. Test Study of Salty Paper Mill Waste in a Bubbling Fluidized Bed Combustor. In Proceedings of the 15th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1999; pp 1574-1588. (18) Ryabov, G. A.; Litoun, D. S.; Dik, E. P. Agglomeration of Bed Material: Influence on Efficiency of Biofuel Fluidized Bed Boiler. In Proceedings of the 17th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineering: New York, 2003; pp 1-13. (19) Ghaly, A. E.; Ergu¨denler, A.; Laufer, E. Agglomeration Characteristics of Alumina Sand-Straw Ash Mixtures at Elevated Temperatures. Biomass Bioenergy 1993, 5, 467-480. (20) Ghaly, A. E.; Ergu¨denler, A.; Laufer, E. Study of Agglomeration Characteristics of Silica Sand-Straw Ash Mixtures Using Scanning Electronic Microscopy and Energy Dispersion X-ray Techniques. Bioresour. Technol. 1994, 48, 127-134. (21) Latva-Somppi, J.; Kurkela, J.; Tapper, U.; Kauppinen, E. I.; Jokiniemi, J. K.; Johanson, B. Ash Deposition on Bed Material Particles during Fluidized Bed Combustion of Wood-Based Fuels. In Proceedings of the International Conference on Ash BehaVior Control in Energy ConVersion Systems; Society of Chemical Engineers: Yokohama, Japan, 1998; pp 119-126.

Energy & Fuels, Vol. 20, No. 1, 2006 121

(promoted by particle collisions) of micron-sized ash particles on sand surface and their consecutive sintering, forming a porous layer around the particles and (2) the diffusion of alkali species (especially potassium) to the quartz interface where a compact layer was formed. The latter mechanism was imputed both to solid-state diffusion of alkali from the ash layer to the quartz core and to direct attack of quartz by volatilized alkali species. Skrifvars et al.23,24 stressed the importance of the sintering tendency by partial melting of biomass ash as a cause of bed agglomeration. They proposed a combined method based on compression strength tests and thermodynamic multicomponent multiphase equilibrium analysis for predicting the sintering tendency of the ash. The main limit of this procedure was that the interaction of the ash with the inert bed material was not taken into account. They also tested the usefulness of the standard ASTM ash fusion test, which failed to predict the experimental bed agglomeration temperatures.24 On the whole, the results suggested that ex situ methods, not recreating the real environment experienced by the ash in a fluidized boiler, are likely to be unsuitable to correctly predict the agglomeration tendency of biomass fuels. O ¨ hman and Nordin25 proposed a new controlled fluidized bed agglomeration method for biomass fuels. This method consisted of the preliminary ashing of a certain quantity of biomass in a fluidized bed followed by a temperature increase (by external heating) of the inert bedash system until bed defluidization was achieved (as detected by a drop of the differential bed pressure). The method was applied to a number of biomass fuels to rank their agglomeration tendencies during FBC.24-27 The measured agglomeration temperatures were fairly accurate and reproducible, but uncertainties remained as to how well bed agglomeration in fullscale units compares to that of simulated experiments. In particular, the local overheating near-burning char particles and the progressive accumulation of ash in the bed during steady combustion were not taken into account with this method. Further analysis of bed samples collected throughout the tests27 indicated that all of the bed particles were coated with a relatively homogeneous ash layer mainly composed of silica, calcium, and potassium in variable amounts. The outermost layer of the coating appeared to be heterogeneous, suggesting that small ash particles deposited on the partially molten coating and gradually homogenized with time. The melting behavior of the coating was very sensitive to the potassium content and was ultimately responsible for bed agglomeration. The presence of an inner layer around the bed particles composed mainly of alkali and silica and of an outer layer resembling the fuel ash composition was recently confirmed by extensive SEM-EDX (22) Valmari, T.; Lind, T. M.; Kauppinen, E. I.; Sfiris, G.; Nilsson, K.; Maenhaut, W. Field Study on Ash Behavior during Circulating FluidizedBed Combustion of Biomass. 1. Ash Formation. Energy Fuels 1999, 13, 379-389. (23) Skrifvars, B.-J.; Backman, R.; Hupa, M. Characterization of the Sintering Tendency of Ten Biomass Ashes in FBC Conditions by a Laboratory Test and by Phase Equilibrium Calculations. Fuel Process. Technol. 1998, 56, 55-67. (24) Skrifvars, B.-J.; O ¨ hman, M.; Nordin, A.; Hupa, M. Predicting Bed Agglomeration Tendencies for Biomass Fuels Fired in FBC Boilers: A Comparison of Three Different Prediction Methods. Energy Fuels 1999, 13, 359-363. (25) O ¨ hman, M.; Nordin, A. A New Method for Quantification of Fluidized Bed Agglomeration Tendencies: A Sensitivity Analysis. Energy Fuels 1998, 13, 90-94. (26) Natarajan, E.; O ¨ hman, M.; Gabra, M.; Nordin, A.; Liliedahl, T.; Rao, A. N. Experimental Determination of Bed Agglomeration Tendencies of Some Common Agricultural Residues in Fluidized Bed Combustion and Gasification. Biomass Bioenergy 1998, 15, 163-169. (27) O ¨ hman, M.; Nordin, A.; Skrifvars, B.-J.; Backman, R.; Hupa, M. Bed Agglomeration Characteristics during Fluidized Bed Combustion of Biomass Fuels. Energy Fuels 2000, 14, 169-178.

122 Energy & Fuels, Vol. 20, No. 1, 2006

analysis on bed samples collected during biomass FBC both in lab-scale and full-scale tests.28-31 A major contribution to the understanding of the agglomeration mechanism came from the work on straw FBC carried out at the Technical University of Denmark.32-36 Systematic experiments in a lab-scale fluidized bed combustor showed that the variable which exerted the largest influence on agglomeration was the operating temperature and that the bed defluidization time decreased exponentially with the temperature. A gradual buildup of potassium in the bed was measured during the run, and interestingly, the data showed that bed temperature had no significant influence on the potassium accumulation rate. This result, together with simultaneous thermal analysis data on ash,33,37 indicated that only a small amount of the potassium compounds were evaporated during combustion. During purposely carried out batch experiments, it was noted that several agglomerates were already formed a few minutes after straw feeding had started and that they presented a blackish ash core and a shape similar to that of the straw pellets. This observation suggested that agglomerates had likely originated from potassium-rich ash transferred by collisions from the burning char particles to the sand particles. The high temperatures experienced near burning char might have further promoted potassium transfer and the subsequent melting. Two mechanisms for bed defluidization were proposed.32 (1) At low operating temperatures, an increasing number of agglomerates (consisting of a number of bed particles joined together by necks) is formed in the bed during combustion. These agglomerates would progressively build up a defluidized layer at the bottom of the bed, which would eventually lead to the total bed defluidization. (2) At high operating temperatures, instead, the stickiness of the bed particles would increase to a certain level where sudden defluidization occurs. In these conditions, the fluidization gas cannot balance drag, gravity, buoyant, and adhesive forces. Evidence reported in later studies38,39 seem to somewhat justify the above framework. Recent research has demonstrated that FBC of most biomass fuels takes place with extensive generation of carbon fines.40,41 This is a consequence of the tendency of these fuels to yield (28) Nuutinen, L. H.; Tiainen, M. S.; Virtanen, M. E.; Laitinen, R. S. Coatings on Bed Particles from FB-Combustion of Different Biomasses. In Proceedings of the 17th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 2003; pp 705-710. (29) Nuutinen, L. H.; Tiainen, M. S.; Virtanen, M. E.; Enestam, S. H.; Laitinen, R. S. Coating Layers on Bed Particles during Biomass Fuel Combustion in Fluidized-Bed Boilers. Energy Fuels 2004, 18, 127-139. (30) Brus, E.; O ¨ hman, M.; Nordin, A.; Bostro¨m, D.; Hedman, H.; Eklund, A. Bed Agglomeration Characteristics of Biomass Fuels Using Blast-Furnace Slag as Bed Material. Energy Fuels 2004, 18, 1187-1193. (31) Brus, E.; O ¨ hman, M.; Nordin, A. Mechanisms of Bed Agglomeration during Fluidized-Bed Combustion of Biomass Fuels. Energy Fuels 2005, 19, 825-832. (32) Lin, W.; Krusholm, G.; Dam-Johansen, K.; Musahl, E.; Bank, L. Agglomeration Phenomena in Fluidized Bed Combustion of Straw. In Proceedings of the 14th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1997; pp 831-837. (33) Lin, W.; Dam-Johansen, K. Agglomeration in Fluidized Bed Combustion of BiomasssMechanisms and Co-Firing with Coal. In Proceedings of the 15th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 1999; pp 1188-1191. (34) Lin, W.; Dam-Johansen, K. Modelling of Agglomeration in Straw Fired Fluidized Bed Combustors. In Proceedings of the 16th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 2001; pp 1-13. (35) Lin, W.; Jensen, A. D.; Johnsson, J. E.; Dam-Johansen, K. Combustion of Biomass in Fluidized Beds-Problems and some Solutions Based on Danish Experiences. In Proceedings of the 17th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 2003; pp 945-953.

Scala and Chirone

highly porous or even incoherent chars after pyrolysis that are rather fragile and very susceptible to attrition or fragmentation during combustion. Attrited char may experience, upon further burnoff, peak temperatures largely exceeding the bed temperature, while coarse particles burn at temperatures only slightly above the bed temperature.42,43 As a consequence, modifications of the ash constituents, such as softening, melting, or even vaporization, may occur even at nominal bed temperatures at which no such change in mineral matter would take place. The transportation of the fuel ash to the bed-particle surface by random collisions between burning fines and inert particles was proposed to be a relevant mechanism responsible for the formation of ash-layered bed material.42 Upon impact with bed particles the formation of a softened or fluid phase at the contact point may be responsible for the adhesion of fines on the bedparticle surface. In the event that the bed temperature is itself higher than the relevant eutectic temperature, fused potassiumsilicate is permanently formed yielding a sticky surface layer responsible for bed agglomeration and defluidization. If the bed temperature is, instead, below the eutectic melting temperature, the adhesion of fines can still take place, but the formation of a surface melt on the bed solids is prevented. In addition to understanding the ash-bed material interaction and agglomeration mechanisms, the possibility of early detection of incipient bed agglomeration during biomass FBC would be of extreme importance. In this way, appropriate countermeasures to prevent total bed defluidization and boiler shutdown could be taken in due time. The simple measurement of the temperature and pressure profiles inside the bed is not appropriate for this purpose because the change of these variables is significant only upon bed defluidization. On the other hand, a suitable indicator of the quality of fluidization should detect the formation of agglomerates in the bed early enough to give the boiler operator the chance to intervene on the process conditions. A technique based on a nonlinear pressure-fluctuation timeseries statistical analysis (chaos analysis) for the early detection of agglomeration in fluidized beds was recently presented.44-48 While the method was fairly accurate at lab-scale conditions, a (36) Lin, W.; Dam-Johansen, K.; Frandsen, F. Agglomeration in BioFuel Fired Fluidized Bed Combustors. Chem. Eng. J. 2003, 96, 171-185. (37) Arvelakis, S.; Jensen, P. A.; Dam-Johansen, K. Simultaneous Thermal Analysis (STA) on Ash from High-Alkali Biomass. Energy Fuels 2004, 18, 1066-1076. (38) Olofsson, G.; Ye, Z.; Bjerle, I.; Andersson, A. Bed Agglomeration Problems in Fluidized-Bed Biomass Combustion. Ind. Eng. Chem. Res. 2002, 41, 2888-2894. (39) Visser, H. J. M.; van Lith, S. C.; Kiel, J. H. A. Biomass Ash-Bed Material Interactions Leading to Agglomeration in FBC. In Proceedings of the 17th International Confernece on Fluidized Bed Combustion; American Society of Mechanical E, 2003, pp 563-570. (40) Salatino, P.; Scala, F.; Chirone, R. Fluidized Bed Combustion of a Biomass Char: The Influence of Carbon Attrition and Fines Postcombustion on Fixed Carbon Conversion. Proc. Combust. Inst. 1998, 27, 3103-3110. (41) Scala, F.; Salatino, P.; Chirone, R. Fluidized Bed Combustion of a Biomass Char (Robinia Pseudoacacia). Energy Fuels 2000, 14, 781-790. (42) Chirone, R.; Salatino, P.; Scala, F. The Relevance of Attrition to the Fate of Ashes During Fluidized-Bed Combustion of a Biomass. Proc. Combust. Inst. 2000, 28, 2279-2286. (43) Scala, F.; Chirone, R.; Salatino, P. The Influence of Fine Char Particles Burnout on Bed Agglomeration during the Fluidized Bed Combustion of a Biomass Fuel. Fuel Process. Technol. 2003, 84, 229-241. (44) Schouten, J. C.; van den Bleek, C. M. Monitoring the Quality of Fluidization Using the Short-Term Predictability of Pressure Fluctuations. AIChE J. 1998, 44, 48-60. (45) van Ommen, J. R.; Schouten, J. C.; van den Bleek, C. M. An EarlyWarning Method for Detecting Bed Agglomeration in Fluidized Bed Combustors. In Proceedings of the 15th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineering: New York, 1999; pp 1501-1513. (46) van Ommen, J. R.; Coppens, M.-O.; van den Bleek, C. M.; Schouten, J. C. Early Warning of Agglomeration in Fluidized Beds by Attractor Comparison. AIChE J. 2000, 46, 2183-2197.


Energy & Fuels, Vol. 20, No. 1, 2006 123 Table 1. Properties of Olive Husk exhausted husk

virgin husk

1250 17.5

1310 19.8

proximate analysis (wt %) moisture volatiles fixed carbon ash

13.04 56.25 26.23 4.43

8.54 73.21 16.77 1.48

ultimate analysis (dry basis) (wt %) carbon hydrogen nitrogen sulfur ash oxygen (diff)

51.76 5.46 1.25 0.09 5.09 36.35

47.06 6.82 0.82 0.11 1.62 43.57

ash composition (wt %) CaO MgO K2O Na2O Fe2O3 Al2O3 SiO2 P2O5 SO3

22.34 1.80 25.56 1.74 2.75 6.22 34.78 1.77 0.92

12.34 4.70 41.65 0.50 2.34 3.23 18.99 8.81 2.03

particle density LHV (MJ/kg)

Figure 1. Experimental apparatus: (1) thermocouple, (2) temperature PID controller, (3) preheater, (4) thermocouple and pressure transducer, (5) acquisition data unit, (6) personal computer, (7) gas analyzers, (8) condenser, (9) filter, (10) cyclone, (11) gas distributor, (12) mass-flow controller, (13) ceramic heaters, (14) windbox, (15) fuel-air mixer, (16) screw feeder, and (17) fuel hopper.

significant sensitivity to changes in operating conditions (especially fluidization velocity) was noted. In addition, the real applicability to full-scale boilers was not demonstrated. The present study was directed at investigating the FBC of olive husk, a biomass residue of the olive oil production industry, common in the Mediterranean area. This fuel has shown a high propensity for unwanted bed agglomeration problems during combustion, as a consequence of the high potassium content of the ash. The experimental work was aimed at evaluating the characteristic times of bed agglomeration, the ash accumulation in the bed during FBC, and the influence of the main operating variables (temperature, excess air, fluidization velocity, and bed-particle size) on this phenomenon. Temperature and pressure profiles along the bed height were also measured during the runs to monitor the effectiveness of bed mixing. In addition, a diagnostic tool based on the measurement of the dynamic pressure signal inside the bed was tested for its ability to predict the onset of bed agglomeration. Experimental Section Apparatus. The experimental apparatus, sketched in Figure 1, is composed of a cylindrical fluidized bed, a set of pressure transducers, a set of thermocouples, a fuel-feeding system, a gasanalysis system, and a data-acquisition unit. The reactor was made of a stainless steel tube with a inner diameter of 0.102 m and a height of 1.625 m. The vessel was fitted with a stainless steel perforated distributor plate with 518 holes of 0.5 mm in diameter disposed in triangular pitch. The windbox was packed with ceramic rings to act as a gas preheater. The distributor and the windbox packing resulted in a uniform fluidizing gas feeding into the fluidized bed. (47) van Ommen, J. R.; Schouten, J. C.; Coppens, M.-O.; Lin, W.; DamJohansen, K.; van den Bleek, C. M. Timely Detection of Agglomeration in Biomass Fired Fluidized Beds. In Proceedings of the 16th International Conference on Fluidized Bed Combustion; American Society of Mechanical Engineering: New York, 2001; pp 1146-1159. (48) Korbee, R.; van Ommen, J. R.; Lensselink, J.; Nijenhuis, J.; Kiel, J. H. A.; van den Bleek, C. M. Early Agglomeration Recognition System (EARS). Proceedings of the 17th International Confereence on Fluidized Bed Combustion; American Society of Mechanical Engineers: New York, 2003; Paper 151.

(kg/m3)

The vessel was equipped with five tubes flush to the column wall located at a vertical distance z from the distributor plate equal to 0.045, 0.095, 0.145, 0.245, and 0.775 m. The tube located at z ) 0.095 m was used for fuel feeding; those located at z ) 0.045, 0.145, and 0.245 m were fitted with a pressure tap and a thermocouple (K-type), while the tube located at z ) 0.775 m was fitted only with a thermocouple. The combustor was electrically heated by means of ceramic mat heaters. The reactor temperature was controlled by two PID controllers driven by the signal from the thermocouples inserted near the column wall by means of the tubes located at z ) 0.045 and 0.775 m. The fluidizing gas was preheated to 500 °C in an electrical heater. The flue gas exiting the combustor was directed to a high efficiency cyclone for fine particle collection. After the cyclone, the flue gas was sampled for gas analysis. A paramagnetic analyzer and four NDIR analyzers were used for the online measurement of O2, CO, CO2, SO2, and CH4 concentrations, respectively, in the exhaust gases. NOx concentration was also determined by passing the sampled gas first in a NOx/NO catalytic converter and then in a UV analyzer for the measurement of NO concentration. The fluidization column was equipped with an air-assisted solid metering/feeding system for continuous injection of the fuel at the bottom of the bed. The feeding system consists of a fuel hopper mounted over a screw feeder that further delivers the powder in a mixing chamber where a swirled air flow pneumatically conveys the fuel within the bed. Fluidizing and fuel feeding air flows were metered with two mass-flow controllers. The column was equipped with three high-precision piezoresistive pressure transducers, characterized by a high temporal resolution. These transducers were used to measure the gas pressure profile along the fluidized bed height. The acquisition data unit consisted of a personal computer equipped with a 16 bit A/D data acquisition board characterized by a high sample rate. The acquisition data unit was used for measuring gas concentration, temperature, and pressure signals at a sampling frequency of 100 Hz. Materials. Technical air was used both as primary fluidizing gas and to assist fuel feeding. Fresh quartz sand, sieved in two size ranges, 212-400 and 600-850 µm, was used as the inert bed material. The bed inventory was kept constant at 3.3 kg corresponding to a bed height at minimum fluidization conditions of 0.3 m. The fuel used was olive husk, either exhausted or virgin, whose properties are reported in Table 1. Olive husk is a solid residue of


Scala and Chirone Table 2. Summary of Operating Conditions and Main Results of the Combustion Experiments run

fuela

T (°C)

U (m/s)

e (-)

dS (µm)

η (%)

tD (min)

Wash (%)

5 3 4 2 9 1 7 6 8 10 11 12 13 14

exhausted exhausted exhausted exhausted exhausted exhausted exhausted exhausted exhausted virgin virgin virgin virgin virgin

850 850 850 850 850 850 900 900 900 850 850 850 850 850

0.48 0.48 0.48 0.47 0.44 0.47 0.49 0.46 0.45 0.59 0.61 0.92 0.38 0.61

1.17 1.17 1.42 1.44 1.67 2.10 1.22 1.38 1.65 1.47 1.77 1.76 1.78 1.76

212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 212-400 600-850

96.07 96.06 97.30 97.14 99.22 99.68 96.01 97.09 98.82 98.52 99.64 98.67 98.26 99.21

196 200 207 287 285 310 156 168 225 147 197 119 230 348

3.00 3.17 2.70 3.42 2.94 2.68 2.37 2.08 2.31 0.48 0.57 0.51 0.41 1.01

a

Figure 2. Fuel particle size distribution.

the olive oil production industry. Virgin husk is obtained after mechanical extraction of oil by olive pressing, while exhausted husk is obtained after solvent extraction of residual oil from virgin husk. Both kinds of husk are characterized by a high potassium content in the ash. The particle size of the fuel used in the experiments was in the range of 20-4000 µm (Figure 2). Procedures. Steady combustion tests were carried out at a fluidization velocity varying between 0.38 and 0.92 m/s. The bed temperature was fixed at either 850 °C or 900 °C. The combustor start up was accomplished by electrically heating the bed of fresh inert sand. When the bed temperature reached a value of 750 °C, fuel feeding started. The fuel feed was adjusted in the range of 0.33-0.82 kg/h to reach the desired excess air value (in the range 17-110%; i.e., e ) 1.17-2.1). During the run, temperature, pressure, and gas concentration data were continuously logged in the PC. Every 15-20 min, the elutriated material collected at the cyclone was measured and analyzed for carbon concentration to determine the unburned carbon flow rate at the exhaust. Mass balance closures on carbon and oxygen were always within a 3% error. The run ended when defluidization of the bed occurred, as indicated by a jump in the temperature and pressure profiles within the bed. The total time interval from the beginning of the fuel feeding until the defluidization onset was recorded. After the end of the run, the bed was cooled and discharged from the combustor and weighed and sieved for agglomerate separation. Selected agglomerate samples were observed under a scanning electron microscope (SEM) and subjected to energy-dispersive X-ray (EDX) elemental analysis. A few samples were embedded in epoxy resin and then cut and polished for SEM observation and EDX elemental mapping of agglomerate cross-sections.

Results and Discussion Olive Husk Steady Combustion Behavior. Table 2 reports a summary of the operating conditions and the main results of the experiments carried out in this work. The olive husk combustion tests were performed at different bed temperatures (Tbed), fluidization velocities (U), excess air factors (e), and inert bed-particle sizes (dS). In addition, two different kinds of olive husk (exhausted and virgin) were used alternatively in the experiments. In contrast to that in other biomass fuels, the air-assisted inbed olive husk feeding was fairly smooth, thanks to the relatively high density (∼1000 kg/m3) and sphericity (∼0.84) of the particles. Figure 3 shows the typical gas concentration profiles measured at the outlet of the reactor during a test (run 4). Time zero in Figure 3 was arbitrarily chosen within the steady state combustion period of the run. It can be observed that gas concentrations at the outlet of the reactor were reasonably

Exhausted or virgin olive husk.

Figure 3. Gas concentration profiles measured at the outlet of the reactor at steady state: run 4; fuel, exhausted husk; T ) 850 °C, U ) 0.48 m/s; e ) 1.42; dS ) 212-400 µm.

constant during the run, except at the end, after bed defluidization had occurred. In the figures, the vertical lines represent the time of occurrence of bed defluidization, as inferred from pressure and temperature profiles (this will be discussed later on). After bed defluidization, a slight change of the gas concentration profiles was observed, followed by a more abrupt change when the fuel feeding was stopped (approximately two minutes after defluidization had occurred). The slight change in the outlet gas concentration profiles upon defluidization was most probably caused by a change in bed fluid-dynamic conditions and, in turn, by a change of the local environment (e.g., temperature, oxygen concentration) experienced by the burning fuel particles. All of the experiments, both with exhausted and virgin husk, showed a similar qualitative behavior. Figure 4 reports the average CO, CH4, SO2 and NOX steadystate outlet concentrations for all the combustion runs, as a function of the excess air factor. As expected, CO and CH4 concentrations sharply decreased as the excess air increased. It



Figure 5. (A) Average steady-state carbon elutriation rate and (B) total combustion efficiency as a function of the excess air factor for all combustion experiments.

Figure 4. Average steady-state outlet gas concentrations as a function of the excess air factor for all combustion experiments.

must be underlined that the combustion of volatile species in the freeboard was not optimized in these experiments and no secondary air was used because bed agglomeration behavior was the focus of this work. The SO2 concentration at the outlet was relatively high because the sulfur content of the biomass is not negligible. No limestone was added to the bed for SO2 capture. The SO2 concentration decreased as the excess air increased as a consequence of gas dilution. The NOX outlet concentration apparently exhibited a minimum with the excess air. A possible explanation might rely on two conflicting effects as excess air increases: (1) NOX would tend to decrease because of a larger gas dilution and (2) it would tend to increase because a higher oxygen concentration promotes nitrogen oxidation. A secondary effect can also be attributed to the influence exerted by the excess air on the steady char loading in the bed, which is known to affect the nitrogen chemistry in fluidized beds.49-51 Figure 5 reports the average steady-state carbon elutriation rate (EC) at the combustor outlet and total combustion efficiency (η) as a function of the excess air factor for all the experiments. The total combustion efficiency takes into account both elutriated unburned carbon and CO and CH4 concentrations measured (49) Winter, F.; Liu, X.; Stingl, C.; Hofbauer, H.; Liu, D. The Effect of Fragmentation on NO/N2O Formation. In Fluidization X; Kwauk, M., Li, J., Yang, W., Eds.; United Engineering Foundation: New York, 2001; pp 661-668. (50) Liu, H.; Gibbs, B. M. Modelling of NO and N2O Emissions from Biomass-Fired Circulating Fluidized Bed Combustors. Fuel 2002, 81, 271280. (51) Kallio, S.; Konttinen, J.; Kilpinen, P. Effects of Char Inventory and Size Distribution on Nitric Oxide Formation in CFB Combustion of Coal. In Fluidization XI; Arena, U., Chirone, R., Miccio, M., Salatino, P., Eds.; United Engineering Foundation: New York, 2004; pp 827-834.

Figure 6. (A) Bed temperature, (B) temperature variance, and (C) relative temperature difference profiles at steady state: run 8; fuel, exhausted husk; T ) 900 °C; U ) 0.45 m/s; e ) 1.65; dS ) 212-400 µm.

at the outlet. As expected, the carbon elutriation rate decreased and the combustion efficiency increased as the excess air increased. Total combustion efficiencies higher than 96% were found for all the experiments, and they were higher than 98% when the excess air was above 1.5. Temperature and Pressure Profiles. Figure 6 A shows the typical temperature profiles measured at three different heights within the bed during a steady combustion test (run 8). Again, time zero in Figure 6 was arbitrarily chosen within the steady state combustion period of the run (and it is different from that chosen in Figure 3). The bed temperature was fairly constant during the run varying (10 °C from the set point. It can be observed that the higher thermocouple (z ) 0.245 m) measured a temperature slightly lower than the other two. This is probably a consequence of heat dispersion from the bed surface to the


Scala and Chirone

colder freeboard. It is interesting to note that the other two thermocouples measured approximately the same temperature during the first period of the run. However, at a certain point (t ≈ 80 min, for this run, indicated by a vertical dotted line) the two temperatures started to diverge, the central thermocouple measuring a higher temperature. This indicates that bed mixing was less effective in the second period of the run and that fuel combustion was mostly concentrated in the middle section of the bed (just above the fuel feeding tube). Suddenly, at t ≈ 174 min (for this run) the bed defluidized: the temperature measured by the lower thermocouple peaked down, while that measured by the other two tended to increase, before the fuel feed was stopped (this trend is not clearly evident in Figure 6 because the fuel feed was stopped immediately after defluidization in this run). It must be noted that, consistent with data reported in the literature,13,15 the increasing trend of the temperatures measured by the two upper thermocouples was observed in all experiments upon defluidization. This trend should be related to the decrease of the heat transfer coefficients when the bed is defluidized and to the segregated fuel combustion in the upper bed region. Figure 6 also shows the temperature variance (B) and the relative temperature difference (C) profiles during the same experiment. The first variable (σT2) is defined as the average value of the squared difference between the instantaneous measured temperature and the average temperature calculated within a definite time interval

σT2 ) (T - T h )2

(1)

This interval was fixed to 30 s to indicate the variance in a time scale of practical interest. In particular, in Figure 6 the temperature variance calculated from the temperature measured by the higher thermocouple is shown. This variance was found to be the most significant among those calculated from the three measured temperatures. The variance represents the average amplitude of the temperature fluctuations around the average value and is linked to the fluid-dynamic conditions around the measuring thermocouple. The plot in Figure 6 shows that for t < 80 min no particular trend could be observed, the variance varying between 2 and 10 K2. On the other hand, for t > 80 min, a decrease of the variance is evident; it approaches zero just before bed defluidization occurs. This trend is clearly connected to the decrease of solid mixing as bed agglomeration proceeds until final defluidization. It is interesting to note that this behavior was more evident for the higher thermocouple, which is positioned just below the bed surface. This result can be explained by considering that evidence reported in the literature suggest that the bed agglomeration/defluidization phenomena often start near the bed surface.13,15,52 The relative temperature difference (Figure 6, C) is defined as the difference between the instantaneous temperatures measured by the two lower thermocouples (z ) 0.45 and 0.145 m), divided by the temperature at z ) 0.45 m and multiplied by 100. This quantity has been introduced here to quantify the temperature divergence trend noted before, as an indicator of the effectiveness of bed mixing. Consistent with the temperature profiles described before, for t < 80 min, the relative temperature difference is close to zero, while for t > 80 min, it starts to increase reaching a maximum of approximately 1.5% at bed defluidization. Again, this trend is a consequence of the decrease (52) Bakker, R. R.; Jenkins, B. M.; Williams, R. B. Fluidized Bed Combustion of Leached Rice Straw. Energy Fuels 2003, 16, 356-365.

Figure 7. (A) Average gas pressure and (B) pressure variance profiles at steady state: run 8; fuel, exhausted husk; T ) 900 °C; U ) 0.45 m/s; e ) 1.65; dS ) 212-400 µm.

of solid mixing (and, in turn, of the heat transfer rate) within the bed as agglomeration proceeds. Figure 7 A shows the average gas pressure profiles measured at the three different bed heights during the same run. Obviously, the lower the measurement level the larger the measured pressure and the difference between whichever two pressure values represent the weight of the bed material (per unit bed section) located between the two measuring levels. All three pressures increased with time during the run. This is likely to be the result of the accumulation of ash in the bed because there is no drain flow. Also, in this case, the curves can be divided into two periods: at the same point as before (t ≈ 80 min) the slope of the curves changed. This slope change might indicate either a faster accumulation of ash in the bed in the second period or the onset of additional interaction forces between the bed particles as the surface became sticky. Upon defluidization (t ≈ 174 min), all of the pressures peaked down because of the onset of bed channeling. All combustion runs carried out with olive husk showed a similar qualitative behavior of pressures. Figure 7 also shows the pressure variance (B) profile during the same run. This variable (σP2) is defined as the average value of the squared difference between the instantaneous measured pressure and the average pressure calculated within a definite time interval

σP2 ) (P - P h )2

(2)

Again, this interval was fixed to 30 s. In separate blank test runs, we determined that this interval was large enough to have realistic and accurate values in the present conditions by comparing calculations made with time intervals up to 10 min. In Figure 7, the variance calculated from the pressure measured at the middle section of the bed (z ) 0.145 m) is shown. This variance was found to be the most regular among those calculated from the three measured pressures. Variance calculated from the pressure measured at the higher tap (z ) 0.245 m), however, gave similar results. As for the temperature variance defined before, the pressure variance represents the average amplitude of the pressure fluctuations around the average value in the chosen interval, and it is linked to the local fluid-dynamic conditions in the measured bed section. In particular, the pressure variance is affected by phenomena such as gas turbulence and bubble formation, passage, coalescence, and eruption.44 In Figure 7 it can clearly be seen that while in


Figure 8. (A) Defluidization time and (B) weight fraction of ash accumulated in the bed at defluidization as a function of the excess air factor for all combustion experiments.

the first period of the run the variance was approximately constant, in the second period (t > 80 min) it steadily decreased until bed defluidization. A 60% decrease of the variance from the steady average value was measured at the defluidization onset. This decrease in the pressure variance might be connected either to an increase of the average size of the bed particles as agglomerates start to form in the bed or to a change of bed fluid-dynamics caused by a decrease of the effectiveness of solid mixing. Characterization of Bed Defluidization Conditions. Figure 8A shows the measured defluidization time (tD) for all the combustion runs as a function of the excess air factor. Defluidization time is defined as the time interval between the start of the olive husk fuel feeding and the bed defluidization occurrence. An analysis of the data reported in Figure 8 clearly shows that faster agglomeration occurred with a higher operating temperature, with a lower excess air, and with a higher fluidization velocity (when the other variables were kept fixed). The first effect is consistent with literature data,32,33 which reported that agglomerate formation is enhanced at higher temperatures where the low-melting point eutectics at the inert particle surface are more easily reached. On the other hand, the other two effects are simply a consequence of the larger fuel feed rate adopted when a lower excess air value (at the same fluidization velocity) or higher fluidization velocity (at the same excess air) is chosen. A larger fuel feed rate implies a larger ash input rate in the bed and, in turn, a larger ash accumulation rate on the inert bed particles. It is interesting to note that in the experiment carried out with a larger sand particle size (run 14) the defluidization time almost doubled with respect to the corresponding experiment where the smaller sand particles were used (run 11) and all the other variables were kept the same. A possible explanation of this result might rely on the consideration that larger bed particles have more inertia and consequently are associated with more energetic collisions so that adhesion of the particles to form agglomerates should be more difficult. Finally, it is noted that in the combustion runs where virgin husk was used, a shorter time was measured for bed defluidization to occur compared to the corresponding runs with exhausted husk. The reason is clearly connected to the alkali content of the fuel ash (Table 1): it can be noted that virgin husk ash has a potassium content almost two times larger than that of the exhausted husk. Also the phosphorus content of the virgin husk ash is larger, and this compound might participate


in eutectic formation. These results indicate that the time of operation of the combustor before bed defluidization occurs is largely dependent on the kind of fuel that is fired and, in particular, on the ash composition of the fuel. Figure 8B shows the maximum weight fraction of ash accumulated in the bed (Wash) at the end of each combustion run as a function of the excess air value. This quantity is calculated as the total ash inlet with the fuel feed during the run divided by the inert bed weight and multiplied by 100. It must be underlined that this is a theoretical quantity as the elutriation loss of both sand and ash during the run is not taken into account in the calculation. Similar quantities referred to the single ash components could be calculated by simply multiplying Wash with the weight fraction of a particular component in the fuel ash (per ash analysis). An inspection of Figure 8 shows that, regardless of the excess air value and fluidization velocity, a well-defined ash content was necessary to defluidize the bed, dependent on the bed temperature and on the kind of husk used. In particular, the ash content was calculated to be ∼3% and ∼2.3% for the runs carried out with exhausted husk at 850 °C and 900 °C, respectively, and the ash content was ∼0.5% for the runs carried out with virgin husk at 850 °C. This result is in agreement with previously reported data32,33 that indicate an important effect of the bed temperature and fuel ash alkali content in enhancing the agglomeration tendency of the bed. Moreover, consistent with the defluidization time results, in the experiment carried out with larger sand particles, the amount of ash necessary to defluidize the bed was larger than that necessary in the corresponding experiment where smaller sand particles were used. All combustion runs carried out with olive husk showed a similar qualitative behavior for the gas concentration at the outlet and the temperature and pressure in the bed. However, these variables could not be used for agglomeration prevention purposes as the quantitative changes were different from run to run and because the change of these variables was significant only upon bed defluidization. In an effort to find a new variable which could be used to reliably predict bed defluidization (i.e., to act as an early warning variable for bed agglomeration problems), three quantities were tested in the present experiments. These quantities are the temperature variance, relative temperature difference, and pressure variance, as defined in the previous section. In particular, the percentage change of each of these quantities upon bed defluidization with respect to the steady-state value were used as indicators for the reliability of bed agglomeration prediction. Figure 9A shows the reduction of the temperature variance at defluidization (for the higher thermocouple) with respect to the steady-state value for all of the combustion experiments. It can be observed that a 97-98% reduction of the temperature variance upon defluidization was found for most of the experiments. However, the experiment carried out at U ) 0.92 m/s and especially the one carried out with a larger size sand bed showed a lower variance reduction. In addition, as clearly visible in Figure 6, the steady state variance signal was quite scattered, so it was difficult to choose an average representative steady state value. From an operational point of view, this quantity is not well suited for bed agglomeration prediction, because the presence of a large steady state variance scatter would easily produce erroneous agglomeration warnings. Figure 9B shows the relative temperature difference at defluidization as a function of the excess air factor for all of the combustion experiments. A 1-2% relative difference was found for most of the experiments. Again, the experiment carried


Scala and Chirone

Figure 9. (A) Temperature variance reduction, (B) relative temperature difference, and (C) pressure variance reduction at defluidization as a function of the excess air factor for all combustion experiments.

out with a larger size sand bed showed a result much different from the other ones. Also this quantity is considered to be unsuited for bed agglomeration prediction because the measured maximum relative temperature difference is quite low and it is not large enough to be confidently distinguished from values connected to random bed temperature oscillations. In addition, the relative temperature difference is likely to be strongly influenced by the location of both the fuel feeding and bed temperature measurements. Figure 9C shows the pressure variance reduction at defluidization with respect to the steady state value (at z ) 0.145 m) for all of the combustion runs, as a function of the excess air. A fairly constant value around 60-70% was measured for all runs, independent of the operating condition, olive husk type, and bed sand size. This result seems to indicate that the pressure variance reduction at the middle section of the bed might be a useful criterion to confidently predict the onset of bed defluidization. An inspection of Figure 7 shows that the progressive reduction of the pressure variance from the steady value to the one relative to defluidization conditions occurred on a time scale on the order of hours. This result was verified for all the combustion runs, and it indicates that the pressure variance reduction criterion might be applied as a real-time indicator of the formation of agglomerates in the bed giving reasonable time for the combustor operator to apply possible countermeasures to prevent bed defluidization. Clearly, a definitive conclusion on the usefulness of this technique requires further testing on larger scale fluidized bed combustors, which should be more representative of a full-scale operation. Analysis of Agglomerate Samples. At the end of each combustion test the bed was discharged for further characterization. The agglomerates discharged from the bed appeared to be quite weak and were easily broken into smaller ones. Approximately 2% of the bed material was found to have a particle size (>400 µm) larger than the initial size range of the fresh sand (212-400 µm). This number should be taken as a rough estimate as the bed discharge procedure is likely to have influenced the particle size distribution of the bed by partially

Figure 10. SEM micrographs of typical large agglomerate samples.

breaking the agglomerates. About 10% of this larger bed material had a size larger than 1.5 mm, and was made up of big agglomerates containing a considerable number of sand particles stuck together. Figure 10 shows SEM micrographs of two typical large agglomerate samples discharged from the bed after defluidization. Most of these agglomerates showed a hollow structure, indicating that a burning fuel particle was likely located inside and possibly initiated the agglomeration process. An inspection of Figure 10 indicates the presence of two different zones in the agglomerates: in the first one, the sand particles appear to be completely embedded in a fused layer of material (see zone A in the bottom micrograph), while in the second one, the sand particles are only attached to each other by a limited number of contact points, in the form of fused material necks. The first kind of zone was typically found in the innermost part of the agglomerates, indicating that higher temperatures were experienced inside. Figure 11 shows three SEM micrographs at different magnification of selected zones of the agglomerate samples. Many zones of the agglomerates (both the bridges between the sand particles and the surface of the sand particles itself) appear to be fused and resolidified. Within the fused material, however, nonmolten ash inclusions are clearly visible. These ash inclusions are likely to have been deposited first on the sticky fused surface and then progressively embedded in the melt. Results of EDX spot analyses of selected zones of the agglomerates (points A, B, and C in Figures 10 and 11) are reported in Table 3. An enrichment of potassium (and calcium) can be clearly observed in all three zones, especially in the fused



Figure 12. (top) SEM micrographs of fresh sand particles and (bottom) bed sand particles discharged from the combustor after defluidization.

Figure 11. SEM micrographs at three different magnifications of selected zones of the agglomerate samples. Table 3. EDX Spot Analysis of Selected Zones of the Agglomerates Shown in Figures 10, 11, and 12a zone

Na

Mg

Al

Si

A B C D E

4.44 1.29 2.07 1.18

1.23 1.27 1.52 1.95

2.14 2.64 2.00 2.84 8.59

4.44 49.94 77.19 95.61 45.71

a

P

S

K

Ca

2.35 25.61 48.52 6.25 5.51 1.45 16.58 13.85 3.89 2.26 6.58 3.12 1.55 4.89 21.48 10.80

Cr

Fe

1.77 0.68 0.23 0.84

3.26 6.80 1.14 4.53

Data are in wt % of the element (oxygen-free basis).

layer A. Figure 12 shows SEM micrographs of unused fresh sand particles (top) and used (but nonagglomerated) sand particles discharged from the combustor after defluidization

(bottom). The two samples turn out to be very similar, even if the fresh particles appear to be slightly more rough and angular, while the used ones more smooth and rounded. The results of EDX spot analyses of points D and E in Figure 12 are also reported in Table 3. A strong enrichment of potassium on the surface of the used bed particles is evident. It is interesting to note that the other ash elements (which were not originally present in the fresh sand) were found in all EDX analyses on used bed material in relative proportions similar to those of the original ash analysis (Table 1). This result suggests that ash material is likely to be consistently deposited on the sand particles during combustion. Selected large agglomerate samples were embedded in epoxy resin, cut, and polished for SEM observation and EDX elemental mapping of cross-sections. Figure 13 shows a SEM micrograph of the cross-section of a typical agglomerate. Again, a hollow structure is clearly evident with a large cavity surrounded by a continuous layer of fused material embedding a number of sand particles. A second smaller cavity also appears in this sample. Sections a and b in Figure 13 were further observed at a higher magnification under the SEM, and the micrographs are shown in Figures 14 and 15, respectively. These figures also show the EDX elemental mapping of selected elements for the two crosssection areas examined. A lighter area in the maps indicates a higher concentration of the element. It is evident in the figures that a number of sand particles (mainly composed of silica with some K, Al, and Mg impurities) are fully embedded in a melt containing silica, potassium, sodium, and calcium, with minor amounts of magnesium and aluminum (and Fe and P, not shown in Figures 14 and 15). It must be noted that some aluminum spots on the maps derive from impurities from the sample


Figure 13. SEM micrograph of the cross-section of a large agglomerate sample. The dark gray zones represent epoxy resin in which the sample is embedded.

polishing procedure and were not originally present in the agglomerate. Mechanisms of Fuel Ash-Bed Particle Interaction and Agglomerate Formation. Fuel ash-bed material interactions can be analyzed in the light of the scheme given in Figure 16, representing the fate of ash inside the dense phase of a fluidized

Scala and Chirone

bed. Upon injection in the hot fluidized bed, fuel particles devolatilize yielding either coarse nonelutriable char particles or elutriable char fines. The relative extent of these two “phases” is dictated by the attrition/fragmentation phenomena and carbon combustion. It is worth noting that during the combustion of biomass char extensive attrition and fines generation was observed, as discussed by Salatino et al.40 and Scala et al.41 Ash is obviously included in both of these phases. The important difference between the two phases is the combustion temperature with which they are associated. It has been suggested that char fines can burn at extremely high temperatures, much higher than typical temperatures relevant to coarse char.42 Both coarse and fine char particles can collide with inert bed sand, effectively transferring the ash if the stickiness of either the char or sand particle surface is sufficient for permanent adhesion. On the other hand, attrition of the sand particles may bring back part of the adhered ash into the free moving fines phase. Finally, free fines can be elutriated from the bed resulting in a loss of ash from the dense fluidized bed to the freeboard. So far we have described only “solid” flow rates between the phases inside the bed. An additional competitive transfer mechanism implies the vaporization of some of the ash constituents from the burning char particles and the subsequent condensation onto the sand surface. Part of this vaporized material may also exit the bed with the gas flow and condense later

Figure 14. SEM micrograph and EDX elemental maps of the agglomerate cross-section in Figure 13 section a.



Figure 15. SEM micrograph and EDX elemental maps of the agglomerate cross-section in Figure 13 section b.

Figure 16. Scheme representing the fate of ash inside a fluidized bed. Square boxes represent solid phases (CA ) ash in nonelutriable coarse char particles; FA ) ash in elutriable fine char particles; AA ) ash adhered onto inert sand particles). The solid and dashed lines represent flow rates of ash in the solid and vapor forms, respectively.

downstream of the combustion chamber. In principle, ash components also may vaporize from the sand particles to the gas phase, but this path is likely to be negligible because the sand temperature is typically lower than the temperature of the char.

To establish the relevant mechanism of ash-sand interaction and, in particular, the alkali species transfer to the sand surface, the relative importance of the pathways described before should be evaluated. Let us first analyze the vaporization-condensation mechanism. There is evidence in the literature that vaporization of alkali during biomass combustion is rather limited.33,37 On the other hand, the accumulation of ash alkali onto the inert bed was reported to be almost quantitative.33 A further evidence is that the composition of the outermost coating layer around the sand was found to be very similar to the fuel ash composition, with only limited enrichment of alkali compounds.4,9,27-31 Finally, bed temperature was not found to appreciably influence the alkali deposition rate33 or the ash-coating composition.9 With this in mind, it is interesting to note that in the present experiments the excess air factor did not influence the amount of ash accumulated in the bed at defluidization (Figure 8), while it certainly affected the burning temperature of the char particles. On the whole, on the basis of the above considerations, it is reasonable to conclude that the vaporization-condensation mechanism has only a limited importance, if any, on alkali transfer to the sand particles. Vaporization, in fact, should be very sensitive to temperature, and this sensitivity was not observed for alkali deposition. However, it must be stressed that, although limited, alkali vaporization is not negligible. Alkali


vaporization during biomass combustion has been reported to be a significant cause of fouling and corrosion problems on the superheater surfaces.53 It is likely that while the short gas residence time and the high temperature in the bed zone lead to little condensation of the gas-phase alkali species on the sand particles, in the low-temperature zones of the combustor, where longer residence times are available for the gas, alkali condensation is favored. As a consequence of the above reasoning, the most likely mechanism for the transfer of alkali to the bed particles relies on the collision of char particles with sand. Support for this mechanism is given by the results of Manzoori and Agarwal8-10 who found that inorganic elements in the ash tended to form a molten matrix on the char surface increasing its stickiness. Whether collisions of sand with fine char or with coarse char is the main ash transfer path still remains an open question. After they are deposited on the sand surface, the alkali species migrate to the quartz interface by solid-state diffusion through the ash layer, and then they interact with silica leading to the possible formation of an eutectic. The results of Lin and DamJohansen33 and the SEM analysis reported in this work (Figure 10) suggest that alkali transfer by collision is promoted by the high temperatures experienced near burning char. On the basis of this evidence, it is reasonable to assume that bed agglomerates start to form around burning char particles. The higher temperatures experienced near burning char enhance the formation of melt and, in turn, the particles stickiness. Now two different cases are possible. (1) The bed temperature is higher than the relevant eutectic temperature. In this case, after the time necessary to accumulate enough alkali in the bed to reach the eutectic composition at the sand surface (or, from another point of view, the time necessary to reach a critical coating thickness around the sand), the bed catastrophically defluidizes as a consequence of the permanent increase of stickiness of the whole sand bed. (2) The bed temperature is lower than the relevant eutectic temperature. In this case, even if alkali progressively accumulates on the sand, the formation of a surface melt on the bed solids is prevented. However, near the burning char, where the temperatures may be much higher than the bed temperature, (53) Westberg, H. M.; Bysrto¨m, M.; Leckner, B. Distribution of Potassium, Chlorine, and Sulfur between Solid and Vapor Phases during Combustion of Wood Chips and Coal. Energy Fuels 2003, 17, 18-28.

Scala and Chirone

agglomerates can still form by transitory local surface melting. If these agglomerates are strong enough to withstand collisions within the dense bed, their numbers would progressively increase, building up a defluidized layer in the bed which would eventually lead to total bed defluidization. Conclusions The steady fluidized bed combustion of a biomass residue (exhausted or virgin olive husk) common in the Mediterranean area was investigated in a bench-scale reactor. This fuel has a high propensity for bed agglomeration problems during combustion as a consequence of the high potassium content of the ash. Bed defluidization characteristic times were measured, together with the temperature/pressure profiles within the bed at different operating conditions. Results showed that during the experimental runs both the temperature and pressure profiles changed because of the decrease of mixing and the ash accumulation in the bed. A critical ash content in the bed was required for bed defluidization. Bed temperature, sand size, and the type of husk had all strong effects on this quantity, but this was not the case for the excess air. A diagnostic tool based on the measurement of the dynamic pressure signal inside the bed was tested for its ability to predict bed agglomeration. This technique was fairly accurate under the present operating conditions. Mechanisms of the fuel ash-bed particle interaction and agglomerate formation are interpreted on the basis of a SEM/ EDX analysis of agglomerate samples discharged from the bed after defluidization. The most likely ash transfer mechanism relies on collisions of sand with burning char particles. It is reasonable to assume that bed agglomerates start to form near burning char, where the higher temperatures enhance the formation of melt and, in turn, the stickiness of the particles. Acknowledgment. The authors are indebted to Mr. A. Cammarota and Mr. F. Tartaglione for their support in performing experimental tests. The support of Mrs. C. Zucchini and Mr. S. Russo in the SEM/EDX analysis is gratefully acknowledged. The authors are grateful to ENEL Produzione-PSI-Ricerca for providing olive husk samples. EF050236U

Characterization and Early Detection of Bed Agglomeration during the

Recommend Documents