Energiministeriets Forskningsudvalg for produktion og fordeling af el og varme

EFP-95 J.nr. 1323/95-0007

Characterization of Ashes From Biofuels

Flemming J. Frandsen Lone A. Hansen AUS 2 5 1938 Department of Chemical Engineering, Technical University of Denmark

Henning S. Sorensen

Geological Survey of Denmark and Greenland

Klaus Hjuler

dk-TEKNIK Energy & Environment

Edited by Klaus Hjuler and Henning S. Sorensen (February 1998)

ISBN 87-7782-000-2 EFP-95 J.nr. 1323/95-0007

Characterization of Ashes From Biofuels

Flemming J. Frandsen Lone A. Hansen

Department of Chemical Engineering, Technical University of Denmark

Henning S. Sorensen

Geological Survey of Denmark and Greenland

Klaus Hjuler

dk-TEKNIK Energy & Environment

Edited by Klaus Hjuler and Henning S. Sorensen (February 1998)

ISBN 87-7782-000-2 DISCLAIMER

Portions of this document may be illegible electronic image products. Images are produced from the best available original document. Preface

This report contains the results of a co-operative research and development project “Characterization of Ashes from Biofuels ” carried out in the period from Jan. 1995 to Jan. 1998 by the Department of Chemical Engineering at the Technical University of Denmark, the Geological Survey of Denmark and Greenland, and dk-TEKNIK Energy & Environment. The project management was at dk-TEKNIK. Fundings from the Danish Energy Research Programme, Elsam, and Elkraft are gratefully acknowledged. A steering comittee was established with participation of Bo Sander, Elsam, Ole Hede Larsen, Fynsvasrket I/S, and Peter Simonsen, Elkraft, to whom thanks are given for guidance and advises during the project.

Part of the results obtained in the present project was presented at the Engineering Foundation Conference in Kona, Hawaii, Nov. 2-7 (1997). Two papers (found in Appendix B and C) were presented orally and two (Appendix A and D) as poster presentations. The paper in Appendix C was awarded a prize for its quality and content. This and other parts of the work presented in this report was carried out as a Ph.D. study at the Department of Chemical Engineering at the Technical University of Denmark. The Ph.D. thesis entitled “Melting and Sintering of Ash Deposits ” by Lone Hansen is available at the department.

Another more practical result of the project is that the new method for ash fusion characterization based on High-Temperature Light Microscopy (described in Appendix B) has been taken into use by the fuel analysis laboratory at dk-TEKNIK Energy & Environment as an alternative to the standard ISO 540 test. At the time of writing this report, a large number of biomass ashes have been analysed on a commercial basis.

I Contents

Preface i Contents ii List of Appendices iii List of Figures iv List of Tables iv

IA. Summary and Conclusions 1 IB. Resume og Konklusioner 4 2. Introduction to Literature 6 2.1. Fuel and Ash Chemistry Aspects 6 2.2. Developments in Ash Fusion and Sintering Tests 6 3. New Methods of Biofuel Ash Characterization 8 3.1. Computer Controlled Scanning Electron Microscopy (CCSEM) 8 3.2. Simultaneous Thermal Analysis (STA) 9 3.3. High-Temperature Light Microscopy (HTLM) 10 3.4. Electrical Resistance Test 10 4. Fuel and Ash Samples Studied 11 5. Characterization of Samples by CCSEM, STA, and HTLM 13 5.1. Inorganics in Straw, Straw Char, and Grains 13 5.2. Straw Laboratory Ashes 14 5.3. Straw, Ashes and Deposits from CHP 14 5.4. Fuels, Ashes and Deposits from PF Straw-Coal Co-Combustion 22 6. Classification of Straw Ashes and Deposits (Triangular Diagram) 23 7. Prediction of Deposit Formation in Straw and Straw-Coal Fired Boilers 25 7.1. Case Study: Slagelse CHP (grate firing) 25 7.2. Global Equilibrium Calculations, GEA 28 7.3. Modelling Ash Deposition Fluxes 29 7.3.1. Residual Ash Transport: Inertial Impaction 29 7.3.2. Submicron AshTransport: Thermophoresis 31 7.3.3. Summary and Concluding Remarks 32 List of References 34 Appendix Section

ii List of Appendices

A Computer Controlled Scanning Electron Microscopy (CCSEM) Analysis of Straw Ash. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

B Ash Fusibility Detection Using Image Analysis. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

C Ash Fusion Quantification by Means of Thermal Analysis. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

D Ash Fusion and Deposit Formation at Straw Fired Boilers. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

E SEM-lmages and SEM-Analysis of Fuel, Ash, and Deposits from Straw and Straw-Coal Combustion.

F Various Notes on Apparatus, Equilibrium Calculations, Straw Leaching and Ash Mixing, and Samples from Coal/Straw Co-Firing.

G Fusion, Sintering, and CCSEM Analyses of Ashes from Coal/Straw Co-Firing.

H Estimation of Ash Deposition Fluxes in Straw (Co-)Fired Utility Boilers.

I Data Compilation (available as CHEC-report no. 9804 from the Technical University of Denmark, Dept. Chem. Eng, Building 229, 2800 Lyngby).

in List of Figures

BSE image of wheat straw 5.1 BSE image of the rim of a wheat grain 5.2 BSE image of a straw char surface 5.3 BSE image of laboratory ashes 5.4 Pie-diagrams of compositional data (BA and FA from straw firings) 5.5 Pie-diagrams of mineral category data for Haslev (straw fired CHP) 5.6 HTLM melting of Slagelse (straw fired CHP) bottom ashes 5.7 STA melting of Slagelse fly ashes 5.8 STA melting of MKS (coal/straw co-combustion) ashes and deposits 5.9

Triangular diagram with Haslev ashes and deposits 6.1 Triangular diagram with Slagelse BA and FA: Correlation with fusibility 6.2

HTLM melting of Haslev samples incl. leached straw 7.1 Flux of sticking particles on a tube at a 0, 10, and 20% straw share (MKS) 7.2 Flux of sticking particles in different sections (MKS) 7.3 Calculated thermophoretic and measured superheater dep. fluxes (CHP) 7.4

List of Tables

Overview of samples 4.1 Slagelse BA ranking 7.1 Slagelse FA ranking 7.2 Comparison of measured K and S evaporation and GEA 7.3 Comparison of measured HCI and GEA 7.4 Time necessary to build up a deposit layer 7.5

iv 1A. Summary and Conclusions

One motivation for initiating the present project was that the international standard method of estimating the deposit propensity of solid fuels, of which a number of variants exist (e.g. ISO, ASTM, AS, DIN), has shown to be unsuitable for biomass ashes. This is due to the fact that ashes from biofuels differ significantly from coal ashes with respect to mineral composition and fusion characteristics. Moreover, the standard method is based on a subjective evaluation of the deformation of a test specimen, which implies that the repeatability as well as the reproduciblity is poor - also for coal ashes. Quantitative methods are needed, which can give information about the extent of ash fusion as function of temperature and about the onset of sintering.

This goal was adressed by the development of two new methods for the detection of ash fusibility behavior based on Simultaneous Thermal Analysis (STA) and High Temperature Light Microscopy (HTLM), respectively. The methods were developed specifically for ashes from biofuels, but are suitable for coal ashes as well. They have been tested using simple salt mixtures, geological standards and samples from straw CHP and coal-straw PF combustion plants. All samples were run in a nitrogen atmosphere at a heating rate of 10°C/min. In comparison with the standard method, the new methods are objective and have superior repeatability and sensivity. Furthermore, the two methods enable the melting behavior to be characterized by a continuous measurement of melt fraction versus temperature. Due to this two-dimensional resolution of the results, the STA and HTLM methods provide more information than the standard method. Comparing results from the two methods have shown a general agreement, even though the detection principles are very different.

The STA method is based on the fact that melting is an endothermal reaction with no concurrent change of mass that can be detected by combining the well known TGA and DSC analysis. By integrating the endothermal peaks recorded during heating, the total melting enthalpy of the ash sample can be determined. From this a melting curve can be produced presenting the mass fraction of melt versus temperature when taking into account the melting enthalpy of the ash specie. The STA method requires expert knowledge to correctly interprete the results obtained, and therefore it is considered that this method primarily is suitable for research purposes.

The principle of the HTLM method is to quantify the ash melting behavior that can be observed visually in a microscope equipped with a heating stage. The ash is placed loosely on a specimen disc. The quantification is carried out by means of real-time, grey scale image acquisition and image analysis. By selecting a suitable method of converting from grey scale to binary images, the appearance of any melt formed during heating can be extracted from the processed image and the area of the remaining solid part of the ash is calculated at short time intervals (typically seconds). From this a curve presenting the fraction of melted ash versus temperature can be produced. The HTLM method is almost fully automated and requires no expert knowledge. The number of samples that can be run per day is relatively high due to the low heat capacity of the heating stage and the robustness of the apparatus.

Another important motivation for initiating this project was that more knowledge about fuel and ash mineralogy was essential in order to interprete experimental data available from full-scale combustion and deposition studies that recently have been carried out on straw- (Jensen et al. (1997)) and straw/coal-fired plants (Andersen et al. (1996), Hansen et al. (1996)). Fuels, laboratory prepared ashes, chars, bottom ashes, fly ashes, and deposit

1 samples from these studies were examined by: STA, HTLM, ISO 540 standard fusibility testing, bulk chemical analysis as well as by Computer Controlled Scanning Electron Microscopy (CCSEM) and Scanning Electron Microscopy equipped with Energy Dispersive X-ray analysis (SEM-EDX). The basic idea was to study the fuel inorganic matter and the corresponding residuals to provide understanding, input and calibration data for mathematical modelling of deposit formation and rate of build-up. The study of fuels and residuals also provided information on the chemistry of inorganic elements during combustion

SEM-EDX and CCSEM have during the recent years found widespread use for the characterization of coal mineral matter, but have not been used for biofuels so far. In order to do so, suitable mineral categories were defined in this work to characterize the inorganic matter of especially straw and straw based ashes. The CCSEM technique does not detect the dispersed inorganic elements in biofuels, but is efficient in characterizing the distribution of elements in particulate matter. This information is particularly interesting in combination with the fusion results. As far as possible the categories developed previously for coals were adapted (Laursen, 1997). Among other results, the SEM-EDX study of straws showed that potassium and chlorine are found mainly dispersed in the straw and only a minor part as submicron particulate KCI. In general, straws have relatively high contents of silica found as hydrated Si02, which enables the plant to stand erect on the field. The main content of straw bottom ashes from grate firing is mixed calcium and potassium silicates, whereas the fly ash mainly contains KCI. By CCSEM it was possible to characterize all types of straw related ashes with respect to their content of dominating minerals which determines ash fusion behaviour. This information was adequately presented in a triangular diagram and related to ash fusion temperatures. CCSEM studies of samples from co-firing of coal and straw revealed that straw potassium to a large extent reacts with aluminosilicates from the coal. A study of 550°C laboratory ashes of coal and straw blends showed that straw potassium preferentially reacts with coal sulphur instead of straw chlorine. Finally, the SEM- EDX study of the large number of samples in this work has yielded a reference selection of representative images that will be useful for future comparison.

The study of bottom ash and fly ash as well as deposit samples from straw test firings at the Haslev and Slagelse Combined Heat and Power plants (Jensen et al., 1997) resulted in a better understanding of mineral behaviour during straw grate firing. In these tests a number of straws were fired which had been carefully selected for having different qualities with respect to sort and potassium and chlorine contents. By studying bottom ashes from Slagelse it was found that the melting behaviour correlated with the deposition rate on a probe situated at the outlet part of the combustion zone. That is, lower melting bottom ashes resulted in higher rates of deposit formation. This correlation had not been observed previously with the standard ash fusion test. Moreover, it was found that the fraction of potassium evaporated from the combustion zone was dependent on the straw fuel content of silica, e.g. the more silica the less potassium evaporated. This behaviour could also be demonstrated by chemical equilibrium calculations in the present work.

A simple superheater deposition modelling approach has been tested, where the rate of deposit formation is assumed to depend only on condensation and impaction. The sticking probability of the impacting particles was assumed to increase linearly with the melt fraction at the gas temperature prevailing at the deposition probe position. The melt fraction was obtained from the STA and HTLM analysis of fly ashes sampled from Haslev and Slagelse CHP. It was found that the experimental rates were grossly overestimated, but a linear trend was found except for two experiments. A more detailed modelling approach has also been tested, where thermophoresis is taken into account. This approach proved to be successful and work will be continued in this area.

The modelling work resumed briefly above was carried out to understand which factors that mainly influence deposition rates and how deposition formation possibly can be predicted by combining laboratory fuel testing and models. This raises the question about the relevance of the laboratory prepared bulk ash used today for fusibility testing. In fact, it can always be argued that a laboratory prepared ash is not relevant for a specific plant and consequently it is unsuitable for estimating the deposition propensity of a particular solid fuel. However, it should be noticed that the new ash fusion tests are excellent as a relative test, i.e. the results should be related to previous experiences at a particular plant. Work will be continued to find a combination of laboratory testing and modelling that will enable plant operators and design engineers to make better predictions of the deposition propensities of solid fuels and mixtures thereof.

3 1B. Resume og Konklusioner

En vigtig motiverende faktor for igangssetningen af nasrvserende projekt var, at den Internationale standardmetode, hvormed et braendsels tendens til belaegningsdannelse vurderes (et stort antal nationale varianter findes, f.eks. ISO 540), bar vist sig at vaere uegnet til asker fra biobraendsler. Dette skyldes at den uorganiske del af biobraendsler adskiller sig vesentligt fra tilsvarende hos kul, bade med hensyn til kemisk sammen- sastning og med hensyn til elementemes associering. Desuden er standardmetoden baseret pa en subjektiv vurdering af formsendringer pa et askelegeme, og det betyder at reproducerbarheden save! som repeterbarheden er darlig. Derfor er der behov for en bedre og ikke mindst kvantitativ metode til at undersoge askers smelteforhold.

I projektet er der anvendt to nye metoder til detektion af askers smelteforhold: Simultan Termisk Analyse (STA) og Hoj-Temperatur Lysmikroskopi (HTLM). Metodeme er udviklet specielt til analyse af asker fra biobraendsler, men er ligesa velegnede til kulasker. De er testet ved brug af simple salt-blandinger, geologiske standarder samt prover fra ristefyring af halm og fra PF-samfyring af kul og halm. Alle prover blev undersogt i en nitrogen- atmosfaere ved en opvarmningshastighed pa 10°C/min. De nye metoder er objektive og har bedre reproducerbarhed og folsomhed end standardmetoden. Desuden opnas mere information om smelteforlobet, da denne karakteriseres kontinuerligt ved smelteandelen som funktion af temperaturen. Ved sammenligning af STA og HTLM metodeme findes generelt god overensstemmelse pa trods af at detektionsprincipperne er meget forskellige.

STA-metoden er baseret pa det velkendte forhold, at smeltning er en endoterm reaktion uden massesendringer som kan detekteres ved en kombination af termogravimetri og differential termisk analyse. Ved integration af de endoterme toppe uden massesendringer, som registreres under opvarmningen, kan den totale enthalpi til smeltning af en given prove beregnes. Nerved kan fremstilles en kurve, der viser smelteenthalpi-andelen som funktion af temperaturen. STA-metoden krsever ekspertviden for at kunne tolke de registrerede ra-data korrekt. Derfor vurderes det, at denne metode primaert er velegnet til forskningsbrug.

HTLM-metoden er principielt en made at kvantificere det smelteforlob, der kan iagttages visuelt i et lysmikroskop udstyret med et varmebord. Askeproven anbringes lost pa et objektglas. Kvantifikationen foretages ved optagelse af graskala-billeder og billed- behandling i sand tid pa PC. Ved at vselge en egnet made at konvertere billederne fra gratone til binssr form kan den dannede smelte ‘fjemes ’ fra det oprindelige billede, og arealet af den tilbagevserende faste del af asken kan bestemmes. Resultatet er en kurve, der viser areal-andelen af smeltet aske som funktion af temperaturen. HTLM-metoden er nsesten fuldt automatiseret og krsever ingen ekspertviden. Antallet af prover der kan kores per dag er relativt hojt pa grund af varmebordets lave varmekapacitet og udstyrets robusthed.

En anden vsesentlig arsag til projektets igangssetning var, at mere viden om braendsels- og askemineralogi var nodvendig for at kunne forsta forbrsendings- og belsegningsdata fra de seneste fuldskala forsog med halmfyring pa rist (EFP-93: “Karakterisering af biomasse til fyringsformal ”) og fra PF samfyringsforsog (Midtkraft). Samhorende prover af braendsler, laboratorieasker samt koks, asker og belsagninger fra disse forsog blev undersogt i nservaerende projekt med metodeme: STA, HTLM, standard ISO 540 askesmelteforlobs- bestemmelse, bulk kemisk sammensaetning, Computer Controlled Scanning Electron Microscopy (CCSEM) og Scanning Electron Microscopy med Energy Dispersive X-ray analysis (SEM-EDX). SEM-EDX og CCSEM har i de senere ar fundet udbredelse til karakterisering af mineraler i kul, men er hidtil ikke blevet anvendt til karakterisering af biobraendsler. For at gore dette blev der defineret mineralkategorier, som var egnede til specielt at karakterisere halm og halmbaserede asker. Sa vidt muligt blev tidligere for kul udviklede kategorier inddraget uden modifikationer. SEM-EDX studierne af halm viste blandt andet, at kalium og chlor findes disperst I halmen og ikke som partikulsert kaliumchlorid. CCSEM kan ikke detektere det disperse materiale, men er en effektiv metode til karakterisering af elementfordelingen i uorganiske partikler. Generelt har halm et relativt hojt indhold af silicium i form af hydreret Si02, som bevirker at planten kan sta oprejst. Bundasker fra halmfyring pa rist bestar hovedsagelig af blandede calcium- og kaliumsilikater, mens flyveaskeme hovedsageligt er KCI. CCSEM gor det muligt at klassificere halmrelaterede asker ved deres indhold af dominerende mineraler, som er bestemmende for askernes smelteforhold, og denne information kan prsesenteres i et trekantsdiagram. CCSEM studier af prover fra samfyring af halm og kul viste, at kalium fra halm i dette tilfeelde ‘foretrsekker* at reagere med svovl fra kul fremfor chlor fra halm. Den samme effekt sas ved laboratorieforaskning af kul og halm ved 550°C. Endelig har SEM-EDX studierne af det store provemateriale i projektet resulteret i udvselgelsen af et repraesentativt billedmateriale, der kan finde anvendelse som reference ved fremtidige undersogelser.

Studiet af prover fra fyring med forskellige halmarter og -kvaliteter pa Haslev og Slagelse Kraftvarmevaerker har givet en bedre forstaelse for uorganiske elementers opforsel ved ristefyring. STA og HTLM smelteundersogelser af bundasker fra Slagelse KW viste en sammenhaeng mellem askernes smelteforhold og deres tendens til belaegningsdannelse pa en probe, der var anbragt ved udgangen fra fyrrummet. Jo lavere smeltende aske, desto hojere belsegningsdannelseshastighed. Denne sammenhaeng kunne ikke pavises med ISO 540 standardmetoden. Desuden var der en sammenhaeng mellem bundaskernes smelteforhold og deres totale indhold af kalium samt med de laboratoriefremstillede askers indhold af kalium. Dette indikerer, at det er muligt pa basis af en laboratorieanalyse at estimere tilbojeligheden til dannelse af fyrrumsbelaegning. Endelig har det vist sig, at maengden af fordampet kalium fra fyrrummet oges ved et lavere indhold af silicium i halmen. Dette forhold kunne ogsa pavises ved udforelse af kemiske ligevasgtsberegninger.

En simpel made at modellere belaegningsdannelse pa overhedere er blevet testet, hvor det antages at impaktion er den eneste dannelsesmekanisme. Askepartiklernes tilbojelighed til at blive haengende pa en overflade antages at stige lineaert med smelteandelen ved den aktuelle gastemperatur. Smelteandelene blev aflasst fra STA og HTLM kurveme for bund- og flyveasker udtaget fra Haslev og Slagelse KW. Det viste sig, at modellen over- estimerede de malte belaegningsfluxe, men en lineaer sammenhaeng kunne pavises. I projektet er der ogsa arbejdet med en mere detaljeret model, hvor der tages hensyn til betydningen af thermophorese og denne model giver gode resultater. Modelarbejdet blev udfort for at afklare hvilke faktorer, der er mest betydende ved belaegningsdannelse, og hvordan den kan forudsiges ved kombination af braendselsanalyser og modelberegninger. Dette har sat sporgsmalstegn ved relevansen af den laboratoriefremstillede aske, der idag anvendes ved smelteforlobsanalysen, da askedannelsesprocessen i et givent fyringsanlaeg naturligvis ikke er sammenlignelig med laboratorieforaskningen. Pa den anden side er de udviklede STA og HTLM metoder udmaerkede som relative analyser, d.v.s. at en braendselskvalitet, der afviger fra tidligere anvendte kvaliteter, hurtigt vil kunne afslores med disse metoder. Der skal arbejdes videre pa at finde en kombination af laboratorie-analyser og beregninger, der vil gore det muligt for driftsledere og anlaegsdesignere at forudsige problemer med belaegningsdannelse ved fyring med faste braendsler og blandinger heraf.

— 5 2. Introduction to Literature

In the recent years the use of renewable fuels (biofuels) has gained increasing importance as a substitute for or by being co-fired with fossil fuels. This development is particularly forced by the international concern about carbon dioxide emissions. However, the use of biofuels for power production is so far restricted by the fact that biofuels generally have a high content of potentially deposit forming and corrosive elements. Another important aspect is that the characteristics of ashes from biofuel- and co-firing differ significantly from that of coal ashes, which affect possible uses of the ashes. In the following a brief introduction is given to recent literature on ash and deposit formation as well as on ash fusion and sintering determinations.

2.1. Fuel and Ash Chemistry Aspects

The elements contained in fossil and biofuels can be grouped in three concentration levels (Benson et al. (1993)): 1) the major elements, C, O, H, S, and N forming the organic matrix of the fuel, 2) the ash forming elements, Al, Ca, Fe, K, Mg, Na, and Si, present in the concentration range of about 1000 ppmw to a few wt% on a dry fuel basis, and 3) trace elements, e.g. As, B, Cd, Cr, Hg, Pb, Se, and Zn, typically present in concentrations below 1000 ppmw on a dry fuel basis.

The ash-forming elements occur in fuels as internal or external mineral grains, simple salts such as Na 2S04 and KCI or associated with the organic matrix of the fuel. Depending on the gas/particle temperature and redox conditions during fuel particle heat-up, devolatilization and char burnout, the simple salts may vaporize, while the mineral grains will undergo phase transformations and approach each other to form fly ash. In the furnace, some of the fly ash will be removed as bottom ash while the rest will be entrained with the hot flue gases. The vaporized metal species may undergo several reactions: nucleation, coagulation, heterogeneous condensation and/or interactions with mineral inclusions in the char or fly ash particles, depending on the total specific surface area of the fly ash particles, the cooling rate of the flue gas, the redox conditions, and mixing in the gas phase. Local supersaturation with respect to certain chemical species such as Na 2S04(g), KCI(g), and K2S04(g) may lead to the formation of submicron ash by homogeneous nucleation (Christensen (1995)). Vapors and fly ash particles may be deposited on heat transfer surfaces in the boiler through a number of mechanisms (Baxter et al. (1992), Rosner et al. (1992), and Baxter (1993)): inertial impaction, thermophoresis, condensation, chemisorption, and eddy turbulence. The stickiness and thereby rate of deposit formation is a strong function of the amount of molten phases formed in the system. Ash deposits may cause several operational problems, e.g. changes in the heat uptake of the boiler (Wall et al. (1994)), corrosion of heat transfer metal surfaces (Harb and Smith (1990), Jacobson et al. (1990), Ahila and Iyer (1992), and Michelsen et al. (1996)), and/or in extreme cases plugging of the convective pass of the boiler. Thus the ash-forming elements in fuels constitute a potential operational and cost problem. Traditionally, boiler designers and operators, and fuel manufacturers have tried to predict ash deposition propensities by use of standard laboratory methods for ash fusion and sintering.

2.2. Developments in Ash Fusion and Sintering Determinations

The current standard fusion test is subjective and has shown to be unsuitable for biofuel ashes (Westborg and Nielsen, 1994). In the standard test a ash test body shaped as a cone or cube is prepared and heated in an oven. The deformation of the body during heating is observed visually and three characteristic temperatures are reported: the initial deformation temperature (IDT), the hemispherical temperature (HT), and the flow temperature (FT). It is not uncommonly observed that the biofuel ash test body ‘blows up’ like a baloon or melt flows out of it without overall changes in shape, leaving a skeleton composed of e.g. silicon and calcium (Eriksen, 1993). This complicates the interpretation and reporting of the test, especially for automatic equipments, where the characteristic temperatures are determined only from the height and width of the test specimen. Another main criticism of the test has been for unreliability in predicting the behavior of ashes in real combustion processes. One of the possible explanations for this is that the initial deformation temperature is not the temperature at which the melting actually begins, which may take place at 300°C lower or more (Hansen et al., 1997).

An alternative approach for the determination of (coal) ash fusion temperatures which have been adressed in a number of studies is to calculate characteristic temperatures from bulk ash analysis results by weighing the effect of various compositional variables (e.g. Slegeir et al. (1988), Lloyd et al. (1993)). Lloyd et al. (1993) believe that for most blends of North American coals, ash fusion temperatures increase with increasing amounts of Al203, CaO, K20, and Na 20, and decrease with increasing amounts of MgO, S03, and Si02. However, the regression results cannot be used in general for all types of coals because the ash elements are not present as simple oxides but are associated in complex ways.

An electrical ash resistance method for the detection of ash fusibility has been studied by Gumming et al. (1985), Conn and Austin (1984), Gibson and Livingston (1991), Raask (1982), Wain et al. (1992), among others. The basis of this method is that the resistance of the ash drops at a certain temperature, TR, due to the formation of a liquid phase. Gumming et al. (1985) have shown an apparent relationship between the ash fusibility, including TR, and" the build-up of sinter strength. In general, it has been shown that the presence of Na and K reduce sintering temperatures and promote sinter strength. The presence of Ca, on the other hand, has an inhibiting effect on the development of sinter strength. Gibson and Livingston (1991) studied the sintering and fusion behaviour of coal ashes, both lab. ash and fly ash, over the temperature range 900-1 300°C. The electrical resistivity and the linear shrinkage of a range of ashes were measured as function of temperature under both oxidizing and reducing atmospheres in a custom-built apparatus. Characteristic curves were related to the chemical composition of the ash and the results of the standard ash fusion test. In general, reasonably good agreement was obtained between the characteristic ash sintering temperatures, the ash chemical parameters, and the standard test. Livingston (1994) reported that the recent activities at the Babcock Technology Centre, Scotland, have involved the extension of a data base of coal ash measurements. At that time, Babcock had not yet attempted to use the electrical resistance method for biofuel ashes.

Vassilev et al. (1995) discuss the influence of mineral and chemical composition of coal ashes on their fusibility. The ashes were analyzed using ICR (Inductively Coupled Plasma) for chemical composition and XRD (X-ray Powder Diffractometry) for major (> 5 vol-%) and minor (1-5 vol-%) minerals. Standard ash fusion tests (cones, air, 5°C/min) were carried out as well as DTA and TGA studies (air, 10°C/min). In addition, various samples were observed in reflected light (air, 107min). A negative correlation was found between the average sum of some fluxing minerals (feldspars, Ca-silicates, hematite, anhydrite, barite) and the average hemispherical ash-fusion temperature for 815°C laboratory ashes. Correspondingly, a positive correlation was found between main refractory minerals (quartz, kaolinite, mullite, Ti-oxides) and the hemispherical ash fusion temperature. It was suggested that such relations are used together with mineralogical data for predicting ash-fusion characteristics.

7 — 3. New Methods of Biomass Ash Characterization

This chapter contains brief presentations of the experimental methods and apparatus which have been developed in this work. A more detailed description and verification of the methods is found in Appendix A (CCSEM), B (HTLM), and C (STA). In Appendix D some of the results obtained using the HTLM and the STA method, respectively, are compared.

3.1. Computer Controlled Scanning Electron Microscopy (CCSEM)

A Scanning Electron Microscope equipped with an Energy-Dispersive X-ray analysis system (SEM-EDX) is a strong tool for characterization of a wide variety of materials. It can be used to acquire X-ray spectra in points or areas that are manually selected on the basis of backscatter - or secondary electron images for qualitative or quantitative compositional analysis. However, advanced automation of stage and electron beam movement combined with image analysis facilities allows Computer Controlled Scanning Electron Microscopy (CCSEM) to be performed.

The CCSEM analysis technique is designed to characterize the shape and chemical composition of individual particles. The use of CCSEM for characterising coal and coal ashes is well documented, but this technique has not yet been used for samples of biofuels and their ashes. The CCSEM data in this work were obtained on a Philips XL40 Scanning Electron Microscope equipped with a Voyager 2.7 EDX analysis system with Feature Sizing and Chemical Typing.

The samples were embedded in epoxy resin and subsequently cross sectioned to give a section through any gravitative sedimentation that may have occurred before hardening of the epoxy. The cross sectioned block was epoxy embedded and ground on P1000 silicon- carbide grinding paper (approximate grain size 18pm) and subsequently polished on cloth- discs saturated with first 1 pm and second 1/4 pm diamond spray. The polished sample were coated with carbon in a Polaron TB500 Carbon Coater, to avoid charging of the sample during analysis.

The Philips XL40 Scanning Electron Microscope were operated at 15kV with a beam diameter of approximately 0.1 pm. Backscattered electron (BSE) images were obtained with a solid state diode backscatter detector to give information of the average atomic number of particles in the sample, i.e. inorganic (mineral) material appears significantly brighter than the organic matrix and the embedding resin.

A binary image in which inorganic particles are singled out as white on a black background is produced on basis of the BSE image by choosing an appropiate threshold grey-level. The binary image is used for the CCSEM analysis to control the electron beam to perform a raster scan on each individual particle (or adjoined particles) larger than 1 pm in diameter. Data of particle size and form is collected together with an X-ray spectrum during the raster scan. In this way information about typically 1500 to 4000 particles were obtained for each sample depending on the area proportion of the inorganic particles.

The CCSEM compositional data are based on the obtained X-ray spectra, i.e. each particle is assigned to one of a series of compositional categories based on ratios between characteristic X-ray peaks. These categories are conventionally called “mineral ” categories even though the particles may consist of amorphous phases or salts. The data are semi- quantitative in the respect that the X-ray counts are not corrected for ZAP matrix effects (atomic number, absorbtion and fluorescence). This is, however, not critical for the CCSEM analysis because the X-ray spectra are used to quantify the proportions of characteristic compositional categories and not to give strict quantitative compositions. Therefore the CCSEM data are not directly comparable with traditional bulk chemical analyses, but results for different samples are internally comparable. The data reduction into mineral categories is based on detailed observations and semi-quantitative analysis of inorganic constituents in biomass (mainly straw) and straw ashes.

3.2. Simultaneous Thermal Analysis (STA)

The simultaneous thermal analysis (STA) technique is generally used to characterize the change in material properties as a function of temperature. The results in this work were obtained using the NETZSCH STA409.

Using STA, the sample (2-30 mg) and an inert reference material are placed in a furnace and subjected to a desired heating programme, typically heating from 20 to 1400°C with 10°C/min. The analysis implies the continuous measurement of sample weight (Thermo- gravimetry, TG) and sample temperature (Differential Thermal Analysis, DTA, or Differential Scanning Calorimetry, DSC) as the sample is heated. The weight measurement reveals any mass changes occuring in the sample and by comparing the sample temperature to the temperature of the inert reference material, any heat producing or heat consuming processes taking place in the sample is detected. The resultant DSC signal expresses the measured temperature difference between sample and reference material, meaning that the larger the peak in the DSC curve, the larger is the measured temperature difference between the two materials. The heat evolved or consumed in the detected process is quantified by calculation of the area under the peak in the DSC curve and multiplication with an apparatus and temperature dependent factor.

Using STA, melting is detected as an endothermic process involving no change in mass. For a combustion ash the melting will result in several overlapping DSC peaks corresponding to the melting of various chemical species in the ash, melting at different temperatures and involving different melting enthalpies. Obtaining a melting curve (that is the fraction of melt as a function of temperature) based on this kind of data can be done in two ways:

1) The area below the total melting curve from the first point where melting is detected and to the temperature where melting is completed, Amitota |, reflects the total energy used for melting of that ash. Taking the area under the DSC curve from any temperature T, to T2, and dividing this area, A^, by the total area under the melting curve yields the fraction of energy used for melting of ash in the specified temperature interval (compared to the total energy used for melting of the ash). And this energy fraction is a simple estimate of the mass fraction melted in the specified temperature interval. Using this estimate it is assumed that the melting enthalpy of all chemical species in the ash is equal, which is not true in reality. So, for a better estimation the next method can be used.

2) Any given peak below the DSC curve corresponds to an absolute quantity of energy used for melting. The position of the peak (the peak temperature) gives an indication of the identity of the substances that melts, which means that a reasonably good estimation of the relevant melting enthalpy can be given. Based on this information, the mass of material which is melted in the given temperature interval can be calculated, and by relating this to the mass of ash analysed, the mass fraction of ash melted in the given temperature interval is obtained. 3.3. High-Temperature Light Microscopy (HTLM)

This method is a development of the hot stage light microscopy known from e.g. metal ­ lurgical studies. Instead of making an ash body as in the standard method, the basic idea in HTLM is to place the ash sample loosely and randomly on a specimen disc. The process of shrinkage and melting during heating is observed using a stereo microscope with a digital camera, and the process is quantified by means of real-time image analysis on a personal computer. The result is a melting curve showing the melt fraction versus temperature. The apparatus and hardware is commercially available, whereas the software was programmed specifically for the image analysis and data acquisition, and for the operation of the heating stage. The HTLM method is suitable for routine analysis as no expert knowledge is required, it is almost completely automated, and the sample throughput per day can be high due to a low thermal mass of the heating stage.

The quantity of sample needed for the analysis is small, typically 1 pg (or about 1-2 mm3). In special cases information about the local melting behavior of eg. a deposit is needed. In this case a small piece, say 1 mm3, may be broken off and analysed directly. The Leitz heating stage used in this work is gas tight when a top glass seal is mounted and can be operated in a controlled atmosphere (normally nitrogen). The microscope is focused on a random part of the ash, the magnification is set at typically 100-120 X, and the light conditions are optimized. If desired, the melting behavior can be stored using a video tape recorder.

The ‘heart ’ of the method is the image analysis metod. The image analysis is based on grey level images and binary images. In the binary images, the ash sample appears black and the background white. The resolution is 384x288 pixels, which gives a manageable file size of 1=M kB. At a maximum magnification of 275 X, the corresponding specimen area is 0.15 mm2 and the resolution is 1.4 pm2. At the typical magnification of 100 X, the corresponding figures are 1.1 mm2 and 11 pm2 respectively. The way the grey level images are converted to binary images affects the calculated melt fraction (which is actually an area fraction), and a number of metods for this conversion can be applied depending on the application.

3.4. Electrical Resistance Test

Since no commercial apparatus was available, it was attempted to design a simple apparatus for routine measurements of the electrical resistance vs. temperature only (not for linear shrinkage measurement) which could be operated with a controlled atmosphere at atmospheric pressure. The equipment consists of a quartz glass tube with a support for an alumina crucible containing the ash sample between two wire net electrodes. The reactor was placed vertically in a tube oven. The design is outlined in Appendix F.

The electrical resistance method was preliminary tested using straw fly ash. The tests revealed some operational problems that limited the suitability of this method for routine analysis. The apparatus was laborious to use and had to be handled carefully. Relatively large quantities of sample were needed in order to make sure that the electrodes were separated. Low viscosity melt flowed down into and through the lower electrode Pt/Rh net, which was eventually plugged and had to be cleaned or disposed. The heat capacity of the system is relatively high, which reduces the number of runs per day significantly. Further­ more, an obvious determination of the resistivity breakpoint is not obtained. This means that the intersection of two regression lines has to found, and this is not obvious either, as it is not exactly clear which part of the curves that should be taken into account. Results seem to be comparable with the IDT of the standard method, but by using HTLM and STA on the same samples, changes were detected at temperatures typically about 100°C lower. As the STA and HTLM methods show much more promising, the electrical resistance method is not discussed any further in this work.

4. Fuel and Ash Samples Studied

Corresponding fuel and ash samples were selected for analysis in order to study mineral transformations and the fusion and sintering behaviour of laboratory ashes as well as ashes and deposits from test firings at full scale plants. The laboratory ashes were prepared in various ways to investigate the influence of the method of preparation. Laboratory ashes of fuel samples from the test firings were prepared for comparison with the inorganic and fusion characteristics of the full scale ashes.

The full scale samples (bottom ash, fly ash, deposits) originated from test firings on two grate fired CHP-plants (Haslev 5 MWe with cigar burner and Slagelse 8 MWe ) and a PF- boiler co-combusting coal and straw (Studstrup 150 MWe ). The CHP-plants and the measurement campaign have been presented in Jensen et al. (1997). Measurements at the Studstrup boiler has been reported in Andersen et al. (1996) and Hansen et al. (1996).

The major part of the fuel and ash samples studied in this work are listed in table 4.1.

11 Experiment Fuel Sample type Ref. name

Lab. ashing Wheat straw 550°C LA LAS1 815°C LA LAS2 Wash + 550°C LA LAS3W Part, wash + 550°C LA LAS4W Mix. LAS1 and LASS LAS5W Wheat grains 550°C LA LASS 815°C LA LAST Haslev 1 Wheat straw (Kontra) 550°C LA HAS1LA Wash + 550°C LA HAS1 LAW BA HAS1BA FA HAS1FA Superheater DEP HAS1SD-1 Outer furnace exit DEP HAS1FD-1 Inner furnace exit DEP HAS1FD-3 Slagelse 1 Wheat straw (Hevera) BA SLA1BA FA SLA1FA Slagelse 3 Barley straw (Canut) BA SLA3BA FA SLA3FA Outer superheater DEP SLA3SD-1 Inner superheater DEP SLA3SD-3 Outer furnace exit DEP SLA3FD-1 Inner furnace exit DEP SLA3FD-3 Slagelse 6 Wheat straw (Ritmo) BA SLA6BA FA SLA6FA Slagelse 7 Wheat straw (Ritmo) BA SLA7BA FA SLA7FA Slagelse 8 Rape (Bristol) BA SLA8BA FA SLA8FA Studstrup 3 20% wheat straw/coal 550°C LA (straw) MKS3LAS Wash + 550°C LA (straw) MKS3LASW 815°C LA (coal) MKS3LAC BA MKS3BA FA MKS3FA Deposit, pos. 1, 540°C probe MKS31-1 Deposit, pos. 2, 540°C probe MKS32-1 Deposit, pos. 3, 540°C probe MKS33-1 Studstrup 4 10% wheat straw/coal 550°C LA (straw) MKS4LAS Wash + 550°C LA (straw) MKS4LASW 815°C LA (coal) MKS4LAC BA MKS4BA FA MKS4FA Deposit, pos. 1, 540°C probe MKS41-1 Deposit, pos. 2, 540°C probe MKS42-1 Studstrup 5 Coal 815°C LA MKS5LAC BA MKS5BA FA MKS5FA

Table 4.1. Overview of experiments, fuels, and the majority of the samples considered in this work. The reference names in the last column are used troughout the report. 5. Characterization of Samples by CCSEM, STA, and HTLM

Inorganic components in straw, wheat grains and ashes and deposits from both straw and straw-coal combustion were characterized by SEM-EDX and CCSEM. A detailed account of the SEM work is given in Appendix E (straw) and G (straw-coal). The ash and deposit samples were also characterized with respect to melting behavior using STA and HTLM.

5.1. Inorganic Elements in Straw, Straw Char, and Grains

The mode and occurrence of mineral components in various fuel samples was examined by SEM-EDX. Fig. 5.1 shows a backscatter electron (BSE) image of a crushed wheat straw pellet in which bright areas represent inorganic components. The inorganic or inorganic-rich constituents show up bright compared to the organic material. A typical feature of the wheat straw is the presence of elongated rims that outline straw fragments. These rims are mainly composed of Si (>90%) and represent parts of the hydrated silicate skeleton that makes the straw rigid enough to stand erect on the field (Marschner (1995)). Additional inorganic particles consist mainly of terrigeneous derived clay or quartz particles. Analyses of the straw sample in areas without any included particles revealed only 0.2% K20 and less than 0.25% Cl on a weight basis. Concentrations in this range are too low to allow quantification on EDX-results, meaning that elements that are dispersed in the organic matrix can be detected, but not quantified by SEM-EDX.

There is an obvious difference in the composition of inorganic particles in the straw and the grain, the straw being dominated by silica whereas the grain is dominated by K and P rich particles. However, inorganic elements that are disseminated or bound in the organic structure is not readily discernible in the BSE-image. This means that, since CCSEM analysis is controlled by contrasts in BSE images, organically associated elements as for example K and Ca are not detected.

Fig. 5.2 illustrates a BSE-image of the rim of a wheat grain. The major part of the inorganic- rich particles are located close to the grain boundary as small spherical particles arranged in clusters. SEM-EDX analyses show that the particles are rich in K, Mg and P, typically present in the approximate ratio of 1:1:3. These particles are phytates which function as the main storage-location of K, P and N in the grains (Marschner (1995)). The association with Mg may ease the accessability of K for growth. Apart from the phytates, only a few terrestrially derived inorganic particles were observed in the grains.

SEM-EDX investigations have been performed of char particles obtained from straw-coal co-combustion experiments at Studstrupvasrket. The chars were sampled from the flame zone using a pyrometer (see Appendix F) and contains straw chars as well as coal chars. Fig. 5.3 shows a backscatter image of a straw char surface. The straw chars are typically covered or partially covered with an appearingly fragile Si-K rich rim sometimes with subordinate Ca. Attached to the rims are small KCI and sometimes K2S04 crystals and a thin appearing K2S04 coating covers much of the rim. The rim has a smooth, irregular outline suggesting they have been partially melted. The observations indicate that KCI and K2S04 have condensed on the Si-K compound either in the flame zone or during sampling. During sampling the flue gas continually flows past the collected, cooled chars and through the filter. Therefore condensation on the chars is likely to occur during this process. Coal chars are typically smaller and rounded compared to the elongated straw chars.

13 Fig.5.1. The BSE photomicrograph shows embedded and polished milled straw particles (grey). The embedding material is dark whereas inorganic material appears bright. A bright rim outlines the rim of an elongated straw particle. This consists mainly of silicon (>90%). Other inorganic par­ ticles are mainly quartz and clay mineral fragments (alumino-silicates).

Fig.5.2. The BSE photomicrograph shows part of the margin of a wheat grain. Spherical K, Mg and P rich particles are arranged in clusters close to the grain margin.

Fig.5.3. The BSE photomicrograph shows typical elongated straw chars up to 2 mm long, retaining parts of the straw structure. Conspicuous Si-K rich rims are commonly present on the surface of the chars. Coal chars are typically much smaller (around 100 microns) and rounded.

— 1 ^ — Fig.5.4. The BSE photomicrographs show laboratory ashes of wheat straw and wheat grains. a)Wheat straw ashed at 815°C (LAS2) b) Wheat grains ashed at 815°C (LAST), c) Wheat straw ashed at 550°C (LAS1). d) Wheat grains ashed at 550°C (LAS6).

— 'j 5 — Occasionally coal derived fly ash particles are attached to the straw chars, but condensation of KCI or K2S04 have not been observed on coal chars.

The CCSEM analyses of wheat straw, char and wheat grains show that the inorganic fraction differs strongly from what is usually observed in coal. For example do important mineral components in coal, such as clay and pyrite, only constitute an insignificant proportion of the minerals in straw. The most important mineral categories in the wheat straw pellets are quartz (mainly hydrated amorphous Si02 x nH20), various phosphor - and chloride rich phases and potassium-sulphate. Wheat grains have a distinctly different association of inorganic particles, being dominated by phospor-rich phases.

5.2. Straw and Grain Laboratory Ashes

The wheat straw and wheat grain samples discussed above were laboratory ashed at 550°C and 815°C. SEM backscatter images of the laboratory ashes are shown in Fig. 5.4. It is seen that the 550°C straw ash has a partly particulate nature, whereas the 550°C grain ash and especially the high-temperature ashes are sintered or melted together, making them unsuitable for CCSEM analysis. Moreover, high-temperature ashing involves some evaporation of K and Cl making it an unsuitable method for compositional characterization of biomass fuels.

A strong difference between the compositions of low temperature ashes from straw and grains was found. The grain ash is dominated by K(Mg, Ca)-phosphates in accordance with the phospor-rich composition of the grains. The low-temperature straw ash is dominated by silicates, KCI and smaller amounts of potassium-sulphate and inclassified chlorides. The high content of silicates is expected due to the high content of silicates in the straw. The silicates in the ash is, however, mainly K-Ca silicates, without any significant amount of aluminium. These particles are believed to have formed during ashing by reaction between dispersed K and Ca with remains of the straw silica skeleton or terrestrially derived quartz. In this respect it is interesting that the straw inorganic was found to contain more than 50% of the quartz category, whereas the low temperature ash contains only 6%. KCI constitutes 14% of the laboratory straw ash whereas this category was not detected in the straw fuel, indicating that KCI mainly is formed from evaporated K and Cl during the ashing process. Alternatively, KCI occurs in the straw as particles that are too small (i.e. < 1 pm) to be detected by CCSEM.

5.3. Straw, Ashes, and Deposits from CHP

The CCSEM study of Haslev straw, ashes, and deposits has shown that KCI only constitutes 2% of the straw fuel inorganic matter, whereas it amounts to 27% in the flyash and as much as 61% to 81% in the deposits. This substantiates that a large fraction of the K and Cl in biomass is present dispersed in the organic material and is therefore not detected by the CCSEM analysis of distinct mineral particles (> 1 pm). Chloride is, however, also present in the unclassified chloride category which may consist of various alkali chlorides. K- Ca silicates, which are characterized by a very low Al-content in contrast to various alumino­ silicates (see Appendix E), constitutes only a minor part of the straw, but is a dominant constituent of the bottom ash. This indicates that a significant part of the organically bound or dispersed K - and to a minor extent Ca - reacts with either remains of the silica-skeleton of the straw or with terrestrially derived quarts grains.

16 In the pie-diagram, Fig. 5.5, the difference in CCSEM mineral composition between bottom and fly ash is readily observed. The bottom ash is dominated by K- and Ca-silicates, quartz and various silicates (alumino-silicates), whereas the fly ash is dominated by KCI and composite particles of KCI and silicate. These composite grains are found mainly in two different forms: they are either formed by KCI condensing on silicate particles, or they represent KCI and silicate particles that are located so closely during the sample preparation that they cannot be distinguished on the BSE image. The SEM backscatter images found in Appendix E show that KCI condensed on silicate particles is a common feature for the straw fly ashes. Fig. 5.6 shows mineral category data for Haslev samples illustrating the dominance of KCI in the deposits compared to the fuel inorganics, bottom ash and fly ash. It is also seen that the compositions of deposits HAS1SD-1 and HAS1SD-2 are similar, as was indeed expected since both are loose deposits from the first and second half of the superheater probe, respectively. Similarly, the deposits HAS1FD-1 and HAS1FD-2, which were taken from the two halves of the furnace probe, are quite alike. It seems, however, that the content of K- and Ca-silicates is a little higher in the first of these. Sample HAS1FD-3 (hard fraction of the furnace deposit) is depleted in K- and Ca-silicates compared to the loose furnace deposits, indicating that formation of the inner hard deposit layer may be dominated by condensation of KCI relative to the outer more loose deposit in which impaction of particles plays a higher role.

The Slagelse 8 experiment distinguishes itself by using rape as fuel rather than wheat or barley as in the other experiments. The difference in ash composition is especially evident for the fly ash which is very different from what is observed in the other experiments, as shown in Fig. 5.5. It contains a high amount of potassium-sulphate (arcanite) and Ca-rich categories in accordance with the Ca- and S-rich character of rape relative to wheat and barley. Additionally the fly ash contains a high amount of un-categorized particles (60% unknown) mainly due to the fact that the data reduction algorithms are primarily designed to categorize wheat and barley ashes. These unknowns have an average composition rich in Ca, K, S and Cl indicating the presence of composite KCI and K-Ca sulphates. The bottom ash from Slagelse 8 does, however, not distinguish itself significantly from the bottom ashes formed during wheat and barley firing. Probably the reason is that some carry-over from the preceeding experiment occurs, meaning that part of the sampled bottom ash may originate from an earlier experiment. The Slagelse 8 experiment only lasted for 4 hours in contrast to 8 hours for the other experiments giving an adverse effect on the carry-over problem.

The compositional differences of bottom ash, fly ash and deposits are reflected in the fusion characteristics of the samples as shown in Fig. 5.7 (HTLM, Slagelse BA), and 5.8 (STA, Slagelse FA). It is seen that two general temperature intervals appear, in which major melting occurs: the low temperature interval ranging from about 650° to 750°C and the high temperature interval ranging above 1000°C. Melt formation in the low temperature melting interval is supposed to be due to melting of salts (predominantly KCI), whereas the high temperature melting interval is supposed to include melting of various silicates. In Fig. 5.7 increasing solid fractions are seen in the temperature range 850° to about 950°C. This appears to be caused by melting of species acting as cement between individual ash particles. These particles are then free to float in the melt and are detected as ‘new ’ solid. It is possible to introduce an automatic correction for this phenomena. SLA1BA SLA1FA SLA7BA SLA7FA

EZ3KCI OCa-Si rich I ~~iKCl+silicate m Ca-rich I—I K-Ca silicate HU unci, chloride I ~l Various AI-silicateslB unci, sulphate 05! Quartz □unclassified rr~1 K2S04

Fig. 5.5. Pie diagram illustrating the difference in CCSEM mineral composition between bottom and fly ashes from the Haslev and Slagelse straw fired plants. The bottom ash is dominated by K- and Ca-silicates, quartz and various silicates (alumino-silicates), whereas the fly ash is dominated by KCI and composite particles of KCI and silicate.

— 18 — Haslev 1

Bottom ash Fly ash HAS1BA HAS1FA

HAS1FD-2 HAS1FD-3 EUKCI r I KCI+silicate I I K-Ca silicate F~n diverse silicates Wa quartz 1 arcanite unclassified chloride H unclassified sulphate I---- 1 unclassified I I iron-oxide

Fig. 5.6. Pie diagram showing mineral category data for Haslev samples, illustrating the dominance of KCI in the deposits compared to the fuel inorganics, bottom ash and fly ash.

— 19 — Temperature (deg. C)

Fig. 5.7. Fusion characteristics measured by HTLM of bottom ashes from the Slagelse plant. Notice the relatively large differences in the fusibilities that reflect variations in the fuel (straw) quality. For comparison, the horisontal lines indicate results obtained by the standard ISO 540 method.

SLA3FA SLA6FA SLA7FA SLA1FA SLA8FA

200 400 600 800 1000 1200 1400 Temperature [°C]

Fig. 5.8. Fusion characteristics measured using STA of fly ashes from the Slagelse plant.

— 20 — Fly ash Bottom ash Dep. pos. 1 Dep. pos. 2 Dep. pos. 3

Temp./°C

Fig. 5.9. STA melting characteristics of ashes and deposits from full-scale, straw-coal co­ firing experiments at Midtkraft.

— 21 5.4. Fuels, Ashes and Deposits from PF

CCSEM analyses of three sets of fly ashes, bottom ashes and deposits collected during co ­ combustion of respectively 0, 10, and 20% straw (on an energy basis) in a coal-/straw PF- fired boiler revealed that far the largest part of the ashes consisted of metal silicates. For the fly ashes, the chemical composition was quite alike on an oxide basis, but for the CCSEM data a clear trend was seen: when potassium was available for reaction (i.e. when straw was burned), a very large part of all present alumina-silicates had reacted to form potassium alumina silicates. Comparing fly ash compositions to deposit compositions, it was found that less different species were present in the deposits. Species being present (concentrated) in deposits included K-, Ca-, Fe-, and Fe-AI-silicates, iron oxide, quartz, illite derived species, and mixed phosphates.

All plant ashes examined showed melting in the temperature range from 1000°C to 1390°C, where the deposits typically showed a higher melt fraction at 1100°C than the matching fly and bottom ashes, Fig. 5.9. This is consistent with the chemical composition of the ashes, which show deposit enrichment in K and Ca, and deposit depletion in Si, compared to the fly and bottom ash. Both these phenomena are known to increase the fluxing of silicate ashes.

In addition, sintering experiments were carried out to investigate the relationship between the melting behaviour of a certain fly ash and the corresponding densification/strength developing process. Strength was found to build up in all ashes at temperatures below the first melt appearance, probably due to the flow of viscous material. For the fly ash collected during coal combustion, high strengths were built up with out any presence of a melted phase, whereas for the ashes produced during partly straw burning, strengths above 4 N/mm2 were not obtained without melt present. Viscosity calculations revealed that for all ashes the sintering onset was equivalent with a viscosity of 1-3-106 Poise. Relating sintering experiment results to real boiler performance, it was interpreted that severe strength build up could happen for deposits in probe position 1 and 2, but not in position 3. These interpretations were in agreement with operational experience.

Blends of straw and coal from the full scale experiments were ashed at 550°C, and their melting behaviors were studied (Appendix F). For comparison a straw/coal ash was prepared by mixing the individual ashes. The tests with mixtures of straw and coal indicate that 20, 40, or 60% of straw on an energy basis have minor fluxing effect in the temperature range 600-1200°C. By adding 80% straw the melting behavior approached that of straw alone. The way the blends were prepared (from fuel mix or ash mix) did not influence the melting behaviour, but CCSEM studies revealed that reaction between straw K and coal S to form potassium sulphate occurs during ashing of the fuel mix. This reaction was not found to take place during full scale co-combustion, where significant quantities of K2S04 was found only in deposits and especially in the superheater deposit from pos. 3 formed at 20% straw share. During HTLM analysis of this sample, pure K2S04 evaporated at about 1150°C (Appendix F).

22 6. Classification of Straw Ashes and Deposits (Triangular Diagram)

Some of the CCSEM data presented in chapter 5 for the Haslev and Slagelse experiments have been plotted in a triangular diagram of K-Ca silicates, quartz+ aluminosilicates and KCI in Figs. 6.1 and 6.2. The background for creating the diagrams was to test a classification into more or less problematic (fusible) ashes seen from a combustion point of view. The three end-members were chosen to represent: 1) relatively refractory quartz and alumino ­ silicates, mostly originating from included terrestrial particles and the straw-derived silica, 2) more fusible K-Ca dominated silicates that most probably are formed by reaction of evaporated K (and to a lesser extent Ca) with quartz grains and the straw silica skeleton, and finally 3) low-melting point KCI. Together the three end-members constitute more than 90% of the bottom ash analyses and typically more than 70% of the fly ash analyses. The straw fuel inorganics are not shown because a significant fraction of the K and Cl present is not represented in the CCSEM data, as discussed previously.

In Fig. 6.1 the strong compositional differences between the bottom ash, fly ash, and deposits from the Haslev experiment are evident. It can also be seen that the outer deposits of the furnace probe (FD-1 and FD-2) are significantly richer in K-Ca silicate than outer deposits from the superheater probe (SD-1 and SD-2) and inner deposit from the furnace probe (FD-3). This indicates that 1) impaction of silicate particles is a more important deposit forming mechanism in the furnace compared to the superheater (where the deposit almost solely consists of KCI), and 2) the initial deposit in the furnace is probably formed by condensation of KCI.

In Fig. 6.2 the CCSEM data from the analysis of Slagelse bottom and fly ashes are plotted together with numbers indicating the percentage of each composition that is represented by the diagram. It can be seen that the bottom ashes are all located at or close to the silicate line, whereas the fly ashes (except FAS) stretch from the KCI apex towards a composition of approximately 85% (quarts + aluminosilicates) and 15% K-Ca silicates. This ratio is higher than in any of the bottom ashes that are all richer in K-Ca silicates. The FAS (rape fly ash) has a distinctly different composition and only 11% of its constituents are represented by the three mineral categories of the diagram.

In Fig. 6.2 the applicability of the triangular diagram as a basis for classification into more or less fusible straw ashes can also be considered. The color of each data point refers to a melting curve with the same color in Fig. 5.7 for bottom ashes and in Fig. 5.8 for fly ashes. A clear trend is evident for the bottom ashes towards higher melting temperatures with an increasing content of quarts + alumino-silicates. This is in accordance with a more refractory nature of quartz and most alumino-silicates compared to K-Ca silicates which may melt at temperatures as low as around 700°C. A similar trend is seen for the fly ashes with an increase in silicates relative to KCI. This is in accordance with the low melting point of KCI (770°C) compared to most alumino-silicates.

Thus it seems that the proposed triangular diagram of CCSEM data may be useful to classify the relative fusibility of biomass ashes, and Fig. 6.2 furthermore illustrates the qualitative agreement between the HTLM/STA results and the CCSEM analysis in terms of assessing relative fusibility of the ashes. In Appendix D this is discussed in more details. K-Ca silicate ■ Bottom Ash A Fly Ash □ Deposits

HAS1BA HAS1FD-1

HAS1FD-2 ,S1FA A\ HAS1SD-1 1 .HAS1SD-2

HAS1FD-3

Fig. 6.1. Triangular diagram showing the compositional differences between the bottom ash, fly ash, and deposits from the Haslev experiment.

K-Ca silicate

Quarts+ aluminosilicates

Fig. 6.2. Applicability of the triangular diagram as a basis for classification into more or less fusible straw ashes. The numbers in the colored boxes indicate the weight percent of the total sample represented by the three categories in the diagram.

— 24 — 7. Prediction of Deposit Formation in Straw and Straw-Coal Fired Plants

Traditionally, three approaches of prediction of ash deposition propensities have been applied: 1) empirical prediction (based on bulk ash chemistry indices, bulk ash fusibility, and plant/operational parameters), 2) thermochemical equilibrium calculations, and 3) mechanistic modelling of the ash formation and transport, and deposit build-up.

Empirical prediction of ash deposition propensities is easy to perform and requires a mini­ mum of fuel (proximate, ultimate and bulk ash analyses) and boiler configurational data (size of furnace, heat input, steam data and production rate, geometry of burner belt etc.). Anyhow, recent work, has revealed that the existing U.S. empirical data material is not always suitable for Danish coal-boiler combinations (Frandsen (1997a)). Moreover, for biofuel firing the traditional (coal) bulk ash chemistry indices are not suitable as the behavior of coal ash is very different. Section 7.1 contains a case study on deposition formation in the Slagelse grate fired CHP, where correlations between fuel and plant data are studied.

A Global Equilibrium Analysis, GEA, based on minimization of the total Gibbs energy of a well-defined chemical system of known temperature, pressure and total elemental composition, may provide valuable flash-type partitioning data, i.e. information about the distribution of elements among and within phases in the system, see Frandsen (1995,1997b). The limitations in this type of analysis was briefly addressed by Frandsen et al. (1996). In section 7.2 comparisons between GEA and plant data from the Haslev and Slagelse CHP are made.

Mechanistic prediction of ash and deposit formation requires detailed models for ash formation, ash transport and deposit build-up. The first basic need is a very detailed information about the abundance and association (mineral grains, organic association, simple salts) of the inorganic metal species in the fuel. Next, a model for fly ash formation is needed. This model should take into account char and mineral grain fragmentation, and coalescence of ash droplets. Then, a model for transport of ash species is needed, taking into account diffusion (Brownian, thermoforetic and eddy) and inertial impaction. Finally, a model for deposit build-up is needed. In section 7.3 mechanistic modelling is discussed in more details.

7.1. Case: Results from the Grate Fired Slagelse CHP

During combustion experiments at the Slagelse grate fired CHP a large number of plant data were monitored, among others the deposition fluxes on a furnace and a superheater probe, respectively, and the quantity of aerosols formed. The bulk chemical compositions and the standard ISO 540 ash fusion temperatures of the straw fuels were also obtained.

In Table 7.1 some of these data are arranged with decreasing fusibilities of the Slagelse bottom ashes detected using HTLM (cf. Figs. 5.8 and 6.2). Table 7.1 shows that the fusibility of the bottom ashes closely correlates with the content of potassium in the ash and in the original straw, and with the total furnace deposition flux. This indicates that by grate firing it may be possible from laboratory analysis to estimate the intensity of furnace deposition. The results from the standard ISO 540 method correlates only to some degree, illustrating the problem about using this method for ash characterization.

25 Kin Si in Kin IDT HT FT Aerosols F dep. flux BA straw straw BA °C °C °C mg/Nm 3 g/m 2/h % dry % dry % dry SIS 1.83 0.56 20 870 970 1090 1142 103 SI6 2.07 1.73 17 830 1110 1440 768 67 SIS 1.69 0.16 17 830 1150 1390 2408 40 SI7 0.66 1.59 14 900 1150 1410 225 19 SI1 0.41 2.03 9 910 1190 1430 86 3

Table 7.1. Various parameters arranged with decreasing fusibility of bottom ashes from Slagelse based on HTLM, Fig. 5.7.

Kin Si in Kin IDT HT FT Aerosols S dep. flux FA straw straw FA °C °C °C mg/Nm 3 g/m 2/h % dry % dry % dry SIS 1.83 0.56 47 650 660 750 1142 17

SI6 2.07 1.73 51 710 740 830 768 - SI7 0.66 1.59 39 710 770 1010 225 7 SI1 0.41 2.03 27 750 930 1170 86 1 SIS 1.69 0.16 33 790 800 1490 2408 120

Table 7.2. Various parameters arranged with decreasing fusibility of fly ashes from Slagelse based on STA, Fig. 5.8.

Table 7.1 also shows that the Slagelse 8 experiment produced a very large quantity of aerosols leading to a high rate of superheater deposition flux. Slagelse 8 was carried out with rape straw as fuel, which has a very low content of silicium and a high content of phosphorus, sulphur, and calcium. Also appearing from Table 7.1 is a negative correlation between the fuel content of silicium and the aerosol production. This means that even though the content of potassium in rape straw is not higher than in the wheat straws from Slagelse 3 and Slagelse 6, the aerosol production is much higher.

In the first column of Table 7.2 the fly ashes from Slagelse are likewise arranged with the lowest melting ash at top and the highest melting ash in the bottom, based on the melting curves obtained using STA (cf. Fig. 5.8). From this it can be seen that the melting behaviour of the fly ashes generally agree with the IDT, HT and FT detected by the standard method. Some correlation is found between the melting behaviour of the fly ash and the quantity of potassium in the ash, but no obvious correlation is found to the content of potassium in the straw fuel or the superheater deposition rate. These findings show that it is not possible to estimate the intensity of superheater deposition from traditional laboratory fuel testing.

The straw fuels used in the Slagelse experiments 1 and 7 (that have the lowest deposition rates) were leached on the field by rain before being combusted. The effect of leaching on the melting behaviour of the ash was also studied in this work by leaching a number of straw samples in water and ashing at 550°C (see Appendix F). The melting behavior of the ash prepared from leached straw was compared with that of the unleached. In addition, blends of straw and leached straw were prepared. The effect of leaching on the chemical composition of straw is to remove most of the potassium and chlorine. Also calcium and carbonates are major elements in the leachate. As the presence of potassium typically lowers the melting of the ash, it was expected that leaching resulted in higher melting ashes.

An example of this is shown in Fig. 7.1, where the fusibilities of bottom ash and fly ash from Haslev 1 are compared with the fusion behaviour of the corresponding laboratory ashes prepared from leached and unleached (standard) straw, respectively. It appears that the melt fraction of the standard laboratory ash is comparable with that of the bottom ash at about 800°C and with that of the fly ash above about 900°C. The fusibility of the leached straw laboratory ash is lower than that of the bottom ash and significantly lower than that of the standard laboratory ash. The blending of leached and unleached straw mentioned above indicated that 80% of leached straw had to be added in order to influence the melting significantly. However, for the straw used in this test the difference between the melting of the standard ash and the ash obtained after leaching was not very remarkable.

One conclusion that can be drawn from the leaching and mixing tests is that the ash content of potassium influences the melting behavior, but the effect is not proportional to the content of potassium in the ash. Generally, leaching results in higher melting straw ashes, but in some cases almost no effect is seen. Therefore no clear conclusion can yet be made in this respect.

----- FA IDT (ISO540)

—ba icrr(isc640) x LA (550 deg. C)

x LA (washed)

x BA

Temperature (deg.C)

Fig. 7.1. Fusibilities of bottom (BA) and fly (FA) ash sampled from the Haslev plant as well as standard 550°C laboratory prepared ash (LA) and laboratory ash prepared from leached straw. The laboratory ashes were prepared from samples of the straw used for the full-scale combustion test.

— 27 — 7.2. Global Equilibrium Calculations, GEA

Global equilibrium calculations were carried out to study the possible agreement with plant data from the Haslev and Slagelse straw fired plants (Appendix F). The input were the elemental concentrations in the straw fuel, the measured combustion temperature, and the air/fuel ratio. A large number of solid phase minerals and gas phase species were taken into account.

For potassium and chlorine the calculated results agree well with plant data. To obtain this agreement, it was found necessary to exclude K2S04(s) as possible specie because otherwise the sulphur and potassium capture in the bottom ashes was far higher than observed. CCSEM analysis of these ashes also showed no presence of K2S04. With this modification the level of fuel potassium evaporated to the gas phase is correctly predicted, as can be seen from Table 7.3. The calculations generally overestimate the measured values somewhat, but notice that the measured values are minimum possible values calculated from aerosol measurements. Variations in the emission levels of HCI from almost zero to more than hundred ppmv are also correctly predicted, Table 7.4. The calculated release of sulphur in the rape experiment, SIS, is very low due to formation of CaS0 4 in the ash as rape straw has a relatively high content of calcium and sulphur.

It was also found that the calculated compositions of the ash phase reflect the trend observed by CCSEM analysis of the bottom ashes sampled during the combustion experiments. That is, the sum of the contents of calcium and potassium silicates in the ash increases in the order SI1 (40.5%w/w), SI7 (53.7%w/w), and SI6 (82.7%w/w). If potassium carbonate is summed as well, SIS fits into this picture with 91 .4%w/w (K2CC3 was not stable in any of the other calculation results). The order SI1, SI7, SIS, SIS is also found if the calculations are carried out at 900°C. At this temperature K2CC3 is unstable and is replaced by K2SiC3 as the stable potassium species. Unfortunately, the presence of the predicted minerals in the bottom ashes can not be verified by CCSEM because the elements K, Ca, and Si are mixed non-stoichiometrically and the individual particles are non-homogeneous.

Meas./Calc. I Experiment % evaporated | SI1 SIS SI6 SI7 SIS Hai Ha2 K 4/8 15/22 10/27 8/11 32/48 24/25 28/26 S 59/35 36/20 43/20 52/52 46/0.1 56/100 47/36

Table 7.3. Comparison of ‘measured ’ and calculated fraction of evaporated fuel potassium. ‘Measured ’ values are actually the minimum values calculated by adding contributions from aerosol measurements and measured flue gas concentrations. For potassium only the aerosols contribute, assuming that no potassium is present in gas phase at the sampling temperature of 120°C (Christensen, 1995).

Meas./Calc. I Experiment ppmv in gas, wet | SI1 SIS SI6 SI7 SIS Ha1 Ha2 HCI 38/22 2.9/0.3 147/288 83/48 18/0.1 87/104 43/185

Table 7.4. Comparison of measured and calculated gas phase concentrations of HCI. In order to make this comparison it must be assumed that HCI passes unreacted from the end of the combustion zone (calculated values) to the flue gas channel (measured values upstream of filter).

28 7.3. Modeling Ash Deposition Fluxes

Applying the simple theory outlined by Rosner (1986), Sarofim and Heible (1994) have reported the times necessary to build up a uniform layer of deposit on a single tubular cylinder, considering each of the mechanisms, diffusion, thermoforesis, and inertial impaction, at a time, see Table 7.5.

Mechanism: Diffusion: Thermoforesis: Inertial Impaction: Time: 46 days 12 days 37 min.

Table 7.5. The time needed to build up a 2 mm uniform layer of deposit on the surface of a cylindrical tube by condensation, thermoforesis and inertial impaction. Source: Sarofim and Heible (1994).

As seen in Table 7.5, inertial impaction is by far the fastest of the three transport mechanisms considered by Sarofim and Heible (1994). Thus, below, as the starting point, a simplified model for quantification of the inertial impaction mass flux, is outlined.

7.3.1. Residual Ash Transport: Inertial Impaction

A model for quantification of inertial impaction, see Appendix H, was tested on fly ash compositional data from a number of Danish and foreign power stations fired with coal, coal blends and co-fired with coal and straw. An example of output from the model is given in Figs. 7.2 and 7.3. Based on proximate and ultimate analyses of the feedstock mixtures utilized in the Midtkraft-Studstrup Power Station, Unit 1, Demoprogramme, experiments no. 3-5 (covering 0-20 % straw on an energy basis) the amount and composition of the flue gas is calculated following the procedure outlined by Frandsen (1995). This gives a fixed flue gas composition as a function of temperature and pressure. The mass concentration of residual ash, based on conversion of 1 kg of feedstock mixture (fuel) is determined as outlined in Appendix H, assuming validity of the ideal gas law.

The impaction probability and stickiness of the fly ash particles are calculated based on CCSEM analyses (see Frandsen et al. (1998)). For each size bin and chemistry in the CCSEM scheme, the impaction and the sticking coefficients are calculated as outlined in Appendix H. It is assumed that the particles and the gas have equal temperatures and the deposit is assumed dry. The mass loading of entrained fly ash particles is calculated from fuel compositional data (Andersen et al. (1997)), assuming that 90 % of the ash introduced to the furnace is entrained as residual fly ash, while the rest is removed as bottom ash (no significant vaporization occurs). 60 £ 50 E 40 - f 30 - a 20- £ 10

0- 0 % 10 % 20 % Straw Straw Straw Feedstock Mixture

Fig. 7.2. Calculated flux of sticking fly ash particles on a single 38 mm outer diameter cylindrical tube in a position equal to the platen superheater, in experiment no. 3 (20 % straw on an energy base), no. 4 (10 % straw) and no. 5 (0 % straw) of the Demopro- gramme at the Midtkraft-Studstrup Power Station, Unit No. 1. For further details on the Demoprogramme: See Andersen et al. (1997).

60

PLSH SSH RH UPSH Boiler Profile

Fig. 7.3. Calculated flux of sticking fly ash particles in different sections of the Midtkraft- Studstrup Power Station, Unit 1, boiler, during experiment no. 3 (20 % straw on an energy base) of the MKS Demoprogramme. PLSH - platen superheater, SSH - secondary superheater, RH - reheater and UPSH - upper primary superheater. For further details on the boiler configuration and operational conditions: See Andersen et al. (1997).

In Fig. 7.2, it is seen that the flux of sticking particles by this model is equal at 0 and 20 % straw share. According to Andersen et al. (1997), the amount of deposit is increased when the straw fraction fired is increased from 0 to 20 %. The reason for this discrepancy may be that the deposit is assumed dry in all simulations shown in Fig. 7.2. Deposit analyses performed at GEUS have indicated that the content of K is significantly increased in the 20 % straw deposits as compared to the 0 % straw deposits. This increase in K-content could very well induce a sticky (liquid) condition on the surface of the 20 % straw deposits compared to the deposits from coal-firing (0 % straw). A sticky surface of the deposit is reported to cause an increase in the deposit formation rate (Walsh et al. (1990)).

30 A calculated inertial impaction ash deposition flux of 55 mg/m 2/s correspond to approximately 200 g/m 2/h. Based on the deposition measurements at the Midtkraft- Studstrup Power Station, Unit 1, Andersen (1998) estimated deposition fluxes in the range 0.9 - 86 g/m 2/h. Considering the simplicity of the model for inertail impaction applied in this work, and the fact that shedding is not taken into account, theoretical and experimental results agree very well.

In Fig. 7.3, the flux of sticking particles in different sections (PUSH - platen superheater, SSH - secondary superheater, RH - reheater and UPSH - upper primary superheater) of the MKS1 boiler is simulated. The reason for the drop in flux from the PUSH to the UPSH is that the viscosity of the fly ash particles decreases significantly during cooling of the flue gas between the PUSH and UPSH. The exponential nature of the Urbain (1981) high-Si ash viscosity model is clearly seen in Fig. 7.3.

Hansen (1997) and Hansen et al. (1997) tried to apply the fraction of melt in a biomass ash directly as sticking coefficient, the basic idea being that the dissipation of the kinetic energy of the in-coming particles can be correlated to the fraction of melt in the ash particles. The conclusion of the work of Hansen (1997) and Hansen et al. (1997) is that a model based on inertial impaction alone or a serial combination of models for condensation (building up eg. a 20 pm thick layer of condensed KCI) and inertial impaction (based on bulk ash chemistry of the fly ash) over-predict the deposition flux measured in the superheater section of straw- fired utility boilers, as long as the fraction of melt in a biomass fly ash is applied directly as sticking coefficient.

As noticed above and seen in Table 7.2, inertial impaction is by far the fastest of the three transport mechanisms considered by Sarofim and Helble (1994). Anyhow, Christensen (1995) reported very high aerosol mass loadings, > 1000 mg/Nm 3, in Danish utility boilers, fired with wheat, barley and rape, which may lead to a significant deposit formation by thermoforesis.

Thus, it was considered important to be able to rank thermoforetic deposition fluxes in biomass-fired boilers. Below, as the starting point, a simplified model for quantification of the thermoforetic mass flux, is outlined.

7.3.2. Submicron Ash Transport: Thermoforesis

Stenholm et al. (1996) and Jensen et al. (1997) have investigated the combustion of twelve well-defined batches of different straws (wheat, barley, rape) at two small combined heat and power production (CHP) boilers in Slagelse and Haslev, Denmark.

Aerosol measurements showed formation of high concentrations of submicron aerosols, 195 - 960 mg/Nm 3 during wheat-firing, and 1010 - 1660 mg/Nm 3 during barley-firing. In the rape experiment at Slagelse CHP, a submicron aerosol concentration of more than 2000 mg/Nm 3 was measured (Christensen (1995), Christensen et al. (1997)). The aerosols consisted almost solely of K, Cl and S.

Based on the aerosol mass loadings, the metal surface and mean gas temperatures, a thermoforetic velocity and thereby a thermoforetic mass flux of aerosols can be calculated and compared to the measured mass deposition flux (Jensen et al. (1997)).

Calculated thermoforetic mass fluxes has been compared with measured total superheater deposition fluxes from the Haslev and Slagelse experiments. The thermoforetic velocity was determined for a 0.1 pm aerosol in a 500 K/cm and a 700 K/cm thermal gradient, see Appendix H.

Fig. 7.4 shows data from the wheat-fired experiments at Haslev and Slagelse, ie. experi ­ ments no. 1, 2 and 4 at Haslev, and experiment no. 1,2, 5, 6 and 7 at Slagelse.

A 500 K/cm A700 K/cm

Measured Flux (g/m2/h)

Fig. 7.4. Calculated thermoforetic mass fluxes vs. measured total superheater deposition fluxes from the Haslev and Slagelse wheat-fired experiments. The thermoforetic velocity was determined for a 0.1 pm KCI aerosol in a 500 K/cm and a 700 K/cm thermal gradient, using the Brock (1962) equation.

Notice in Fig. 7.4, that for deposition fluxes below 15 g/m 2/h, this simple model give reasonable deposition fluxes compared to experimental data.

7.3.3. Summary and Concluding Remarks

A simple engineering model for thermoforetic deposition of submicron ash particles in biomass (straw) fired utility boilers has been set up. Notice in Fig. 7.4, that for deposition fluxes below 15 g/m 2/h, this simple model give reasonable deposition fluxes compared to experimental data. In systems with higher deposition fluxes the thermoforetic model under­ predicts the deposition flux. One reason for this may be that in these systems inertial impaction play a major role in the deposition of ash particles on heat transfer surfaces. As stated by Hansen (1997), a model based on pure impaction will highly over-predict the deposition flux. The ideal case may be a serial or parallel combination of models for thermoforetic and inertial impaction ash transport.

The thermoforesis model applied in this work need to be improved in many points. First, a model for estimation of the thermal gradient above a deposit coated heat transfer tube is needed. Such a model should focus on heat transfer to a tube from a hot surrounding gas and on the thickness of the laminar boundary layer above the tube (deposit) surface. A simple empirical model for this thickness as a function of the fluid dynamic (flow regime) in the system will be tested in near future.

32 — Another point that need improvement is the correlation between the fuel composition and the composition and size of the aerosols. In this study experimental data from a previous work lead to a correlation between the K-content in the fuel and the aerosol mass loading in the Haslev and Slagelse boilers. List of References

S. Ahila and S. R. Iver : “Fireside Corrosion of Superheaters and Reheaters in Coal Fired Boilers ”; Tool & Alloy Steels, 26,45,1992.

K. H. Andersen. P. F. B. Hansen, K. Wieck-Hansen, F. Frandsen and K. Dam-Johansen: "Co-Firing Coal and Straw in a 150 MWe Utility Boiler: Deposition Propensities", Proc. 9th European Bioenergy Conference, Copenhagen, June 24-27,1996.

K. H. Andersen. F. J. Frandsen, P. F. B. Hansen and K. Dam-Johansen: “Full-Scale Deposition Trials at a 150 MWe PF-Boiier Co-Firing Coal and Straw: Summary of Results ”, Proc. Eng. Found. Conf. ‘Impact of Mineral Impurities in Solid Fuel Combustion’ , Kona, Hawaii, November 2-7,1997.

K. H. Andersen : “Deposit Formation in Coal-Straw Co-Combustion ”, Industrial Ph.D.-Thesis, Midtkraft I/S, J. No. EF-582, 1998.

L. L. Baxter. M. F. Abbott and R. E. Douglas: “Dependence of Elemental Ash Deposit Composition on Coal Ash Chemistry and Combustor Environment ”, In: Inorganic Transformations and Ash Deposition During Combustion, S.A. Benson (Ed.), Proc. Eng. Found. Conf., Palm Coast, FA, USA, March 10-15,1991; 1992.

L. L. Baxter : “Ash Deposition During Biomass and Coal Combustion: A Mechanistic Approach ”, Biomass. Bioenergy, 4, 85,1993.

S. A. Benson. M. L. Jones and J. N. Harb: “Ash Formation and Deposition ”; In: Fundamentals of Coal Combustion - For Clean and Efficient Use, L. D. Smoot (Ed.); Coal Science and Technology, Vol. 20; Elsevier, Amsterdam, 1993.

J. R. Brock . J. Col. Sci., 17, 768,1962.

K. A. Christensen : “The Formation of Submicron Particles from the Combustion of Straw ”, Ph.D. Thesis, Dept. Chem. Eng., Techn. Univ, of Denmark, 1995.

K. A. Christensen . M. Stenholm and H. Livbjerg: “The Formation of Submicron Aerosol Particles, HCI and S02 in Straw-Fired Boilers ”. To appear in J. Aerosol Sci., 1997.

I. W. Cummino . Joyce, J. H. Kyle: “Advanced Techniques for the Assessment of Slagging and Fouling Propensity in Pulverized Coal Fired Boiler Plants ”. Journal of the Institute of Energy, No. 58, pp. 169-175,1985.

R. E. Conn . L. G. Austin : “Studies of Sintering of Coal Ash Relevant to Pulverized Utility Boilers. 1. Examination of the Raask Shrinkage-Electrical Resistance Method ”. Fuel, Vol. 63, 1984.

K. E. Eriksen : “Rapport vedrerende bestemmelse af askesmelteforlob pa halmaske og forbrasndingsprodukter i forbindelse med halmfyringsforsog pa Grena Fjemvarmevasrk ”. I/S Midtkraft, Studstrupvasrket, 1989 (in Danish).

F. J. Frandsen : “Trace Elements from Coal Combustion ”, Ph.D.-Thesis, Dept. Chem. Eng., Techn. Univ. of Denmark, ISBN 87-90142-03-9,1995.

34 F. J. Frandsen. L. A. Hansen, H. P. Michelsen, K. Dam-Johansen: “Alkali Metal Partitioning in Hot Flue Gases - Aspects of Equilibrium Modelling ”, Proc. 9th Eur. Bioenergy Conf., Copenhagen, Denmark, June 24-27, 1996.

F. J. Frandsen : “Empirical Prediction of Ash Deposition Propensities in Coal-Fired Utilities ”, CHEC Rep. No. 9703, January 1997a.

F. J. Frandsen : “Ash Chemistry Mapping ”, EU Joule 3 Coal Blends Project 12 - Monthly Progress Report, January 1997b.

F. J. Frandsen . L. A. Hansen, H. S. Sorensen and K. Hjuler: “EFP-95 Project: Characterization of Biomass Ashes - Data Compilation ”, CHEC Report No. 9804, February 1998.

J. R. Gibson . W. R. Livingston : “The Sintering and Fusion of Bituminous Coal Ashes”. In “Inorganic Transformations and Ash Deposition During Combustion ”, Engineering Foundation Conference, Palm Coast, FL, USA, March 10-15, 1991, S.A. Benson (Ed.), New York, NY, USA, ASME.

L. A. Hansen : “Melting and Sintering of Ashes”; Ph.D.-Thesis, Dept. Chem. Eng., Techn. Univ. Denmark, 1997.

L. A. Hansen . F. J. Frandsen, H. S. Sorensen, P. Rosenberg, K. Hjuler and K. Dam- Johansen: “Ash Fusion and Deposit Formation at Straw Fired Boilers ”, Proc. Eng. Found. Conf. ‘Impact of Mineral Impurities in Solid Fuel Combustion ’, Kona, Hawaii, November 2-7, 1997.

P. F. B. Hansen . K. H. Andersen, K. Wieck-Hansen, P. Overgaard, I. Rasmussen, F. J. Frandsen, L. A. Hansen, K. Dam-Johansen: "Co-Firing Straw and Coal in a 150 MWe Utility Boiler: In-Situ Measurements", Proc. Eng. Found. Conf. 'Biomass Usage for Utility and Industrial Power', Snowbird, UT, April 28 - May 3, 1996.

J. N. Harb and E. E. Smith: "Fireside Corrosion in PC-Fired Boilers ”, Prog. Energy Combust. Sc/., 16, 169, 1990.

N. S. Jacobson. M. J. McNallan and E. R. Kreidler: “High-Temperature Reactions of Ceramics and Metals, with Chlorine and Oxygen ”, High. Temp. Sc/., 27, 381, 1990.

P. A. Jensen . M. Stenholm and P. Hald: “Deposition Investigation in Straw-Fired Boilers ”, Energy and Fuels, 11(5), 1048,1997.

K. Laursen : “Characterization of Minerals in Coal and Interpretations of Ash Formation and Deposition in Pulverized Coal Fired Boilers ”. Ph.D. Thesis, Geological Survey of Denmark and Greenland. Report 1997/65, 1997.

W. G. Llovd. J. T. Riley, M. A. Risen, S. R. Gilleland and R. L. Tibbitts: “Estimation of Ash Fusion Temperatures ”. Journal of Coal Quality, Vol. 12, No. 1, pp. 30-36,1993.

W. R. Livingston, Babcock Energy - Fuel Technology Group, Scotland. Personal communication, 1994.

35 H. Marchner : “Mineral Nutrition of Higher Plants ”. 2 Ed., Academic Press, 1995.

H. P. Michelsen. O. H. Larsen, F. J. Frandsen and K. Dam-Johansen: "Deposition and High Temperature Corrosion in a 10 MW Straw Fired Boiler". Proc. Eng. Found. Conf. 'Biomass Usage for Utility and Industrial Power', Snowbird, UT, April 28 - May 3, 1996.

R. E. Raask ACS Div. Fuel Chem., Vol. 27, p. 145,1982.

D. E. Rosner : “Transport Processes in Chemically Reacting Flow ”; Butterworth, London, 1986.

D. E. Rosner . A. G. Konstandopoulos, M. Tassopoulos and D. W. Mackowski: “Deposition Dynamics of Combustion-Generated Particles: Summary of Recent Studies of Particle Transport Mechanisms, Capture Rates and Resulting Deposit Microstructure/Properties ”; In: Inorganic Transformationsand Ash DepositionDuring Combustion, S.A. Benson (Ed.), Proc. Eng. Found. Conf., Palm Coast, FA, USA, March 10-15,1991; 1992.

A. F. Sarofim and J. J. Helble : “Mechanisms of Ash and Deposit Formation ”; In: The Impact of Ash Depositions on Coal Fired Power Plants, J. Williamson and F. Wigley (Eds.), Proc. Eng. Found. Conf., Solihull, UK, June 20-25,1993; Taylor & Francis Ltd., London, UK, 1994.

W. A. Sleaeir . J. H. Singletary, J. F. Kohut: “Application of a Microcomputer to the Determination of Coal Ash Fusibility Characteristics ”. Journal of Coal Quality, Vol. 7, No. 2 pp. 48-54, 1988.

M. Stenholm . P. A. Jensen and P. Hald, “The Fuel and Firing Characteristics of Biomass - Combustion Trials ”, Final Report, EFP Project No. 1323/93-0015,1996 (In Danish).

G. Urbain . F. Cambier, M. Deletter and M. R. Anseau, Trans. J. Br. Ceram. Soc., 80,139, 1981.

S. V. Vassilev . K. Kitano, S. Takeda and T. Tsurue: “Influence of Mineral and Chemical Composition of Coal Ashes on Their Fusibility”. Fuel Processing Technology, Vol. 45, No. 1, p. 27, 1995.

S. E. Wain . W. R. Livingston, A. Sanyal and J. Williamson: “Thermal and Mechanical Properties of Boiler Slags of Relevance to Sootblowing ”. Inorganic Transformations and Ash Deposition During Combustion, Engineering Foundation Conference, Palm Coast, FL, USA, May 10-15, 1991, S.A. Benson (Ed.), New York, NY, USA, ASME, p. 459,1992.

T. F. Wall . S. P. Bhattacharya, D. K. Zhang, R. P. Gupta and He, X.: “The Properties and Thermal Effects of Ash Depositions in Coal Fired Furnaces: A Review ”; In: The Impact of Ash Depositions on Coal Fired Power Plants, J. Williamson and F. Wigley (Eds.), Proc. Eng. Found. Conf., Solihull, UK, June 20-25, 1993; Taylor & Francis Ltd., London, UK, 1994.

P. M. Walsh. A. N. Sayre, D. O. Loehden, L. S. Monroe, J. M. Beer and A. F. Sarofim: “Deposition of Bituminous Coal Ash on an Isolated Heat Exchanger Tube: Effects of Coal Properties on Deposit Growth ”, Prog. Energy Combust. Sci., 16, 327,1990.

— 36 S. Westb.org and C. Nielsen: Analysis of Straw and Straw Ashes, Part 2. Basic Parameters which Demands Further Investigation of Suitable Analytical Methods. Report prepared for IEA Biomass Utilization Task X, ISBN 87-7782-049-5, 1994. Appendix A:

Computer Controlled Scanning Electron Microscopy (CCSEM) Analysis of Straw Ash. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

A Computer Controlled Scanning Electron Microscopy (CCSEM) analysis of straw ash

Henning Sund Sorensen Geological Survey of Denmark and Greenland Thoravej 8, DK-2400 Copenhagen NY, Denmark Phone: +45 3814 2000, Fax: +45 3814 2050, e-mail: [email protected]

Keywords: CCSEM, Scanning Electron Microscopy, biomass, straw ash, combustion.

Abstract This work presents the use of CCSEM analysis for characterization of straw ash. The data-reduction algorithms has been modified from coal mineral analysis to encompass compositions that arise during combustion of straw, in particular wheat. Initial SEM and SEM/EDX studies shows that Si-K rich and Al-poor material is common in wheat straw chars and ash. The separate mineral phases in the straw itself is dominated by Si-rich material whereas dispersed and organically associated material such as K and Cl is not detected by SEM/EDX. CCSEM analyses of 12 bottom and fly ash samples from two straw fired power plants in Denmark show a marked difference between the compositions of bottom and fly ashes. The bottom ash is dominated by silica-rich particles in general, and K-Ca silicates (K-Ca-Si-rich particles) in particular. These particles are suggested to be formed primarily by reaction between straw derived silica and vaporized potassium. In contrast the fly ash is dominated by KC1 and composite particles of KC1 and silicates, whereas K2S04 is rare. K-Ca silicates are also present in the fly ash but in minor amounts than in the bottom ash. In addition to straw ash analysis the modified CCSEM algoritms will show useful for characterization of ash from co-combustion of straw+coal as well as powdery deposits in straw combustion.

Introduction One of the main problems involved in straw combustion is related to the behaviour and fate of the ash forming inorganic species contained in the fuel. These elements can end up in fly ash, bottom ash or may be incorporated in various types of deposits. The inorganic species in straw either occur disseminated or ionically bound in the organic structure, are located in inorganic straw constituents or they are constituents of terrigeneuos dirt that is incorporated in the straw during harvesting and handling. Clearly the fate of the inorganic species during combustion depends on their mode of occurence in the straw. Knowledge of straw ash composition and melting behaviour is important to be able to predict, at least to some extent, the fate of the inorganic elements in terms of practical operation of boilers. The aim is to be able to predict and possibly avoid formation of troublesome deposits and potential corrosion. Additionally, knowledge of straw ash compositions and morphology is useful in terms of possible utilization or disposal of the ash. This paper presents part of a project that aims to acquire knowledge about the formation, composition and physical-chemical behaviour of ash from combustion of straw and co-combustion of straw and coal. One of the goals is to develop or modify useful analysis techniques for straw ash characterization. The techniques utilized in the project involves Computer Controlled Scanning Electron Microscopy (CCSEM), High Temperature Light Microscopy (HTLM) (Hjuler [1997]) and Simultaneous Thermal Analysis (STA) (Hansen [1997]. The CCSEM analysis provides information of the composition and morphology of the ash whereas HTLM and STA provides information about melting temperatures and melting behaviour. The combined data on composition and melting has been utilized to study straw ash and deposits in two Danish straw fired power plants (Hansen et al. [1997]). The CCSEM method is well suited to characterize minerals in coal (e.g. Birk [1989], Zygarlicke and Steadman [1990], Laursen [1997a], Jones et al. [1992], Huggins et al. [1980]). This is especially the case for bituminous or subbituminous coals where the inorganic elements are preferentially located in distinct minerals wheras in brown coals a significant part of the inorganic elements are situated in the organic structure and are therefore not detected by CCSEM analysis. Additional to coal mineral analysis CCSEM analyses of coal ash supplies valuable data on particle composition and morphology, parameters that are important for the behaviour of ash particles in a boiler. Therefore CCSEM data can provide important inputs to modelling of ash deposit formation in coal fired boilers. In straw and other biomass fuels a large proportion of the inorganic species are dispersed in the organic structure and is therefore not detected by CCSEM analysis. Hence, CCSEM analysis of biomass fuel is of limited value on its own, and should at least be coupled with for example chemical fractionation analysis by progressive leaching. However, the CCSEM technique is a strong tool for characterization of biomass ash due to the fact that it provides data on the diversity of ash particles, i.e. each individual particle is analyzed with respect to both size and composition. This is in contrast to traditional bulk chemical analyses which combine all constituents into one batch yielding an “average” composition. This paper deals with the modifications and development of the CCSEM method to be suited for characterization of straw ash and ash from co-combustion of straw and coal. Additionally, the algorithms will be suitable for characterization of deposits formed during straw combustion.

Analytical method Between 0.2 g and 1 g of each sample were embedded in epoxy and after hardening sectioned into two pieces parallel to the settling direction. The two pieces were embedded with the new face pointing upwards to avoid density bias due to settling. During the work it was found that many particles settled to touch each other. This induces an adverse effect on the CCSEM analysis. Therefore some samples were mixed with a filler consisting of 10-50 pm diameter acrylic balls to keep ash particles from touching. The blocks were ground and polished using 1/4 pm diamond powder for the final step. No water or alcohol were used during preparation since both can readily dissolve various salts present in the sample material. Either Buehler Oil or distilled petroleum were used as lubricant and cooling during grinding and polishing. The polished blocks were coated with a thin carbon layer in a Polaron TB500 coater. The SEM and CCSEM work was performed at The Geological Survey of Denmark and Greenland (GEUS) on a Philips XL 40 Scanning Electron Microscope with a Noran Instruments Voyager II X-ray analysis system attached. The use of CCSEM analysis for characterization of minerals in coal and coal ash at GEUS has recently been described in detail by Laursen ([1997a], [1997b]) and therefore only the basics of the technique will be mentioned here:

1. A number of points is set up randomly to cover a large part of the sample surface. At each point an analysis is performed at three different magnifications (25x, lOOx and 500x) to achieve good resolution in the whole size range. 2. A Backscatter Electron (BSE) image with a good separation between inorganic and organic material is acquired (Fig.la). 3. An appropiate grey-level threshold value is set to create a binary image with only inorganic particles singled out (Fig.lb). 4. The binary image is used to control the beam to perform a scan across each individual particle for five seconds. A particle is here defined as a cluster of connecting pixels in the binary image (Fig.lc). 5. An energy-dispersive X-ray-spectrum and morphological data are acquired during each raster scan yielding compositional and morphological data for each individual particle.

Fig.l. Shows the images used for the CCSEM analysis, a) BSE image showing atomic number contrasts. Particles with the highest average atomic number appear brightest, i.e. inorganic particles appear brighter than the organic embedding material, b) Binary image based on the BSE image in Fig.la. The inorganic particles are singled out as white on a black background, c) Sizing image which shows all analyzed particles. Note that only particles which satisfy specific size criteria for this magnification are analyzed. Particles whose center of mass are located outside the marked guard region are omitted.

It shall be stressed that the CCSEM results are semi-quantitative for several reasons: 1. The EDX spectrum is not corrected for ZAP effects (Z: atomic number, A: absorbtion and F: fluorescence). 2. Particle size is based on cross sectional area and weight percent is estimated on the assumption that all particles are spherical. This involves an uncertainty for particles with more complicated shapes. 3. Particles with average diameters below 1 pm are omitted from the CCSEM analysis because much of the detected X-rays are emitted from a volume beyond the particle boundary, potentially leading to erroneous or even meaningless results. 4. Organically bound or otherwise disseminated elements are not detected by the CCSEM method which only “sees” particles with high BSE reflection, i.e. inorganic or inorganic-rich particles. These aspects are, however, not critical for the CCSEM analysis because the X- ray spectra are primarily used to group particles into mineral categories with rather wide ranges of composition. The strength of a CCSEM analysis lies in information about element distribution between different particle types and different samples and in the fact that each particle is characterized with respect to both size and composition. However, it is emphasized that a CCSEM analysis is not directly comparable to a traditional bulk chemical analysis. The raw X-ray data are used to group the analyzed particles into a number of catagories, commonly called mineral catagories. However, the catagories are only defined by composition and bears no information on crystallinity. The data reduction will be described in detail below.

Sample material The sample material used for the present work consists of fly ash, bottom ash and straw chars from full scale experiments and wheat straw and grains from the harvest of 1995. Bottom ash and fly ash were sampled at two danish straw fired boilers, the 23MWth Haslev Combined Heat and Power Plant (CHP) equipped with four cigar burners and the SlMWy, stoker fired Slagelse CHP, in due of an earlier biomass characterization project (Stenholm et al. [1996]). Fly ash were collected in a bag filter at Haslev CHP and in an electro filter at Slagelse CHP. Bottom ash were collected from material that were automatically scraped off the grate in both cases. Four of the experiments utilized wheat straw, one used barley straw and one rape straw. Wheat chars were obtained from a co-combustion (coal-straw) pf-experiment at Studstrupvasrket, Denmark. The chars were sampled from the burner zone with a water-cooled suction pyrometer. The samples are listed in Table 1.

Table 1. Sample description: CODE: FUEL: Ash samples Slagelse 1 bottom ash SL1BA wheat straw Slagelse 1 fly ash SL1FA wheat straw Slagelse 3 bottom ash SL3BA barley straw Slagelse 3 fly ash SL3FA barley straw

Slagelse 6 bottom ash SL6BA wheat straw

Slagelse 6 fly ash SL6FA wheat straw Slagelse 7 bottom ash SL7BA wheat straw Slagelse 7 flyash SL7FA wheat straw

Slagelse 8 bottom ash SL8 BA rape straw

Slagelse 8 flyash SL8 FA rape straw Haslev 1 bottom ash HA1BA wheat straw Haslev 1 fly ash HAIFA wheat straw Chars Studstrup wheat straw char MKS3CH wheat & coal Fuel samples Wheat grains WG Haslev 1 wheat straw HA1ST Results The work was divided into two phases. At first selected ash, char and fuel samples were studied by SEM and SEM/EDX with the intention to provide a basis of knowledge for modifying the CCSEM analysis procedures and data-reduction. Secondly a number of straw ashes and some fuel samples were analyzed by CCSEM and the raw data were used to develop a new set of data-reduction algorithms suitable for characterization of straw ashes and deposits.

Initial SEM investigations Selected ash and fuel samples were investigated by SEM to provide a basis for setup of the CCSEM analysis. The results of this study is summarized below. Figure 2 shows BSE-images of wheat straw, wheat grains, wheat straw char, fly ash and bottom ash. Figure 2a shows the elongated structure of this milled straw sample. Inorganic particles (bright) occur either as equant grains or as elongated rims on the straw fragments. Qualitative SEM/EDX analysis indicate that the former are mainly terrigeneous alumino-silicates and quartz that were incorporated during harvesting and handling. The elongated rims consist almost solely of silica and represent amorphous hydrous silica (opal) which is known to be common in cereal straw JMarschner, 1995]. The presence of these silica- ”skeletons ” is to a large extent responsible for typical high silica-contents in straw and straw ashes. It must be noted that any inorganic species that are disseminated in the organic material is not visible on the BSE-image. Figure 2b shows a BSE-image of wheat straw char, sampled from the burner- zone of a pf-fired boiler co-firing coal and wheat straw. The straw chars are readily distinguished from coal chars by showing relict elongated straw-structures and by being relatively large (up to 2 mm long) compared to the rounded and smaller coal chars (< 200 pm). Many of the straw chars are partly outlined by inorganic rims rich in silica and potassium typically constituting close to 90%. The rims stand out bright in contrast to the organic char material. The Si02-K20 ratio of the rims ranges typically between 4 and 6 on a weight basis and under equilibrium conditions 60-80% will be melted at the peritectic temperature of 769°C in the binary K20-Si02 system [Morey et al., 1931]. Clearly the presence of this low-melting compound is important for the melting behaviour of the inorganic material in the chars and probably in the ash itself. Figure 2c shows a BSE image of a polished embedded sample of mixed straw and coal chars corresponding to the one in Fig.2b. The Si-K rich rims discussed above are readily observed. Qualitative SEM/EDX analysis show that remaining inorganic constituents of the straw chars consist mainly of small (1-5 pm) particles of KC1 and K2S04, that are either attached to the Si-K rims or are located in the char interior. In detail these particles appear to be condensed on the larger char particles. It is very likely that condensation of KC1 and K2S04 occurs during sampling, as both chars and gas are cooled on their passage through the cooled sampling probe. Nevertheless, the presence of condensed KC1 and K2S04 shows that K, Cl and S were present in the gas phase indicating that these species were least to partly evaporated already during this early stage of combustion. This is in agreement with studies of aerosol particles in the plants in question which indicated respectively that 18% K, 66% Cl and 43% S [Christensen et al., in press] and 22% K, 74% Cl and 52% S [Christensen and Livbjerg, 1996] were evaporated from the straw at combustion temperatures. Inorganic particles in wheat grains are clearly different from those observed in the straw (Fig.2d). They consist predominantly of small K, Mg and P-rich, spherical particles that are located in clusters close to the grain margins. These particles are phytates, which are salts of K, Mg and P that function as the main storage sites of these elements in cereal grains [Marschner, 1995]. Based on these observations it is evident that ashes deriving from the studied straw and grains will be of contrasting compositions and that they therefore will pose different practical problems in combustion. The grains will produce ashes (and potentially deposits) rich in phosphor whereas straw will lead to ashes (and potentially deposits) rich in Si, K and possibly Cl. Bottom ash from wheat straw combustion consists mainly of relatively large (up to 1mm) composite Si-rich particles (Fig.2e). However, the bottom ash were milled before analysis so the size distribution does not represent the original conditions. The particles are commonly zoned with grey Si-rich cores surrounded by brighter margins that are enriched in K and to a minor extent Ca (Fig.2f). Minor elements such as Fe, Mg and Na are relatively enriched in the margins as well. The suggestion is that this type of ash particles are formed by reaction between dehydrated fragments of Si-rich inorganic straw components or terrigeneous quartz grains with vaporizedspecies in the gas. As discussed above based on observations of straw chars significant parts of K is likely to evaporate in the early combustion phases. The fly ash samples exhibit a much wider range in particle size than the bottom ash. Fly ash ranges from sub-micron particles that appear as a greyish mass on a BSE image and up to around 150 pm large particles with compositions an features similar to those observed in the bottom ash. Rims (Fig.2g) or aggregates (Fig.2h) of primarily KC1 and to a minor extent K2S04 are commonly located on the larger particles Si-rich particles. Most of the Si-rich particles have subordinate contents of K and Ca and many exhibit some zoning with K-enriched margins similar to the features observed i the bottom ash. Since potassium is known to lower the melting point of silicium glass [Morey et al., 1931] it is evident that the stickiness of such fly ash particles and thereby their propensity to adhere to for example heat exchange surfaces is controlled more by composition of the margins rather than bulk- or core compositions.

CCSEM analyses The strategy for modifying the CCSEM procedures was to obtain an elaborate set of CCSEM raw data on straw ashes from Slagelse CHP and Haslev CHP and subsequently to utilize them to develop a new data reduction algorithm. In total 12 ashes were analyzed by CCSEM. Initially the raw data were reduced by using coal mineral characterization algorithms developed by Laursen [1997b]. Due to the contrasting composition of straw ash and coal ash this resulted in a high number of unclassified particles (35-70%), i.e. particles that do not fit the criteria for any of the remaining catagories. The compositions of the unclassified particles were subsequently studied in detail, and it was found that particles rich in chlorine and sulphur constituted a significant part pointing to the presence of chlorides and sulphates, compounds that have been reported present in biomass ashes elsewhere (e.g. danders and Steenari [1994], Steenari and Danger [1988] and Bryers [1994]). Fig.2. BSE images, a) Wheat straw, b) wheat and coal chars (powder sample), c) wheat and coal chars (embedded and polished), d) wheat grain, e) bottom ash, f) zoned bottom ash particle, g) fly ash particle with rim of KC1, h) Si-rich flyash particles with attached small KC1 and K2SO4 particles. On the basis of semi-quantitive SEM/EDX analysis and corresponding ratios between X-ray intensities a number of new catagories were established in order to characterize the ash particles more completely. After each modification of the algorithms the datasets were re-catagorized and the amount of unclassified particles were taken as a measurement of the success of the algorithm.

Modifications of the algorithm included addition of a number of new catagories. A composite catagory of KC1 associated with silicates (KCl+silicate) were included, an association that is commonly observed in especially fly ash samples. These composite particles either represent Si-particles with attached rims or agglomerates of KC1 (Figs.2g and h) or they result from close association between KC1 and silicate induced during preparation as discussed above. Such closely adjoined particles are seen as one single particle by CCSEM analysis and the result will be an “average” composition with characteristics from all of the adjoined particles. A catagory of particles rich in K, Ca and Si called “K-Ca silicate” were included in the algorithm. Characteristic for this catagory is a very low Al-content in contrast to terrigeneous derived clay-particles. The K-Ca silicates are believed to form by reaction between silica from the straw and vaporized inorganic gas species as discussed above, and are therefore considered to be glassy, amorphous alkali-containing silica glass particles. However, some of the K-Ca silicates may form by reaction between volatilized inorganic species and terrigeneous derived quartz grains, but notably not by reaction with alumino-silicates as evidenced by the low Al-content. Additional catagories for K2S04 and KC1 were included in the algorithm as these phases constitute significant parts of especially the fine fly ash fractions. A range of phosphates: K phosphate, Ca phosphate and K-Mg phosphate were included in order to encompass ash particles derived from wheat grain. The unclassified group were subdivided into unclassified phosphates, -sulphates, -chlorides and -silicates on the basis of their contents of P, S, Cl and Si, respectively, based on the assumption that these catagories will respond differently when subjected to combustion temperatures. Most of the catagories from the coal mineral algorithms were retained to facilitate future comparison between CCSEM analyses of coal and straw ashes and especially for CCSEM analysis of ash from co-combustion of straw and coal. The proportions of catagories typical of coal did not differ outside the estimated analytical uncertainty when going from the initial coal data-reduction to the modified straw algorithms as should be the case if such comparison shall be made.

The modification of the algorithms resulted in a decrease in the proportion of unclassified to between 4% and 25% for bottom ash from wheat and barley straw, whereas fly ash contained between 6% and 27% unclassified. In the case of rape straw unclassified still constitute 60% of the fly ash (SL8FA), underlining that it is problematic to encompass ash from biomass fuels with contrasting compositions into one single data-reduction algorithm. The algorithm presently used is primarily tuned to characterize ash from combustion of wheat and barley straw as well as ashes from co­ combustion of these straw types and coal. Table 2 shows examples of criteria used in the modified algorithm. Table 2 Catagory: Criteria quartz: Al <= 10, Si >= 80 aluminasilicate: Na <= 5, A1 > 20, Si > 2, Si + A1 >= 80, K <= 5, Ca <= 5, Fe <= 5 mixed silicate: Na< 10, Al >20, Si>20, K< 10, Ca< 10, Fe< 10,Na +AI + Si + K + Ca+ Fe>=80 Ca silicate: Na <= 5, Al <=5, Si >20, S <= 5, K <= 5, Ca> 10, Ca+Si >=80, Fe <= 5 KC1: K >= 30, Cl >= 20, K + Cl >= 70 K2S04: K >= 30, S >= 20, K + S >= 70 K phosphate: K>=35,P>=30,K + P>=70 K-Mg phosphate: K >= 25, Mg >= 15, P >= 20, K + Mg + P >= 80 K-Ca silicate: Si >= 20, Si <= 80, K + Ca >= 25, Al<=2 Ca-Si-rich: Si >= 20, Ca >= 20, Si + Ca >= 70, S <= 10, Cl <= 10, P <= 10 KCl+silicate: K >= 20, K >= (0.9 * Cl), K <= (1.7 * Cl), Si >=15

Results from Haslev CHP and Slagelse CHP The CCSEM results for bottom ashes and fly ashes are presented on a catagory basis in Table 3. Some general observations can be made on basis of the ash results. The bottom ashes are typically rich in silica-rich catagories, especially K-Ca silicate, quartz and Si-rich. Only minor KC1 is present and K2S04 is only detected in HA1BA. Unclassified constitute less than 5% in the case of wheat and barley combustion and only 4% in the case of rape combustion, i.e. the data reduction algoritm are considered to characterize the bottom ashes well. The fly ashes are generally rich in KC1 and KCl+silicate and has a rather low but variable content of K-Ca silicates. K2S04 is present in minor quantities except in the rape fly ash (SL8FA) where it constitutes 8%, underlining the contrating composition of rape compared to wheat and barley. Unclassified constitute between 4% and 14% in the case of wheat combustion, 27% for barley and 60% for rape combustion, illustrating that the data reduction algorithm is suited to characterize fly ash from wheat and barley combustion. However, in the case of a fuel of contrasting composition such as rape, a high amount of unclassified particles remain.

Table 3 Catagory HA1BA SL1BA SL3BA SL6BA SL7BA SL8BA HAIFA SL1FA SL3FA SL6FA SL7FA SL8FA K2S04 1,0 1,1 1,7 1,0 8,3 Ca-rich 9,5 Ca-Si-rich 1,1 1,4 K-AI silicate 1,1 1,3 1,0 1,4 1,1 K-Ca silicate 34,1 30,0 83,8 79,3 58,7 67,9 14,8 4,2 2,2 1,0 4,0 10,9 Kaolinite 1,1 KCI 2,3 1,8 26,9 39,2 62,1 85,0 58,3 2,5 KCl+silicate 8,4 19,9 6,5 11,4 1,1 quartz 32,2 47,0 5,2 5,5 19,9 11,0 17,1 12,4 4,4 Si-rich 19,4 20,7 5,5 9,0 15,7 11,3 8,5 4,9 3,5 1,6 unci, chloride 3,5 4,5 1.4 3,9 unci, phosphate unci silicate 1,3 1,2 1,0 1,2 unci, sulphate 1,0 1.7 unclassified 5,1 2,1 1,7 1,3 4,3 13,8 11,6 26,9 3,7 10,4 60,3 Only catagories which constitute > 1% are shown.

To illustrate the variations in composition and in an attempt to correlate catagory composition with melting behaviour of straw ashes, the results were plotted in a triangular diagram of KC1, K-Ca silicates and quartz+almino-silicates (Fig.3). The last group is to be mainly derived from terrigeneous dust and possibly from straw-derived Si-rich material. The three groups constitute more than 90% of the bottom ashes and more than 60% of the fly ashes with the exception of rape fly ash (16%).

K-Ca silicate

SL3BA SL6BA /

SL8BA, o Bottom ash a Fly ash SL7BA,

SL1BAI

SL1FA Quarts+ aluminosilicates ig,3. Fly ash and bottom ash samples from Haslev and Slagelse experiments.

The basis for creating this type of diagram is that it is expected that the melting of the ash will correlate with position due to the contrasting melting temperatures of the end-members. KC1 melts at 770°C whereas K-Ca silicates may melt at temperatures as low as around 740°C if present in proportions close to the eutectic composition [Morey et al., 1931]. However, most of the K-Ca silicate particles have compositions in the range of 5-15% CaO, 70-85% Si02 and 10-25% K20 as illustrated in the three dimensional triangular diagram in Fig.4. These compositions will begin to melt at the eutectic temperature, but the major part will, under equilibrium conditions, not melt until higher temperatures have been reached. The group of “quartz+aluminosilicates” will clearly have a wide range of melting temperatures, but it is assumed that in general the melting will occur at higher temperatures than for K-Ca silicates. This is especially true of the quartz catagory which on its own will not melt until 1713°C [Deer et al., 1966]. Separation between bottom ash and fly ash is distinct in the diagram. Bottom ash is situated close to the silicate baseline, whereas fly ash range from close to the KC1 apex roughly towards a ratio of 85% “quarts+aluminosilicates” and 15% K-Ca silicates. This indicates that the silicate association in the fly ash generally has a lower content of particles formed by reaction between silica and gas alkalies (K-Ca silicates) than what is observed in the bottom ash. The location of S18FA (rape fly ash) is not considered since only 16% of the total is represented in the diagram. A discussion of sample location in the proposed triangular diagram compared to their melting behaviour is discussed by Hansen et al. [1997].

Repeatability Repeatability of the CCSEM analysis were evaluated by preparing three embeddings of one fly ash (SL7FA) and one bottom ash (SL7BA). In this way the errors induced by both sample preparation and analysis procedures were taken into account. The results of the repeated analysis are presented in tables 4 and 5. a) S17BA b) S17FA

CaO CaO

/• 'X a. /A# . k2q ——s® X K20xSid ------^ jmki Si02 RX sio 2

Fig.4. Composition ofK-Ca silicates in three-dimensional triangular diagrams. The relative heights of bars indicate the cumulative weight percent.

The repeated analyses indicate that there is a substantial uncertainty involved in the CCSEM data on catagory basis. The agreement is considered good for bottom ashes where the problem with touching particles is smaller than in the case of fly ash whereas the fly ash shows poorer repeatability. The problem with the fly ashes is that many particles are connected in aggregates (composite particles) which can change the catagory composition significantly even though the total raw data are unaltered. For bottom ash there is good agreement between CCSEM bulk data for the three runs (Table 5). However, in the case of fly ash there is a rather large variations in K20 and Si02. This may be a result of slight variations in the setting of threshold values since much of the masses of submicron material in the fly ash has a rather low BSE reflection. Variations in the threshold setting may cause some of the submicron material to be omitted form the analysis.

Table 4. Weight percent on a catagory basis. Catagory SL7BA-1 SL7BA-2 SL7BA-3 SL7FA-1 SL7FA-2 SL7FA-3 K2S04 1,0 1,0 1,0 0,7 K-AI silicate 1,3 1,6 3,2 1,1 0,9 0,8 K-Ca silicate 58,7 51,8 60,6 4,0 3,8 3,5 KCI 58,3 71,4 64,3 KCI+silicate 11.4 6,1 9,5 quartz 19,9 20,6 14,9 4,4 5,9 8,5 Si-rich 15,7 18,8 16,8 3,5 2,6 3,9 unclassified chloride 3,9 1,2 1,3 unclassified 1,3 1,7 0,8 10,4 4,6 5,8

Table 5. Bulk composition based on CCSEM analysis. SLA7BA-1 SLA7BA-2 SLA7BA-3 SLA7FA-1 SLA7FA-2 SLA7FA-3 SI02 79,1 77,1 78,1 25,9 17,9 25,4 AI203 0,9 1,2 1,3 0,4 0,4 0,4 Fe203 0,0 0,0 0,1 0,0 0,3 0,1 CaO 7,9 8,2 8,0 2,0 2,6 3,0 MgO 0,5 1,2 1,1 0,2 0,4 0,3 Na20 0,0 0,1 0,1 0,0 0,0 0,1 K20 10,3 10,1 10,2 13,1 25,3 22,9 P205 0,4 0,7 0,7 1,2 1,0 1,1 S03 0,3 0,6 0,1 7,1 5,3 6,0 CI207 0,5 0,8 0,2 50,1 46,8 40,8 SL7BA-1 and SL7FA-1 were prepared without mixing with acrylic filler. In conclusion it is noted that this test indicates a satisfying repeatability for bottom ash analysis whereas the fly ash is more problematic. The preparation technique will be further standardized and developed in the future to increase repeatability for straw fly ash.

Summary and Conclusions The CCSEM method has been modified from coal mineral and ash analysis to be suited for characterization of ash from straw and straw+coal combustion. The CCSEM results are internally comparable, but it is emphasized that direct correlation to traditional bulk chemical analysis is as yet not realistic. The strenght of the CCSEM analysis lies in combined compositional and morphological data for individual particles, facilitating advanced modelling of ash behaviour. The internal repeatability of results is satisfying for bottom ash, whereas it is worse for fly ash mainly due to the presence of touching particles. The preparation procedure will be further developed and standardized to increase repeatability. CCSEM analysis of ashes from two straw fired plants showed a marked difference between bottom ashes and fly ashes from wheat and barley combustion. The bottom ash is dominated by silica-rich particles in general and K-Ca silicates in particular. These particles are suggested to be formed primarily by reaction between straw derived Si and K. In contrast the fly ash is dominated by KC1 and combined KCl+silicates. K-Ca silicates are also present but in minor amounts than in bottom ash, whereas K2S04 is only present in minor quantities. In conclusion it has been shown that CCSEM analysis is useful for characterization of ashes from straw and straw+coal combustion. The modified algorithms are also useful for characterization of powdery deposits that contain similar compositions as the straw ash.

Acknowledgements This project has been financially supported by ELKRAFT, ELS AM and the Danish Energy Research Programme.

References: Biyers, R.W. (1994). “Analysis of a suite of biomass samples. Foster Wheeler Development Corporation. FWC/FWDC/TR-94/03. Christensen, K.A. and Livbjerg, H. (1996). “A Field Study of Submicron Particles from the Combustion of Straw.” Aerosol Science and Technology. 25:185- 199. Christensen, K.A., Stenholm, M. and Livbjerg, H. (in press). “The Formation of Submicron Aerosol Particles, HC1 and S02 in Straw-Fired Systems”. To appear in Journal of Aerosol Science. Deer, W.A., Howie, R.A. and Zussman, J. (1966). ”An introduction to the rock-forming minerals.” Longman Group Ltd. Hansen, L.A. (1997). “Ash Fusion Quantification by Use of Thermal Analysis.” Proc. Eng. Found. Conf.: Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7,1997. Hansen, L.A., Frandsen, F.J., Dam-Johansen, K., Sorensen, H.S., Rosenberg, P. and Hjuler, K. (1997). “Ash Fusion and Deposit Formation at Straw Fired Boilers.” Proc. Eng. Found. Conf.: Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7,1997. Hjuler, K. (1997). “Ash Fusibility Detection Using Image Analysis.” Proc. Eng. Found. Conf.: Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7,1997. Laursen, (1997a). “Characterization of Minerals in Coal and Interpretations of Ash Formation and Deposition in Pulverized Coal Fired Boilers.” Ph.D. Thesis, Geological Survey of Denmark and Greenland Report 1997/65. Laursen, K. (1997b). “Advanced Scanning Electron Microscope Analysis at GEUS.” Geological Survey of Denmark and Greenland Report 1997/1. Marschner, H. (1995). “Mineral Nutrition of Higher Plants.” Second edition, Academic Press. Morey, G.W., Kracek, F.C. and Bowen, N.L.(1931). “The ternary system K2G-CaO- Si02.” Journal of the Society of Glass Technology, Vol.14, pp.149-187. Glanders, B. and Steenari, B-M. (1995). “Characterization of ashes from wood and straw.” Biomass and Bioenergy, Vol.8, No.2, pp. 105-115. Steenari, B-M. and Danger, V. (1988). “Fasanalys av sintrade och osintrade halmaskor med och utan tillsats av kaolin respektive dolomit.” Report OOK 88:12. Stenholm, M., Jensen, P.A. and Hald, P., (1996). “Biomasses brasndsels- og fyringskarakteristika. Fyringsforsog. EFP-93.” 1323/93-0015. (in Danish) Appendix B:

Ash Fusibility Detection Using Image Analysis. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

B ASH FUSIBILITY DETECTION USING IMAGE ANALYSIS

Klaus Hjuler dk-TEKNIK Energy & Environment Gladsaxe Moellevej 15, DK-2860 Soeborg, Denmark

Keywords: melting, fusion, quantification, image analysis.

ABSTRACT

This paper describes a new method for the determination of ash fusion behavior. The method is based on high-temperature light microscopy (HILM) combined with real­ time automatic image analysis on a personal computer. The ash is subjected to the analysis placed randomly on a specimen disc. With this method it is possible to observe and quantify even very small changes in the size and shape of single particles or agglomerates. The melting behavior of single particles and agglomerates differ because ash is not a homogeneous material. The HTLM method is suitable for routine laboratory analysis and has been tested on a large number of laboratory prepared ashes as well as ash samples from straw combustion and coal-straw co-combustion. Comparison with results obtained using the ISO 540 standard method has shown that the new method is superior in discriminating between different melting characteristics and for the detection of the first melt or shrinkage of the ash. The use of automatic image analysis avoids any subjective element of the test, and opens new possibilities for studying mechanisms and morphology changes during melting.

INTRODUCTION

In the recent years the use of biofuels for power production has gained increasing importance as a substitute for coal or by being co-fired with coal. This development is particularly forced by the international concern about antropogeneous carbon dioxide emissions. However, the use of biofuels for steam raising or in gas turbine cycles is so far restricted by the fact that biofuels generally have a higher content of potentially deposit forming and corrosive elements than coal, especially on a heat value basis.

The international standard method of estimating the deposit propensity of solid fuels, of which a number of variants exist (e.g. ISO, ASTM, AS, DIN), was originally proposed to estimate the suitability of a particular coal for grate firing. The result from the test is valuable for comparing new coal qualities with known coals that behaves satisfactorily in a specific plant. However, the standard fusion test has shown to be unsuitable for ashes from biomass fuels as the ash test body may ‘blow up’ like a baloon or melt may flow out from it without overall changes in shape, leaving a ‘skeleton ’ composed of e.g. silicon and calcium. This complicates the interpretation and reporting of the test, especially for automatic equipments where the characteristic temperatures are determined only from the height and width of the test specimen. Moreover, the standard fusion test is more or less based on a subjective evaluation of the change in shape - i.e. the silhouette - of a test body (cone or cube) while this is being heated. Consequently the reproducibility as well as the repeatability is poor; even a skilled operator cannot obtain a repeatability better than about ±30°C (ISO/TC, 1991). Another serious problem is that the appearance of the first melt is not detected because it takes place ‘inside’ the test body and does not necessarily affect the shape of the body. The appearance of the first melt is important because the presence of a molten phase increases the probability of ash sticking significantly.

The literature has been reviewed elsewhere [Coin et al., 1996; Hansen et al., 1997; Wall et al., 1996] and is therefore not discussed in this paper. However, it should be mentioned that it is well known that visual observation of an ash sample using light microscopy and a heating stage may reveal details in the process of ash shrinkage and melting during heating (e.g. [Vassilev et al. (1995]). Similar equipment is used in metallurgy for studying recrystallization and sintering processes, where the sample are observed using incident or transmitted light or a combination of both. The main task of the present work has been to develop a method to quantify the information.

EXPERIMENTAL

Apparatus

The apparatus is commercially available: the microscope consists of an Olympus SZ 1145 TR stereo microscope body (1.8 to 11 X) with eyepieces for 30 X magnification, a base illuminator, and a fibre optic illuminator. The camera is a Sony RGB XC-711 CCD and the heating stage is a Leitz 1350, which is suitable for examinations in transmitted and incident light (Fig. 1). The heating stage is powered by a Heinzinger LNG 16-30 power unit. The image grabber is a Neotech colour video digitizer for personal computers. The software was programmed specifically for the image analysis and data acquisition, and for the operation of the heating stage

Samples

More than 80 samples have been subjected to the HTLM fusion test. As the main focus in the present work was posed on developing a method which is well suited for ashes from biofuels, mainly fly ash, bottom ash, and deposits from straw firings have been investigated as well as laboratory prepared biofuel ash. Repetitions of runs with the above mentioned samples and tests with an inert (below 1250°C) quartz fiber sample, a geological standard, and analysis quality salt mixtures have also been performed. All runs were performed at a rate of 10°C/min in a nitrogen atmosphere.

Analysis procedure

Before ashing, coal samples are milled to a maximum particle size of 212 pm, whereas biofuels are milled to a particle size of less than 500 pm. The milled fuel sample is ashed in a Pt-boat in air at 815°C for coal and at 550°C for biofuels. The residual carbon content of laboratory prepared ash is lower than 5 %(w/w). Ash samples originating from a combustor are normally glown for determination of the residual carbon content, that is, the glown ash is used for fusion testing.

Fly ash samples do normally not require milling, while most other ash types do. The maximum particle size should be less than about 100 pm in order to observe a representative number of particles/agglomerates. Coals are milled to a maximum particle size of 76 pm in an agate mortar as described in the standard ISO 540 method. Ashes from biofuels are milled in the same way as coal ashes.

The quantity of sample needed is small, typically about 1 pg (or about 1-2 mm3). In special cases information about the local melting behavior of eg. a deposit is needed. In this case a small piece, say 1 mm3, may be broken off and analysed directly. The sample is placed randomly on a 7 mm diameter sapphire specimen disc by means of e.g. a small spatula. The sapphire disc with sample is placed in the sample holder of the heating stage. When a glass cover is mounted the heating stage is gas tight and can be operated in a controlled atmosphere (normally nitrogen). The microscope is focused on a random part of the ash, the magnification is set at typically 100-120 X, and the light conditions are optimized.

The digital camera and the heating stage is operated from a personal computer. The temperature of the heating stage is ramped at typically 10°C/min from an initial temperature of 550°C for ashes from biofuels and 800°C for coal ashes. The ash sample is photographed initially and then every fifth second (0.2 Hz). The image acquisition rate can be adjusted according to the rate of heating/melting, but 0.2 Hz has been found appropriate for 10°C/min. Each image is processed before the next image is acquired. The results are written to a text file. The power for the heating stage is automatically interrupted when the sample is completely melted or when the temperature exceeds 1250°C (present maximum operation temperature).

Image analysis

The image analysis is based on grey level images and binary images. In the binary images, the ash sample appears black and the background white. The resolution is 384x288 pixels, which gives a manageable file size of 111 kB. At a maximum magnification of 275 X, the corresponding specimen area is 0.15 mm2 and the resolution is 1.4 pm2. At the typical magnification of 100 X, the corresponding figures are 1.1 mm2 and 11 pm2, respectively.

Each grey level image that is acquired during heating of the specimen is converted to a binary image using either a constant threshold value or a subroutine, which calculates a proper threshold value. By using the last method one accounts for an eventual varying light intensity. In the binary image the solid part of the ash appears black and the melt and background appears white. This allows the total area covered by the solid part to be calculated. In addition a difference image is created, which contains information about areas in the actual image that have changed from black (B) to white (W). The analysis result comes as the following two fractions: fj(t) = Sample area in the actual image/initial sample area, and f2(t) = Total area of B-to-W changes/initial sample area.

Areas are measured in pixels and t is the actual temperature in °C. Ideally, the area fractions ft and f2 are complementary, i.e. fj + f2 = 1.

Examples of a grey scale image of a laboratory prepared straw ash and the corresponding binary images acquired during melting are shown in Fig. 2 and 3, respectively.

Calibration

The heating stage is calibrated with materials, which do not form a protective oxide layer. Suitable materials are K2(Cr04)2 (394°C), Ba(N03)2 (593°C), LiF (870°C), Ag (961 °C) og Au (1063°C), where the numbers indicate melting points. The heating stage applied has a maximum operation temperature of 1350°C measured at the specimen glass holder. Due to heat transfer primarily through the top glass seal it has been found that the maximum operation temperature corresponds to a specimen temperature of about 1250°C.

5-point calibrations performed using the above mentioned materials show that the temperature measured at the specimen holder is comparable to the specimen temperature within ±5°C up to about 830°C. Above 830°C the specimen temperature is corrected using a third order polynomial approximation.

RESULTS

As mentioned previously tests were conducted using inert quartz glass fibers at a heating rate of 10°C in nitrogen. Figure 4 indicates that the parameter f2 increases significantly in the range 500-600°C and is not further increased above about 800°C. The cause of this error is movements of the specimen holder relative to the microscope objective caused by thermal stresses during heating. The area fraction f, is not as sensitive as f2 in this respect due to the fact that the area of fibers moving into the image field is approximately equal to the area moving out of the field as the fibers are placed randomly. In principle, f, does not detect sample movements, whereas f2 detects any movements. Therefore, the melting curves discussed in the following are based on the area fraction f, by plotting (l-f1)*100% versus the temperature. Also seen in Fig. 4 are the melting curves for 100% analysis quality KC1 and K2S04 (melting points 774°C and 1069°C, respectively). Initially the melt fractions increase slowly to about 8-10%, which is believed as being neither due to melting and particle shrinkage, nor particle movements. It is not yet clear why area reductions are detected initially with pure crystalline materials. Figure 5 shows the measured melting curves of a KC1-K2S04 mixture (approx. 25 %w/w KC1) and a KCl-CaCl2 mixture (approx. 79 %w/w KC1). The mixtures were prepared from analysis quality reagents, which were milled in a mortar. Also shown are predictions obtained from the relevant phase diagrams. It is seen that the temperatures of significant melt formation are correctly identified. Initially the melt fractions increase slowly as discussed above. When more than about 60% melt has formed the measured results deviate from the predicted. These deviations can be reduced by optimizing the method of threshold value calculation. The threshold value has particularly importance when the melt is not totally transparent.

Also shown in Fig. 5 for comparison is a melting curve for fly ash sampled from a straw fired combined heat and power plant. The ash contains about 58 %(w/w) KC1,18 %(w/w) K2S04, 9 %(w/w) Si as Si02 and 5 %(w/w) Ca as CaO. The initial part of the curve is quite similar to that of the salt mixture, but in this case visual inspection shows clearly that the ash shrinks before melting. The ‘tail’ of the melting curve is caused by the presence of Si and probably also Ca. The sample was completely melted at 1015°C (in nitrogen). This particular sample was also subjected to a Round Robin standard method test (ISO 540). As expected, the results (Table 1) confirmed some of the problems about using the standard test for ashes from biofuels.

Lab. IDT HT FT Apparatus A - 1060 1270 LECO AF600 B 830 1040 1230 LECO AF600 C 690 720 980 Leitz

Table 1. Results from Round Robin ISO 540 fusion test of the straw fly ash in Fig. 5. The results were obtained in reducing atmosphere. A cone shaped test body is used in the automatic LECO equipment, whereas the manual Leitz equipment is operated with cubes. The test results indicate that the shape of the test body may affect the results of the standard method.

An important point is the repeatability of the method that may be influenced by the quantity and distribution of sample on the specimen disc, the magnification, and lighting conditions. Figure 6 shows results obtained when melting a laboratory straw ash, a superheater deposit from straw firing, and a geological standard (BCR-l).Each run was repeated two, two, and three times, respectively. BCR-1 is considered as being very homogeneous, and with this material the magnification was varied from 50 to 150 X to enhance any magnification effects on the results. The resulting melting curves of BCR-1 are nearly identical. Generally, the deviations are greater, but still acceptable, with ashes as exemplified by the straw ash and the deposit. The decrease in the melt fractions in some parts of the curves are caused by rearrangement of solid particles due to flow of melt. This is primarily observed when the quantity of melt produced initially is large as observed with the deposit.

Finally, an interesting feature of the method is that material evaporating from the sample may condense on the colder glass cover of the heating stage. This phenoma has frequently been observed when analyzing samples with relatively high salt contents. The resulting reduction in light intensity is tolerated by the method, except when the quantity of condensated material is very large. In this case the top glass seal can be rotated to a clear part. The condensing particles can be analysed using SEM-EDX directly on the top glass seal. Figure 8 shows an example of KC1 particles condensed in the temperature range 700-800°C during melting of a fly ash from straw combustion.

DISCUSSION

Two fundamental features of the ash fusion determination method presented in this paper is that the sample is placed randomly, and that the fusion characteristic is detected by changes in the area covered by the sample. It is implicitly assumed that changes in the area fraction are related to changes in the solid volume fraction or mass fraction. In the following this is shortly discussed. If A is the measured area, hm the mean height and e the porosity, then the solid volume of ash observed in the image field at a given time may be given by:

Solid volume at time t: V = Aehm(l-e)

Initially during heating a reduction in A/A^ to about 0.95-0.90 due to ash shrinking is typically observed. However, the solid volume is unchanged, and that means that the porosity must decrease (the height does not increase). It is also observed that melting mainly takes place from below due to heat transfer from the specimen disc. In this case A is constant but h decreases, approaching zero as the material is melted. In reality shrinking and melting from below take place simultaneously, and consequently the effect of height reduction may be cancelled by a reduction in the porosity. Based on these qualitative considerations it can be argued that changes in the volume fraction may be approximated by changes in the area fraction. The volume fraction is again related to the mass fraction. These findings are supported by comparison of results obtained with the HTLM metod presented in this paper with results obtained using STA (combined TGA and DSC) [Hansen, 1997].

CONCLUSION

The ash fusibility detection method using image analysis which has been presented in this work is very simple and is suitable for routine analysis. The equipment is robust and stable. The number of samples which can be run per day is high due to the uncomplicated sample preparation and a high cooling rate of the heating stage, which has a low heat capacity. Moreover, the resulting fusion characteristics are produced automatically directly from the measurements without using any expert knowledge or corrections.

The method of image analysis can be refined depending on the application. If the sample is highly agglomerated as biomass ashes typically are, an analysis based on detection of changes in the area covered by the sample is suitable. If agglomeration is not predominant, single particles can be counted, measured, and observed when melting. In this way it is possible to evaluate the melting behavior as function of particle size. Normally details of ash morphology are not reported, but it is also possible to quantify such information.

This work was primarily initiated due to the ‘problematic’ melting behavior of biomass ashes with respect to the standard method. Consequently a maximum heating stage operation temperature of 1250°C was not totally unacceptable. But for coal ashes it is, and work is in progress to extend the operation temperature to about 1500°C. In addition, more samples will be analyzed in order to establish a data base of fusion characteristics and in order to obtain more knowledge about the relationship between mineral composition and melting behavior.

ACKNOWLEDGMENTS

This work was made possible by financial support from ELKRAFT, ELSAM, and the Danish Energy Research Programme. This support is gratefully acknowledged.

REFERENCES

Coin C. D. A., Kahraman H., and Peifenstein A. P. (1996). “An Improved Ash Fusion Test”. Proceedings of the Enginering Foundation Conference on Applications of Advanced Technology to Ash-Related Problems in Boilers, July 16-21, Waterville Valley, New Hampshire. Eds. L. Baxter and R. DeSollar (1996).

Hansen L., Frandsen F., and Dam-Johansen K. (1997). This conference proceedings.

Vassilev S. V., Kitano K., Takeda S., and Tsurue T. (1995). “Influence of Mineral and Chemical Composition of Coal Ashes on their Fusibility ”. Fuel Processing Technology 45, 27-51.

ISO/TC 27/SC 5/WG 6 (1991). “International Round Robin for Determination of the Fusibility of Coal Ash”.

ISO 540, (1981). “Coal and Coke - Determination of Fusibility of Ash”. International Standard Organisation.

Wall T. F., Creelman R. A., Gupta R. P., Gupta S., Coin C., and Lowe A. (1996). “Coal Ash Fusion Temperatures - New Characterization Techniques, and Associations With Phase Equilibra ”. Proceedings of the Enginering Foundation Conference on Applications of Advanced Technology to Ash-Related Problems in Boilers, July 16- 21, Waterville Valley, New Hampshire. Eds. L. Baxter and R. DeSollar (1996). FIGURES

Fig. 1. Microscope heating stage. 1: Specimen disc holder, 2: Top glass seal, 3: Gas inlet, 4: Water inlet, and 5: Radiation shield. The bar corresponds to about 2 cm in length.

Fig. 2. Example of grey scale image of laboratory straw ash. The bar corresponds to about 250 pm.

Fig. 3. Example of some binary images acquired during melting of the same straw ash as shown in Fig. 2. The upper left hand image is the initial image, and the numbers indicate the temperature in degree celcius.

Fig. 4. Test with inert quartz glass fibers, and melting of milled, analysis quality KC1 and K2S04.

Fig. 5. Measured melting curves of a KC1-K2S04 mixture (approx. 25 %w/w KC1) and a KCl-CaCl2 mixture (approx. 79 %w/w KC1), as well as the predicted curves. Also shown is the melting curve of a fly ash sampled from a straw fired plant.

Fig. 6. Results obtained during melting of a straw laboratory ash, a bottom ash from straw firing, and a geological standard (BCR-1). Also shown on the figure are results obtained by the standard ISO 540 method (repeated twice) for comparison with the straw ash melting curve.

Fig. 7. Particles of KC1 deposited on the top glass seal of the heating during melting of straw fly ash. Fig. 1. Microscope heating stage. 1: Specimen disc holder, 2: Top glass seal, 3: Gas inlet, 4: Water inlet, and 5: Radiation shield. The bar corresponds to about 2 centimeter in length.

Fig. 2. Example of grey level image of laboratory prepared straw ash. The bar corresponds to about 250 pm. 755

924 1044

Fig. 3. Example of some binary images from melting of laboratory prepared straw ash. The numbers are the temperature in degrees ceicius. The 550°C image (initial image) is the binary version of the grey level image shown above. Fig. 4

1001 m m

80 —•— K2S04 melt frac. —o — KOI melt frac. o Sample area frac., f1 60 + B-to-W frac., f2

Temperature (deg. C)

dk-TEKNIK/KHJ/23/2/98 % Melt 500

550

600

650 700 700 Temperature dk-TEKNIK/KHJ/23/2/98 ; 750 Fig.

(deg. 5 800

C)

850

900

950

1000 I

FIG. 6

a Geol. std. BCR-1 o Lab. straw ash ♦—ISO 540 (1) ♦- ISO 540 (2) + SH deposit

Temperature (deg. C)

dk-TEKNIK/KHJ/21-10-97 Fig. 7. Particles of KCI deposited on the top glass seal of the heating stage. The particles evaporated from a straw fly ash during melting. The bar corresponds to 5 pm. Appendix C:

Ash Fusion Quantification by Means of Thermal Analysis. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

c ASH FUSION QUANTIFICATION BY MEANS OF THERMAL ANALYSIS

Lone A. Hansen. Flemming J. Frandsen, and Kim Dam-Johansen Department of Chemical Engineering Technical University of Denmark 2800 Lyngby, Denmark

Keywords: Melting, fusion, quantification

ABSTRACT

The fusion of coal and biomass ashes has been quantitatively determined by means of Simultaneous Thermal Analysis, STA. Using STA, melting is detected as an endothermic reaction involving no change in mass. The raw measurement signals are transferred into a fusion curve showing the melt fraction in the ash as a function of temperature. This is done either by simple comparison of the energies used for melting in the different temperature ranges or by accounting for the relevant melting enthalpies. The method repeatability is good, onset determinations and completions generally within 10°C, and melt fractions at given temperatures generally within 10 % melt. Results are presented for simple binary salt mixtures, for which the agreement with fusion as determined by phase diagrams is very good and for measurements of straw (salt-rich) and coal (silicate-rich) ashes. Comparing ash fusion curves to index points of current standard ash fusion tests showed initial melting at temperatures typically between 50°C to 100°C below the IDT. The other characteristic temperatures (Softening, Hemispherical, and Fluid Temperature) corresponded to between 15 and 85 % melt in the ash as determined by the STA method. Thus, characterising the fusion by STA provides a more detailed description of the ash fusion as compared to conventional methods, and the onset of ash fusion is more precisely determined. Furthermore, the method enables identification of the chemical species melting in different temperature ranges. As ash melting has a major impact on the deposit formation tendency, the presented detailed ash fusion determination improves the prediction of ash deposition propensities.

INTRODUCTION

The amount of melt present in an ash as a function of its temperature greatly influences the ash deposition propensity in thermal fuel conversion systems. The appearance of melt is believed to increase both the tendency for ash particles to stick to heat transfer surfaces [Srinivasachar et al., 1990; Walsh et al., 1990; Benson et al., 1993; Richards et al., 1993] and the rate of strength build up in ash deposits [Skrifvars et al., 1996; Benson et al., 1993]. For years, laboratory tests have been carried out on fuel ashes to estimate their melting behaviour, and results have been used to estimate the slagging and fouling propensity of the ashes in full scale combustion systems.

Laboratory tests used to estimate the melting behaviour of ashes include a variety of methods. Very widely/commonly used are the conventional ash fusion tests of which many variants appear [ISO 540,1981; DIN 51730, 1984; ASTM D1857, 1987; AS1038.15,1987]. These methods all imply the controlled heat up of an ash specimen of well defined shape, and the simultaneous determination of temperatures corresponding to specified geometrical shapes. The main criticisms of these tests have been 1) their low reproducibility, 2) unreliability in the subsequent prediction of the ash behaviour in real boilers, and 3)the relevance of the laboratory prepared ash which is subjected to the test [Coin et al., 1996]. It has been emphasized that the initial deformation temperature is not the temperature at which the ash melting begins, and many coal ashes have been found to start melting at temperatures far below the initial deformation temperatures [Gerald et al., 1981; Huggins et al., 1981; Coin et al., 1996; Wall et al., 1996].

Alternatively, the ash melting behaviour has been estimated based on electrical resistivity [Raask, 1979; Sanyal and Gumming, 1981; Gibson and Livingston, 1992; Sanyal and Mehta, 1993] or conductance measurements of the ash [Gumming and Sanyal, 1981; Conn and Austin, 1984; Gumming et al., 1985] during heat up. These electrical quantities reflect the conduction path through the ash sample, and thereby the particle-particle contact and fusion. Both methods detect the onset of fusion in the ash as the temperature at which the electrical properties of the ash is drastically changed. The electrical conductance methods have higher repeatabilities than the standard ash fusion tests and give better predictions of field slagging performance [Sanyal and Mehta, 1993]. However, these methods include some practical difficulties since satisfactory contact between ash and electrodes is hard to achieve and maintain [Wall et al., 1996]. Furthermore, the results contribute primarily with information on the onset of fusion and sintering, whereas the further melt quantity increase in the ash is harder to evaluate based on these methods.

Recently, an improved ash fusion characterisation method based on dimensional changes of ash pellets during heating has been reported [Coin et al., 1996]. In this test four ash cylinders are used as pillars to separate two alumina disks. As the assembly is heated, the ash pellets shrink and the distance between the tiles is measured. Significant tile movement over a narrow temperature range is interpreted to correspond to melting of distinct chemical species, and the repeatability and reproducibility of the method is reported to be high, with reproducibilities below ± 20°C for significant tile movement (> 1.5%/minute).

Finally, ash melting behaviour can be estimated based on calculations. The fusion temperatures may be estimated by combining and weighting the effects of several compositional variables [Winegartner and Rhodes, 1975; Vorres, 1979; Gray, 1987; Lloyd et al., 1989; Vassilev et al., 1995], or by use of chemical equilibrium calculations [Backman, 1989].

As indicated above, the estimation of melting behaviour of coal ashes, (and the subsequent prediction of ash behaviour in real boilers) is not a simple job. Still more problems arise, when trying to do the same job for biomass ashes. The chemical composition of biomass (i.e. in Denmark mainly straw) is veiy different from that of coal and thus the same kind of analyses that are useful for characterising coal ashes do not necessarily apply for biomass ashes. The standard AFT has shown to be unsuitable for ashes from biomass combustion and biomass laboratory ashes [Westborg, 1995]. Thus, the present work was initiated to generate a new method to quantify the melting behaviour of biomass ashes in order to improve the understanding and prediction of ash deposition propensities during firing of biomass. In a longer term, the aim was to apply the method also to coal ashes.

EXPERIMENTAL

Apparatus The new method for estimation of ash melting behaviour is based on Simultaneous Thermal Analysis, STA, and the results presented in this paper were obtained using a NETZSCH STA409. STA implies continuous measurement of sample weight (Thermogravimetric Analysis, TGA) and temperature (Differential Scanning Calorimetry, DSC) during heat treatment. The weight measurement reveals any mass changes taking place in the sample and by comparing the sample temperature to the temperature of an inert reference material, any heat producing or heat consuming (chemical or physical) processes occurring in the sample is detected, and the involved energy subsequently quantified. Test Method STA was carried out on the ash samples, while heating them from 20 to typically 1390°C at 10°C/min in aN2 atmosphere. On the resultant STA curves, melting is detected as an endothermic process involving no change in mass. Melting of a pure substance would be seen as a single endothermic peak in the DSC signal, while for ‘real’ ashes, the melting results in several endothermic peaks overlapping each other, corresponding to melting of the different chemical species in the ash, which melt at different temperatures. Conversion of the STA curves into a melting curve is done based on a DSC signal reflecting only melting energies, which implies that energies related to other processes than melting are first subtracted from the (raw) DSC signal. Evaporation is typically occurring simultaneous to part of the melting, and evaporation enthalpies thus typically need to be quantified and subtracted. Evaporation energies are quantified as the product of a reasonably estimated evaporation enthalpy and the derivative of the TG-curve. After subtraction of evaporation enthalpies, the melting curve calculation can be carried out in one of the following two ways.

The total area below the melting curve, An^, i.e. the area below the DSC curve from the first point where melting is detected, T„ [Backman, 1987], and to the temperature where the melting is completed, T100, reflects the total energy consumed for melting of that ash (Fig. 1). Calculating the area below the DSC curve from any temperature TA to TB, and dividing this area, Aa.b , by the total area below the melting curve, the fraction of ‘total energy used for ash melting’ which has been used in the specific temperature interval, is obtained. This energy fraction is a simple quantitative estimate of the mass fraction of ash melted in the specified temperature interval. This estimate is only correct, if the melting enthalpy of all species in the ash are alike, which is not necessarily true. The presented method thus is a simple way of determining the melting behaviour of an ash and the result expresses the melting behaviourin what could be termed an ‘energy- percentage’ of melt as a function of temperature.

Alternatively, a method based on quantitative determinations of the peak areas can be used. Any given peak below the melting curve corresponds to an absolute quantity of energy used for melting. The position of the peak (onset and peak temperature) gives an indication of the identity of the melting substance(-s), i.e. a reasonable estimation of the relevant melting enthalpy can be made. Based on these two figures, the mass of material melted in the given temperature range can be calculated, and by relating this mass to the total mass of ash analysed, the mass fraction of ash melted in the given temperature range is obtained. Area fraction: A/A^ ~ ash fraction melted The latter method gives the most correct estimates of ‘melted mass fraction’, but this method implies that the substances melting at the different temperature intervals can be identified, so that a reasonable estimation of the involved melting enthalpies can be made, unless the species present in the ash have got Temperature equal melting enthalpies. The latter method thus typically implies an identification of the chemical species present in the ash (as provided by e.g. CCSEM) and detailed knowledge on the chemistry between the ash species.

RESULTS

Simple systems

Temperature T, First, the melting behaviour of two simple mixtures Figure 1: Conversion of STA curve to melting consisting each of only two chemical species will be curve presented. KCI-CaCI2 A sample of approximately 85 mole-% KC1 and 15 mole-% CaCl2 was prepared and analysed in the STA409. The resultant STA curves are shown in Fig. 2. 0 The STA curves show evaporation (of crystal water from CaCl2-H20) at 20- 175°C. This is detected by 1) a decrease in mass (TG) and 2) a consumption of energy (the upward DSC peak). At 594°C, a new endothermic peak starts, peaking at 600°C, but also having a long Figure 2: STA curves for KCl/CaCl2 mixture ‘tail’ behind it, so that the peak is not quite ended until the temperature has reached 696°C. The peak corresponds to the formation of a considerable quantity of melt at the eutectic temperature, and the ‘tail’ corresponds to the continuous increase in melt quantity as the temperature is raised from the eutectic temperature and to the liquidus temperature. Above 700°C, the DSC-signal is greatly increased due to the evaporation of KC1 (seen as the decrease in the TG curve). This experiment was repeated three times.

In Fig. 3, a comparison is made

0.8 - between the melting behaviour g 0.7 obtained by using the lever rule in the % 0.6 - phase diagram [Levin et al., 1964] and

x 0.4 - the ones obtained when quantifying s 0.3 ■ - and comparing the areas below the 0.2 ------Exp. DSC curve for each of the three 0.1 ■- experiments. It is seen that the theoretical curve predicts the melting to take place instantaneously, which is not happening in reality, but except for Figure 3: Melting curves for KCl/CaCl2-mixtures this, a very good correlation between theoretically and experimentally obtained melting curves is found. The deviation between the two types of curves is 10 % at maximum (neglecting the temperature range, 594-608°C, just above the eutectic temperature). Melting onset and completion deviate less than respectively 3 and 15°C from the phase diagram values. The method repeatability data is given in Table 1. Repeatability is seen to be very good, deviances are 4°C for melting onset, 4% for the melt fraction obtained at the eutectic temperature (calculated at 608°C, which is the end of the large peak), and 11°C for melting completion. Concerning the completion temperatures, the phase diagram shows that increasing KC1 fractions increases the liquidus temperature, and this tendency is also found in the experimental results.

Xmix[mole%] To [°CJ %melt (Tcut ) Tioo [°C] A«U, [J/g]

theory (0.85) 594 45.2 692 309.5-312.5 »

exp. 1 0.853 597.1 49.4 706.7 99.7 %2)

exp. 2 0.848 593.9 45.3 703.5 99.7 % 2)

exp. 3 0.841 593.7 49.6 695.6 99.0 % 2) !) dependent on system composition;2) fraction of theoretical melting enthalpy Concerning the measured melting enthalpies, a very good correlation to theoretical calculations is seen: the measured energies correspond to between 99.0 and 99.7 % of the theoretical value. Table 1: Repeatability for melting curves for mixture of KCl/CaCl2 The theoretical and experimental description of the melting behaviour are thus judged to correlate well in this case.

KCI-K2S04 The same analysis and comparison was made for a mixture of KC1 and K2S04, the result of which is shown in Fig. 4. As can be seen, the experimental curve overestimates the quantity of melt formed at the eutectic temperature with approximately 4% compared to the theoretical prediction. At temperatures above 700°C, the deviation between experimental and the theoretical prediction varies between 4 and 7%, until the curves meet at 860°C. The temperature differences between the transition temperatures given in the phase diagram and the experimentally determined ones are judged to be acceptable. This example therefore confirms the above and supports the assumption that STA measurements are able to describe melting behaviour of simple systems.

Ash Samples Several fly, bottom and deposit ash samples from 1) grate fired units firing pure straw and 2) PF- fired boilers co-fired with straw and coal have been investigated. In this paper, examples of the results will be given, representing melting behaviour results for a fly ash and a bottom ash collected during a test run of different straws at a grate fired boiler, and a fly ash collected during (pure) coal firing at a PF-fired boiler.

The chemical composition of the ashes is given in Table 2, and a reduced data set of CCSEM results showing the dominant species in the ashes in Table 3. The STA curves for the fly ash

Table 2: Chemical composition of investigated ashes [%(w/w)] 8" SiOz ai2o3 Fe^ CaO MgO Na20 k2o p2o 3 Cl

FA,straw 27.0 0.38 0.66 4.7 0.82 0.7 32 8.7 2.0 17.0

BA,straw 46 0.84 0.34 15 2.5 1.6 24 0.29 2.6 0.29

FA,coal 55.7 18.1 6.4 1.5 1.5 0.49 2.0 0.41 0.15 - FA = fly ash, BA = bottom ash

Table 3: Selected CCSEM data showing main constituents for the investigated ashes (modified from Sorensen, 1997)'

KC1 K-/Ca-silicates* Al.-silicates Si02

FA, straw 49 14 - 12

BA, straw - 84 - 5

FA, coal - - 72 7 * K-, Ca-, and Si-rich amorphous particles collected at the straw fired boiler is shown in Fig. 5, ■100 where the increase in curve complexibility (compared to Fig. 2) is obvious. Referring to 0 - Mtk" Fig. 5, the DSC curve is seen to show a distinct endothermic peak from 641.0°C to 712.0°C corresponding to an energy consumption of 176.14 J/g. As there is no simultaneous decrease in mass (TG curve), Temperature/pC] this peak corresponds to the Figure 5: STA curves for fly ash from straw combustion onset of the ash melting. For increasing temperatures, a general increase in the DSC signal is seen. On top of this general increase, two distinct peaks are seen, one ranging from app. 920°C to 1050°C and one starting at app. 1150°C, which is not completely finished at 1250°C. The first of these peaks is seen to occur simultaneously to a large decrease in mass, and since the shape of the DSC peak and the d(TG) peak are quite - but not totally - alike, a large fraction - but not all - of the energy corresponding to this peak is used for evaporation of material rather than melting.As described earlier, the evaporation energies are estimated as the product of an estimated evaporation enthalpy and the d(TG) curve. In this case the evaporation enthalpy of KC1 has been used, since KC1 constitutes a large part of this ash (app. 40 % (w/w)) and is assumed to evaporate at these temperatures. For the last DSC peak, the simultaneous mass decrease is veiy low, and thus the energy corresponding to the area of this peak is predominantly used for melting of ash: The general increase of the DSC curve is caused by the fact that when great mass losses are occurring (as for this sample) the DSC baseline is shifted upwards [Netzsch, 1995]. This explanation is supported by the slope of the DSC curve, which is quite higher around the temperatures of rapid evaporation (800-1050°C) compared to above 1100°C, where the fast evaporation is ended, and the DSC baseline has found a new level at app. 1.5 mW/mg.

A typical set of STA curves for a silicate rich ash - e.g. a fly ash produced during coal combustion - is shown in Fig. 6. A typical feature is the long lasting DSC peak occurring at temperatures between 20°C to 600°C, which is believed to represent decomposition of clay minerals in the ash. Above those temperatures, nothing seems to occur until the DSC curve „ , starts increasing significantly at Figure 6: STA curves for fly ash from coal combustion 1180=c> which ^presents the melting onset. The DSC curve continuously increases until the termination of the experiment at 1390°C, at which point the DSC peak is not ended; i.e. the ash is not completely melted at 1390°C. Identification of chemical species melting in different temperature ranges above 1180°C is not possible, why the melting curve is calculated based on area comparison (method no. 1), as is typical for silicate dominated ashes. The fraction of melt formed at 1390°C is determined by studying the sample structure (after cooling) in the SEM. The material which has not been molten has maintained its original structure, and the melt fraction determination is made based on an area evaluation of material of original structure to that of fused material. In Fig. 7, the melting behaviour calculated on the basis of the STA straw fly ash | curves for the three ashes is shown. c 0.7 - For the straw derived fly ash, two » 0.6 - - curves are shown, representing 2 0.5- respectively the comparison of areas 0.4- E 0.3 under the DSC curve (method 1), and the calculation that includes the absolute energy represented by the first melting peak and the relevant melting enthalpy (method 2). The Temperature/pc] first melting peak for this fly ash is supposed to represent melting in a salt system containing large amounts ofKCl and only minor quantities of ‘other’ K- and Ca-salts, since the temperature agrees with the eutectic temperature for these systems and since these species have been found in the ash by means of CCSEM. As it is seen, the curve including the melting enthalpy corresponding to the first melting peak shows a larger fraction to melt at 653°C than the curve based on area comparisons (52% > < 21%). This reflects that the melting enthalpies for the potassium and calcium silicates melting at the higher temperatures are higher than the melting enthalpy for KC1.

Comparing the three melting curves, it is seen that generally the fly ash from straw combustion is Tower melting’ than the bottom ash from straw combustion, which is again Tower melting’ than the fly ash from coal combustion. Comparing the Figure 7: Melting curves for two fly ashes and a bottom ash melting curve for the straw-derived fly ash with that for the straw-derived bottom ash, it is seen that at temperatures between 600 and 1100°C, the melt fraction is considerably lower for the bottom ash than for the fly ash. This is due to the high content of simple salts in the fly ash. For temperatures between 1100 and 1250°C, the bottom ash shows higher melt fractions than for the fly ash. This is probably due to the fact that the bottom ash is highly dominated by K- and Ca-silicates (i.e. K-, Ca-, and Si-rich compounds) whereas the silicate part of the fly ash contains larger fractions of the more refractory quartz [Sorensen, 1997]. The coal derived fly ash starts melting at the highest temperatures, with initial melting at app. 1180°C, and only partly melting (—60%) at 1390°C. This is due to the ash consisting almost entirely of various alumino silicates and quartz [Sorensen, 1997] which melt at relatively high temperatures. In conclusion, the melting curves are thus seen to reflect the different chemical composition of the ashes. Repeatability of melting curves Figure 8 shows melting curves for the two fly ashes as obtained during repeating experiments. Starting with the straw-derived fly ash, it is seen on the curves, that the reproducibility of the melting onset is very good, within 5°C, as well as is the slope of the first part of the curve. The first curve part corresponds to the very distinct peak occurring at low temperatures (641-712°C in Fig. 5), and since this peak is dependent on the salt chemistry of the sample, and this chemistry is quite simple (i.e. includes only a few possible reactions between the present species), the melting peak occur at precisely the same temperature every time. The melt fraction obtained at these first peaks do deviate slightly, though; in this case the level obtained is 50, 52 and 47 % melt respectively. This deviation is caused both by method uncertainty but may also be influenced by the inhomogeneity of the sample. For the rest of the melting curve, the uncertainty is somewhat larger, but still within 10 % melt. The larger uncertainty for the last part of the curve (> 800°C) is due to the less distinct peaks corresponding to the melting of the silicate part of the ash. Since the peaks corresponding to silicate melting are less distinct, a precise characterisation is dependent on a very well known baseline. For higher temperatures, drift in the DSC baseline cannot be avoided [Netzsch, 1995]. This leads to larger uncertainties for the melting curve. As stated above, the uncertainty is still within 10% melt, though. For the coal- 5 0.7 FA straw FA coal derived fly ash, which consist = 0.6 5 0.5 mostly of silicates, the problem with not very distinct peaks may 2 0.3 generally lead to a larger uncertainty for the melting onset, but as can be seen, the repeatability is still quite good: melting onset Temperature/TC] varies within 30°C, and melt Figure 8: Repeatability of melting curves fractions at given temperatures are within 10% melt.

To reduce uncertainties from sample inhomogeneity, repeatability was also tested by analysing a well-characterised and homogenized geological standard material (BCR1). Melting curves for these measurements are also shown in Fig. 8, and reveals that onset is determined with a deviation of 15°C, melting completion temperature with a deviation of 5°C, and the melt fractions at given temperatures deviate at maximum 14% (melt). Based on this, method repeatability is generally judged to be quite high.

DISCUSSION

Correlation to standard AFT The results of the standard AFT (DS/ISO540) are marked on Fig. 7. For all ashes, melt is formed at temperatures considerably below the IDT, respectively 105,42 and 65°C for each of the three ashes. This is consistent with previous criticism of the standard ash fusion tests [Gerald et al., 1981; Huggins et al., 1981; Coin et al., 1996, Wall et al., 1996]. For the straw fly ash, the STA predicts significant melt formation (51%) below the IDT. On the other hand, the melt fraction does not increase much for the next characteristic temperatures, since the hemispherical and the fluid temperature corresponds to a melt fractions of respectively 53 and 62 % melt. For the bottom ash, the AFT temperatures seem to give a better description of the increasing melt fraction, with the three characteristic temperatures corresponding to respectively app. 3,14 and 43 (energy) percent melt. For the coal-derived fly ash, the initial deformation temperature corresponds to 5 % melt, whereas the hemispherical and the fluid temperature are both higher than the present maximum analysis temperature of 1390°C and the only relation that can be given is that they correspond to more than 60% melt in the ash. Overall, the standard AFT and the STA melting behaviour curves thus seems to correlate qualitatively, as the three characteristic temperatures are all located in the temperature range corresponding to 3 to 65% melt in the ash. A further and more detailed comparison between the standard AFT and the STA melting behaviour results is presented else where [Hansen, 1997; Hansen et al., 1997].

Correlation of results with ash chemistry (mineralogical changes) The DSC signal is closely related to the chemical composition of the ash, since the melting peaks indicate the melting point of either single or mixtures of ash species. Identification of the species melting in the different temperature intervals can thus be made based on CCSEM and/or XRD analysis of the ash and relevant phase diagrams, supplied with STA of simple synthetic ashes, if necessary. For the less distinct ‘peaks’, the interpretation is correspondingly less certain, but typically wide melting (temperature) ranges can be correlated to the species melting. A detailed comparison between STA melting curves and CCSEM compositional data for ashes collected at straw fired boilers is provided elsewhere [Hansen et al., 1997].

Method Limitations At present, the described method works as an expert tool. The melting behaviour measurements are easy and simple to perform, the repeatability of the results is good, and the measurement procedure can easily be standardized. The interpretation of the STA signal, that is, the conversion from the STA curves to the melting curve, is on the other hand not simple and can not at the moment be standardized. The ‘certain’ interpretation requires detailed knowledge on the chemical species in each ash sample (as provided either by CCSEM or XRD analysis) and the chemistry between these. This is necessary for expressing the melt fraction as a mass percent (method no. 2), but is also important to avoid that energies related to solid phase transitions occurring simultaneously to the melting will be wrongly detected as so (melting).

Result Applicability The new method provides an improved and more detailed characterisation of the melting process occurring in ashes during heating. The results reveal/provide the temperature for which the first melt is formed in the ash, and gives therefore important information for boiler designers. Furthermore, the melting curves can be used as input to mechanistic modeling of ash deposit formation by inertial impaction, which will hopefully improve the understanding of ash deposit formation mechanisms during biomass and/or coal combustion.

CONCLUSIONS

A new experimental method for quantification of ash (solid combustion residues) melting has been developed. Using the new method, a conventional STA apparatus is employed, and the melting is detected as endothermic reactions involving no change in mass. The DSC signal is transferred into a melting curve (showing the melt fraction in the ash as a function of temperature) either by simple comparison of the areas below the melting curve or by accounting for the relevant melting enthalpies. The execution of the measurement is simple and the repeatability of the results is very good. The subsequent conversion of the STA curves to a melting curve requires knowledge of the identity of chemical species in the ash and the involved chemistry.

The method has so far been tested on a number of simple salt mixtures, for which the measured melting behaviour agrees with the predictions from phase diagrams, and a number of ashes collected during combustion of pure straw, co-combustion of straw and coal, and coal combustion, for which melt was detected between 40°C and 110°C below the corresponding IDT. Characterising the fusion by STA provides a more detailed description of the ash fusion as compared to conventional methods, and the onset of ash fusion is precisely determined. Furthermore, the method typically enables identification of the chemical species melting in different temperature ranges. As ash melting has a major impact on the deposit formation tendency, the presented detailed ash fusion determination improves the prediction of ash deposition propensities.

ACKNOWLEDGEMENTS

This work was carried out as part of the Combustion and Harmful Emission Control (CHEC) research program at the Department of Chemical Engineering, Technical University of Denmark. The CHEC research program is cofunded by ELSAM (The Jutland-Funen Electricity Consortium), ELKRAFT, the Danish Technical Research Council, the Danish and the Nordic Energy Research programs. Mrs. Gurli Mogensen (from Haldor Topsoe A/S) is acknowledged for her valuable contributions to the interpretation of the DSC curve behaviour during the long and troublesome running-in of the apparatus.

REFERENCES

AS1038.15 (1987). “Methods for the analysis and testing of coal and coke - fusibility of higher rank coal ash and coke ash”. Australian Standards Association. ASTM D1857-87 (1987). “Standard test method for fusibility of coal and coke ash”. American Society for Testing and Materials. Backman,R (1989). Sodium and Sulfur Chemistry in Combustion Gases. Academic Dissertation, Abo Academy University, Abo, Finland. BackmanJR, Hupa,M., and UppstuJE. (1987). “Fouling and Corrosion Mechanisms in the Recovery Boiler Superheater Area” Tappi Journal 70 (6) Benson,S A., Jones,M.L., and Harb,J.N. (1993). “Ash Formation and Deposition”. In L. D. Smoot (Eds.), Fundamentals of Coal Combustion for Clean and Efficient Use New York: Elsevier Coin, C.D.A., Kahraman, H., and Peifenstein, A.P. (1996). “An Improved Ash Fusion Test ”. In L.L. Baxter and R. DeSollar (Eds.) Proceedings of the Engineering Foundation Conference, 1995, New York and London: Plenum Press Conn,RE. and Austin,L.G. (1984). “Studies of Sintering of Coal Ash Relevant to Pulverised Coal Utility Boilers. 1: Examination of the Raask Shrinkage-Electrical Resistance Method”. Fuel 63 Gumming,I.W. and Sanyal,A. (1981). “An Electrical Conductance Method for Predicting the Onset of Fusion in Coal Ash”. In Richard W. Biyers (Ed.) Proceedings of the Engineering Foundation Conference, Engineering Foundation Gumming,I.W., Jotze,W.I., and Kyle,J.H. (1985). “Advanced Techniques for the Assessment of Slagging and Fouling Propensity in Pulverised Coal Fired Boiler Plant”. Jour. Inst Energy (58) DIN 51730, (1984). “Determination of Fusibility of Fuel Ash”. German Standard Gerald,P.H., Huggins,F.E., and Dunmyre,G.R (1981). “Investigation of the High-Temperature Behaviour of Coal Ash in Reducing and Oxidizing Atmospheres”. Fuel (60) 585 Gibson,J.R and Livingston,W.R. (1992). “The Sintering and Fusion of Bituminous Coal Ashes”. In S.A. Benson (Eds.) Proceedings of the Engineering Foundation Conference on Inorganic Transformations and Ash Deposition During Combustion. New York: ASME Gray,V.R (1987). “Prediction of Ash Fusion Temperatures from Ash Composition for Some New Zealand Coals”. Fuel (66) Hansen, L.A. (1997). Melting and Sintering of Ashes. Ph.D. Thesis; Department of Chemical Engineering, Technical University of Denmark Hansen,LA., Frandsen,F.J., Sorensen,H.S., Rosenberg, P., Hjuler,K., and Dam-Johansen,K. (1997). “Ash Fusion and Deposit Formation at Straw Fired Boilers”; these conference proceedings Huggins,F.E., Deborah,A.K., and Huffman,G.P. (1981). “Correlation between Ash-Fusion Temperatures and Ternary Equilibrium Phase Diagrams”. Fuel (60) 577 ISO 540 (1981). “Determination of Fusibility of Ash”. International Standard Organisation Levin,E.M., Robbins,C.R, and McMurdie,H.F. (1964). Phase Diagrams for Ceramists, Vol. 1. Columbus, Ohio: The American Ceramic Society Lloyd, W.G., Riley,J.T., Risen,M.A., Gilleland,S.R, and Tibbits,RL. (1989). “Estimation of Ash Softening Temperatures Using Cross Terms and Partial Factor Analysis”. Energy and Fuels (4) Netzsch, (1995). Personal Communication RaaskjEJ. (1979). “Sintering Characteristics of Coal Ashes by Simultaneous Dilatometry-Electrical Conductance Measurements”. Journal of Thermal Analysis (16) 91 Richards,G.H., Slater,P.N., and Harb,J.N. (1993). “Simulation of Ash Deposit Growth in a Pulverised Coal- Fired Pilot Scale Reactor ”. Energy & Fuels (7) 774-781 Sanyal,A. and Gumming,I.W. (1981). “An Electrical Resistivity Method for Detecting the Onset of Fusion in Coal Ash”. In R W. Dryers (Eds.) US Engineering Foundation Conference on Fouling and Slagging Resulting from Impurities in Combustion Gases. New York: Engineering Foundation Sanyal,A. and Mehta,A.K. (1994). “Development of an Electrical Resistance Method based on Ash Fusion Test”. In J. Williamson and F. Wigley (Eds.) Engineering Foundation Conference on Impact of Ash Deposition on Coal Fired Plants. Washington:Taylor & Francis Skrifvars,B.-J., Backman,R, and Hupa,M. (1996). “Ash Chemistry and Sintering”. Preprints of Papers Presented at the 211th ACS National Meeting, New Orleans, LA, March 24-28,1996 Srinivasachar,S., Helble,J.J., Katz,C.B., and Boni,A.A. (1990). "Transformations and Stickiness of minerals during pulverised coal combustion". In R W. Bryers and K. Vorres (Eds.) Proc. of the Engineering Foundation Conference on Mineral Matter and Ash Deposition from Coal. New York:Engineering Foundation Sorensen, H.S. (1997) “Computer Controlled Scanning Electron Microscopy of Straw Ash”. These conference proceedings Vassilev,S.V., Kitano,K., Takeda,S., and Tsurue,T. (1995). “Influence of mineral and chemical composition of coal ashes on their fusibility”. Fuel Processing Technology (45) Vorres,K.W. (1979). “Effect of Composition on Melting Behaviour of Coal Ash”. Journal Eng. Power (101) Wall,T.F., Gupta,RP., Polychroniadis,P., Ellis,G.C., Ledger,RC., and Lindner,E.R (1989). “The Strength, Sintering, Electrical Conductance and Chemical Character of Coal Ash Deposits”. NERDDC Project No. 1181 - Final Report, Vol. 1, Summary Report Wall, T.F., Creelman, RA., Gupta, R.P., Gupta, S., Coin, C., and Lowe, A. (1996). “Coal Ash Fusion Temperatures: New Characterisation Techniques and Associations with Phase Equilibria”. In L.L. Baxter and R DeSollar, Proceedings of the Engineering Foundation Conference on Applications of Advanced Technology to Ash-Related Problems in Boilers. New York: Plenum Press Walsh,P.M., Sayre,A.N., Loehden,D.O., Monroe,L.S., Beer,J.M., and Sarofim,A.F. (1990). "Deposition of Bituminous Coal Ash on an Isolated Heat Exchanger Tube: Effects of Coal Properties on Deposit Growth" Prog. Energy Comb. ScL (16) 327-346 Westborg,S. (1995). Round Robin Test - Analysis of Straw and Straw Ash. Internal report in Danish, Biomass Ash Characterisation Project, dk-TEKNIK, Soeborg, Denmark Winegartner,E.C. and Rhodes,B.T. (1975). “An Empirical Study of the Relation of Chemical Properties to Ash Fusion Temperatures” Journal Eng. Power (97) Appendix D:

Ash Fusion and Deposit Formation at Straw Fired Boilers. Paper presented at the Chemical Engineering Conference, Kona, Hawaii (1997).

D ASH FUSION AND DEPOSIT FORMATION AT STRAW FIRED BOILERS

Lone A. Hansen1. Flemming J. Frandsen1, Henning S. Sorensen2, Per Rosenberg 2, Klaus Hjuler3 and Kim Dam-Johansen1 department of Chemical Engineering The Technical University of Denmark DK-2800 Lyngby, Denmark

2Geological Survey of Denmark and Greenland (GEUS) Thoravej 8, DK-2400 Kbh. NV, Denmark

3dk-TEKNIK Energy & Environment Gladsaxe Moellevej 15, DK-2860 Soeborg, Denmark

Keywords: Straw combustion, ash melting, fusion, STA, HTLM, deposit formation.

ABSTRACT

Straw fired boilers used for combined heat and power production generally experience severe problems with ash deposit formation on heat transfer surfaces. In this study fly ashes, bottom ashes and deposits collected at two straw fired boilers were subjected to advanced analyses for determination of ash chemistry and fusion. The chemical composition of the ashes was determined by CCSEM and detailed characterisation of ash fusion was determined by two new techniques based on high temperature light microscopy (HTLM) and simultaneous thermal analysis (STA), respectively. Results from the two techniques were compared and generally found to agree within 10% melt, except for fusion onsets for salt rich ashes. Results were also compared to standard ash fusion test results, revealing a poor correspondence to IDT for fly ashes, as initial melting was detected between 50 and 300°C below IDT. In addition, all fly ashes showed substantial melting, as high as 70% melt, at the IDT. For the bottom ashes the agreement to the IDT was better, only 0 to 22% melt was found at the IDT. These differences are caused by the difference in chemical composition between fly and bottom ashes. The other characteristic temperatures (Softening, Hemispherical, and Fluid Temperature) corresponded to between 40 and 95 % melt in the ash as determined by the two new techniques. The fly ash fusion results were used for simple modeling of deposit formation and related to deposit formation rates as measured in the two boilers. The modeling results overestimated the deposition rates, which is probably due to shedding from the deposit probes. Interpretation of the measured ash fusion data and a critical discussion of the modeling efforts are provided.

INTRODUCTION

Much attention has been drawn towards the burning of biomass for power production in recent years. The Danish Government has agreed to a 20 % reduction in COz emissions from power production by the year 2005 with reference to year 1988. Since biomass is considered COz neutral, the substitution of coal with biomass as fuel for power production is one of the initiatives taken to meet this goal. In Denmark, straw is the most frequently used biomass type, since surplus/reserves of straw often appear. The use of straw for heat production in small scale furnaces, e.g. at individual farms, has been practiced for a number of years. However, power generation from biomass is a fairly new task.

Straw fired boilers have generally experienced serious problems with slagging and fouling [Miles et al., 1995; Michelsen et al., 1996; Stenholm et al., 1996; Jensen et al., 1997] primarily due to the large content of troublesome deposit initiating elements such as potassium and chlorine in the straw. The straw contains sufficient quantities of volatile fluxing alkali to significantly lower the fusion temperature of the ash, so that it partly melts during combustion and increases the tendency of the ash to stick to heat transfer surfaces. Alternatively, the potassium and chlorine species may vaporize and subsequently condense on boiler tubes and refractory surfaces, creating sticky surfaces that accelerate deposit build-up and/or work as ‘glue’, thereby increasing deposit strength.

This paper presents results from two new techniques for characterisation (/quantification) of biomass ash fusion and relates the obtained results to compositional CCSEM data and measurements of deposit formation rates as found in two biomass fired boilers. The aim of this work has been to gain a more detailed understanding of the chemistry and fusion of fly and bottom ashes in biomass fired boilers and the mechanisms that lead to deposit formation.

HASLEV AND SLAGELSE COMBINED HEAT AND POWER PLANTS

Deposition measurements have been carried out at two straw fired boilers, the 23 MW4 Haslev Combined Heat and Power Plant (CHP) equipped with four cigar burners and the 31 MWft stoker fired Slagelse Combined Heat and Power Plant. Details about the measurements are reported by Haslev CHP Slagelse CHP

Straw

Figure 1: Schematic of Haslev and Slagelse Combined Heat and Power Plants. Deposit Sampling position are indicated by crosses.

[Stenholm et al., 1996] and [Jensen et al., 1997]. The boilers and deposit sampling positions are shown schematically in Fig. 1. At the Haslev CHP, furnace temperatures of 720-820°C were measured, and deposit samples were collected at two locations: in the top of the furnace and at the entrance to the third pass, just before the superheaters. At the Slagelse CHP, furnace temperatures of 745-9 10°C were prevailing, and deposit measurements were carried out in the middle of the furnace and in the third pass, the probe being located between the primary and the secondary superheater (Fig. 1).

During the test period at Slagelse CHP, a number of different straws were burned in order to investigate the influence of fuel composition on the deposit formation rate. The chemical composition of these five types of straw and the single straw burned at Haslev CHP are listed in Tables 1 and 2. As seen in the tables the numbers showing the largest deviation are the ash contents (varying between 3.5 and 7.1 %), and the contents of Si02 (4.3 - 70 %), CaO (7.9 - 29 %), K20 (8 - 46 %), S03 (0.1 - 13 %), and Cl (0.06 - 0.86) in the ashes. In the SLA8 test, rape straw was burned, which explains the markedly different ash content and chemistry compared to the other tests, in which wheat and barley straw was used.

Samples of fly ash, bottom ash and deposits were collected during the test period, and the chemical composition of these are shown in Tables 3a and 3b. Since also the association of elements in the fly and bottom ashes is of crucial importance for the ash fusion, the chemical species present in the ashes were identified by means of Computer Controlled Scanning Electron Table 1: Composition of straws fired at Haslev and Slagelse CHP (modified from Stenholm et al., 1996)

Straw C1 O1 H1 S1 N1 Cl1 Ash2 Water2 Vol2 LHV

SLA1 wheat 45.9* 41.1* 6.1 0.09 0.5* 0.06 5.1 15.0 66.3 15.0

SLA3 barley 45.9* 41.1* 6.1 0.07 0.5* 0.32 4.0 11.6 67.6 15.5

SLAB wheat 45.0 41.5 5.9 0.12 0.3 0.86 6.3 13.2 64.7 14.8

SLA7 wheat 46.8 41.7 6.1 0.10 0.5 0.14 4.1 16.1 65.8 14.8

SLAB rape 45.5 39.8 5.9 0.40 0.6 0.33 7.1 14.0 64.8 14.5

HAS wheat 48.4 41.2 5.4 0.12 0.6 0.29 3.5 15.1 66.5 15.1

1 % (w/w) of dry matter;2 % (w/w), as received ; * no specific measurement, average of samples

Table 2: Chemical composition of fuel ashes from Haslev and Slagelse CHP [% (w/w)] (modified from Stenholm et al., 1996)

Si02 ai2o3 FG2O3 030 MgO Na20 K20 S03 P2O5

SLA1 70 0.81 0.42 11 1.4 0.37 8 1.5 0.98

SLA3 25 0.35 0.19 13 1.4 1.7 46 1.4 1.6

SLAB 52 0.14 0.15 7.9 1.4 0.17 35 2 2.2

SLA7 68 2.10 1.0 13 2.2 0.51 16 1.2 3.5

J3LA8 4.3 0.20 0.12 29 1.2 1.7 25 13 3.7

HAS 50 0.50 0.38 11 1.8 0.32 20 0.1 3.5

Table 3a: Chemical composition of fly ashes and bottom ashes from Haslev and Slagelse CHP [% (w/w)] (modified from Stenholm et al., 1996)

Si02 ai2o3 Fe203 CaO MgO Na20 k2o SO, P205 Cl

SLA1FA 27 0.38 0.66 4.7 0.82 0.7 32 8.7 2.0 17.0

SLA1BA 68 0.9 0.56 10 2 0.4 11 0.3 1.8 0.29

SLA3FA 2.7 0.07 0.24 0.94 0.17 0.16 57 9.4 0.49 20.0

SLA3BA 46 0.84 0.34 15 2.5 1.6 24 0.29 2.6 0.29

0.20 SLA6FA 4.6 0.40 1.7 0.27 0.77 61 7.8 1.5 33.0

11 2.2 21 SLA6BA 60 0.57 0.37 0.35 0.35 2.7 1.19

SLA7FA 14 0.35 0.57 3.6 0.68 1.1 47 8.9 2.3 25.0

SLA7BA 63 1.7 0.65 14 2.7 1.0 17 0.27 2.6 0.43

0.20 20 1.6 10.0 SLA8 FA 5.7 0.83 0.99 40 18 2.7

1 20 SLA8 BA 53 0.51 16 2.5 0.9 0.31 3.1 0.33

HASFA 35 0.77 1.4 8.2 1.6 0.46 26 5.7 2.9 10.2

HASBA 69 1.3 0.67 8.5 1.9 0.38 13 0.96 3.4 1.1 Table 3b: Chemical composition of deposits from Haslev and Slagelse CHP [% (w/w)] (modified from Stenholm et al., 1996)

Si Al Fe Ca Mg Na K S P Cl

SH-outer 2.1 0.1 0.8 2.1 0.3 0.8 44.8 2.6 0.2 20.0

SH-inner 5.1 0.3 32.9 4.6 0.7 0.5 28.5 11.4 1.8 15.1

Fur-outer 14.5 0.2 0.3 12.2 1.2 1.3 29.9 1.3 0.9 1.7

Fur-inner 9.3 0.1 3.5 6.9 1.0 1.2 31.6 1.0 0.8 14.0

SH-outer 6.5 0.1 0.5 6.3 0.8 0.3 34.0 6.0 1.3 20.0

Fur-outer 18.7 0.9 1.4 7.2 1.1 0.3 20.8 2.9 1.1 10.0

Fur-inner 20.0 0.9 9.1 9.9 1.4 0.5 35.4 3.9 1.4 10.9

Microscopy, CCSEM. The mineral categories relevant for biomass ashes are developed and described by [Sorensen, 1997]. From the data, it was seen that fly ashes were characterised by a high content of KC1, a minor content (below 5% (w/w)) of other potassium and calcium salts (chlorides and sulphates), and a relatively low content of potassium and calcium silicates. The bottom ashes were characterised by a very high content (> 90 %) of silicon containing species, including quartz, potassium silicates, calcium silicates and small quantities of alumina silicates. Most of the deposits examined were found to contain large quantities of KC1 and varying quantities of potassium and calcium silicates, the ratio between silicates and KC1 varying for different deposit locations in the boiler. The compositional differences between the various ashes and deposits will be discussed later.

DETERMINATION OF ASH SOFTENING

The fusion of ashes is generally of great importance for the ability of these ashes to form problematic deposits on heat transfer surfaces in boilers [Moza and Austin, 1981; Srinivasachar et al., 1990; Walsh et al., 1990; Benson et al., 1993; Richards et al., 1993]. In this study, the fusion of ashes and deposits collected was investigated in order to find a relation between ash fusibility and deposit formation observed. The fusion of ashes and deposits was quantified experimentally using two different approaches: High Temperature Light Microscopy and Simultaneous Thermal Analysis.

Experimental Techniques Using High Temperature Light Microscopy, HTLM, the quantification of melt in ash is based on the change in light transparency of the ash as it is melted. Shortly described the method works as follows: A sample of ash is placed in the high temperature light microscope. During heat up, the area covered by the sample is quantified at short intervals using advanced image analysis. As part of the ash is melted, this part becomes transparent, thereby not contributing to the determined area any more. In this way, the area at any given temperature, T, divided by the area at the initial temperature, T0, A(T)/A(T0), is used as an estimate of the solid fraction of the ash at the temperature T. The method is described in detail by [Hjuler, 1997].

Using Simultaneous Thermal Analysis, STA, for fusion determination implies continuous measurement of sample weight (Thermogravimetric Analysis, TGA) and sample temperature (Differential Scanning Calorimetry, DSC) during heat up of the ash. The weight measurement revealsany mass changes taking place in the sample. By comparing the sample temperature to the temperature of an inert reference material, any heat producing or heat consuming processes (chemical or physical) occurring in the sample is detected, and the energy involved subsequently quantified. On the resultant STA curves, melting is detected as an endothermic process involving no change in mass. The conversion of STA curves to melting curves (showing the fraction of melt in the ash as a function of temperature) can be carried out as follows: Initially, the energies related to other processes than melting (typically evaporation) have to be subtracted from the DSC curve to obtain a DSC curveshowing only melting energies. The area of a given melting peak in the resultant DSC curve, corresponds to an absolute quantity of energy used for melting. The position of the peak (onset and peak temperature) gives an indication of the identity of the melting substance(-s), which means that a reasonable estimation of the relevant melting enthalpy can be given. Based on these two numbers, the mass of material melted in the given temperature interval can be calculated. By relating this melted mass to the total mass of ash analysed, the mass fraction of ash melted in the given temperature interval is obtained. If the identification of discrete melting peaks is not possible, the quantification of melt formation is carried out based on a direct comparison of areas below the DSC curve, i.e. energies used for melting in different temperature ranges. A complete description of the method is given by [Hansen, 1997] and [Hansen et al., 1997].

Comparison of Techniques To illustrate the connection 0.9 - • Fly ash between the two techniques 0.8 -- applied for ash fusion 0.7 -- quantification, results from 0.6 -- respectively HTLM and STA are 0.5 HTLM presented in Fig. 2, showing 0.4 fusion of a fly ash (from exp. 3 0.3 --

at Slagelse CHP), a bottom ash 0.2 - (from exp. 8 at Slagelse) and a 0.1 -- deposit (SH outer deposit from Haslev). For the fly ash and the deposit the figure shows that the 0.9-- Bottom ash two techniques do not agree on 0.8 melting onset: the HTLM detects 0.7 -- the first melting at temperatures 0.6 50-150°C lower than the STA. 0.5 •• 0.4 The initial melt formation 0.3 -- detected by the HTLM method is 0.2 believed, though, to be caused by HTLM 0.1 -- physical rearrangement of the ash, and not necessarily melt formation. At the temperature, where the STA detects the first 0.9 -- Outer superheater deposit melting to occur, the HTLM 0.8 -- detects 10-15% melt. These are 0.6 -- typical numbers/figures. HTLM Generally, the methods agree 0.4 -- nicely at the temperatures for which significant melting

occurs: both the slope of the 0.1 -- curve and the amount of melt obtained agree. At higher 0 200 400 600 800 1000 1200 1400 temperatures the STA Temperature [°C] technique tend to detect slightly higher fractions of melt than Fig. 2: Comparison of results from HTLM and STA the HTLM technique, which could be due to some of the ash species (e.g. some high melting silicates) forming melts of dark colour, which may not be light transparent.

For the bottom ash, the agreement on the melting onset is better, and the quantity of melt detected by the HTLM at the onset temperature determined by the STA is only 0-5%. For half of the bottom ashes investigated, the two techniques agree well at higher temperatures (as for the ash presented), whereas for the others, the STA detects significantly higher degrees of melt than the HTLM. As explained above this may be due to some (silicate) species not being light transparent even when melted. Another feature illustrated is the occasional decrease in melt fraction observed with the HTLM method. This often appear after significant melt formation, and is due to physical rearrangement of the solid ash in the liquid phase.

In conclusion, the two techniques are found to agree qualitatively, and for about half of the ashes, results from the two methods do not deviate more than 10 % (absolute) at temperatures between 650°C and 1100°C. Taking into account experimental uncertainty and the difference in the principles applied, the two techniques are judged to agree reasonably well.

RESULTS

In this section, the ash fusion is evaluated based on the chemical composition and the origin of the ashes. In order to limit the number of fusion curves, only curves produced by the STA method will be shown, and comments will be given on any different trends found from the HTLM method.

Straw, ashes and deposits from Haslev The CCSEM compositional data for the ashes is illustrated more clearly in a triangular diagram of 1) quartz and alumina silicates, 2) K and Ca silicates and 3) KC1 (Fig. 3). The three end- members in the diagram were chosen to represent: 1) relatively refractory quartz and alumina silicates, 2) more easily fusible potassium and calcium dominated silicates (probably formed by reaction of evaporated potassium and calcium with straw derived silica or quartz grains), and 3) low-melting KC1 formed during the combustion process [Sorensen, 1997]. Added up the three end-members account for 67 to 91% of the samples presented.

The compositional differences between the bottom ash, the fly ash and the deposits are evident in the triangular diagram (Fig. 3), with the deposits located close to the KC1 apex, fly ash occupying an intermediate position and bottom ash located close to the ‘silicate line’. It is seen that the compositions of the two outer superheater deposits are very similar, as might be expected since both are bulk sintered deposits collected at each end of the same deposit probe. Similarly, the two outer furnace deposits are quite alike. The inner deposit from the furnace probe is depleted in K and Ca silicates compared to the two outer parts, indicating that the initial Fig. 3: CCSEM compositional data for samples from furnace deposit is formed primarily from Haslev CHP. BA=bottom ash; FA=fly ash; condensation of KC1, whereas for the bulk F=fumace deposit; SH=superheater deposit sintered deposit part, which contains a larger fraction of silicates, impaction of fly ash particles has played a more dominant role. For the bulk sintered part of the super heater deposit, the KC1 content is higher than for the bulk sintered furnace deposit, indicating that impaction of silicate rich fly ash particles is a more important deposit forming mechanism in the furnace compared to the super heater area. An alternative explanation for the higher KC1 content in the outer super heater deposit may be that the lower temperatures in the convective pass may have caused some KC1 to have condensated heterogeneously on the fly ash that impacts.

The fusion of fly ash, bottom ash and deposits collected at Haslev CHP is shown in Fig. 4 as determined by the STA method. It is seen that for all ashes, the melting occurs in two temperature ranges: 1) between 630°C and 750° and Bottom ash 2) above 1000°C. The low temperature range is believed 400 600 800 1000 1200 1400 to be caused by melting of an Temperature [°C] eutectic mixture of potassium Fig. 4: Melting curves for samples from Haslev CHP. and calcium salts (chlorides and sulphates, the exact composition of which is not known), while the high temperature range is believed to include melting of the various silicates. The fusion curves clearly reflect the compositional differences between the ashes: the bulk sintered super heater deposit and the initial furnace deposit containing the largest quantities of KC1, form larger quantities of melt in the low temperature range than the bulk sintered furnace deposit. Likewise, the fly ash shows lower melt fractions than the deposits, but higher melt fractions than the bottom ash, which does not melt at all in the low temperature range. This means that the melting behaviour of the ashes correlate with the position in the ternary diagram, i.e. moving (in composition) away from the KC1 apex results in smaller melt fractions obtained in the low temperature range.

Straw, ashes and deposits from Slagelse 3

For fly ash, bottom ash and the three I K-Ca silicates) deposits from the Slagelse 3 experiment, \F, outer the chemical composition as determined by CCSEM is illustrated in a triangular F, outer diagram (Fig. 5). Also here is seen a dominance of KC1 in the deposits compared to the bottom ash, as was the case for the Haslev samples. However, the fly ash from Slagelse 3 does not extinguish itself from the deposits in respect of KC1 content, as did the Haslev fly ash. It is seen that the compositions of outer furnace and superheater deposits are quite alike, as these are two sets of bulk sintered deposits, from the first and second half of the superheater and the furnace deposit probe, Fig. 5: CCSEM compositional data for samples from respectively. For the deposits collected in Slagelse CHP. BA=bottom ash; FA=fly ash; F= the furnace, the initial layer is seen to furnace deposit; SH=superheater deposit contain a moderate quantity of KC1, whereas the bulk sintered layer contains only a small fraction (between 2.5 and 18.1 %). For the deposit collected in the superheater area, both inner and outer layers contain large quantities of KC1, and it actually seems that the KC1 content of one of the bulk sintered deposit layers is a little higher than that of the initial layer. This is quite surprising, but is probably a consequence of the high KC1 content of the impacting fly ash particles. These findings thus support the trends seen for the Haslev samples: The content of silicates is higher in the bulk sintered layer than in the initial layer of any deposit, meaning that condensation and thermophoresis of KC1 is of highest importance for formation of the initial deposit layer, whereas impaction of fly ash particles is more important for formation of the bulk sintered deposit layer.

The fusion curvesfor fly ash, bottom ash and deposits collected at Slagelse 0.9-- SH, outer 3 is shown in Fig. 6 as determined by 0.8 the STA method. Again, the melting is = 0.7 Inner found primarily to occur in a low- F, inner temperature range between 600°C and 750° due to melting of eutectic salt £ 0.3-• mixtures and a high-temperature F, outer range above 900°C covering melting Bottom ash of silicates. These fusion curves also clearly reflect the compositional Temperature [°C] differences between the ashes: The „. „ , bottom ash that contain no KC1 shows ^ £ Melting curves for samples from experiment no melting in the low temperature at a°e se range, whereas the bulk sintered furnace deposit containing 3% KC1 forms a small amount of melt below 900°C. The inner furnace deposit, which contain moderate quantities of KC1 shows an intermediate fraction of melt at 800°C, and the fly ash and both superheater deposit layers, which have the largest KC1 contents consequently show the largest melt fractions below 800°C. As for the Haslev samples, the position in the triangle thus illustrates the relative fusibility of the ashes.

Slagelse fly and bottom ashes The CCSEM compositional data for the | K-Ca silicate | Slagelse fly and bottom ashes are illustrated in a triangular diagram (Fig. 7) [Hi BAS, identical to the ones shown for the Haslev I 22.01 BAG and Slagelse 3 experiments. The difference [Ho] BAS, in mineral composition between bottom R551BA7I ashes and fly ashes is obvious: the bottom ashes are dominated by K and Ca silicates, quartz and alumina silicates, whereas the PHI BA1 A "\ fly ashes are dominated by KC1 and contain varying fractions of silicates. KC1 and \ /\FA2i\F>2vf73.8 silicate often appear as composite grains which are found mainly in two different V -Vss Vao *0 88 Quartz + I 51.6| 64.3 1^ FA6 KCI forms: they are either formed by aluminosilicates ------[7551 condensation of KC1 on silicate particles, or _ .. . . _ _ , they represent KC1 and silicate particles 7; CCS™ compos,ttonal dam for fly and being located so closely that they cannot be BA-bottom ash; distinguished by the SEM. Observations in FA-fly a* : BAX=BA 6°™ =*P=nm=nt X the SEM backscatter images underline that the presence of condensated KC1 on silicate particles is a common feature in the biomass fly ashes.

In the Slagelse 8 experiment rape is used as fuel instead of wheat or barley as in the other experiments. The difference in ash composition is especially evident for the fly ash, which contains large amounts of K2S04 and Ca rich categories in accordance with the Ca and S rich character of rape relative to wheat and barley. This is the reason that only 11% of the fly ash composition is represented in the diagram. For the remaining ashes, the strong compositional difference between bottom ashes and fly ashes is evident: the bottom ashes being located at or close to the silicate line, whereas the fly ashes stretch from the KC1 apex towards a composition of approximately 85% ‘quartz and alumino silicates’ and 15% ‘K and Ca silicates’. This ratio (of ‘quartz and alumino silicates’ to ‘K and Ca silicates’) is higher for the fly ashes than for any of the bottom ashes.

Fusion curves for the five fly ashes as determined by STA are shown in Fig. 8. 0.8 -- It is seen that all fly ashes show distinct c 0.7 melt formation (> 50%) in the 0.6 2 0.5 - temperature range from 600 to 720°C, 0.4 except for the fly ash from experiment E 0.3 - 0.2 8. The fly ashes from experiment 3 and 0.1 •• 6 reach complete melting first (just below 1200°C). The fly ashes from experiment 7 and 1 are almost, but not Temperature [°d completely melted at 1250°C, whereas Fig. 8: Melting curves for fly ashes from Slagelse CHP the fly ash from experiment 8 showed significantly lower fusion at 1250°C.

Fusion curves for the five bottom ashes are shown in Fig. 9 as determined by HTLM. All bottom ashes start melting at around 800°C. From 900°C to 1000°C the melting process is very slow, and above 1000°C significant melting occurs in 0.2 bottom ashes from experiment 3,6 and 8. Bottom ashes from experiment 1 and 7 do not show any sign of 200 400 600 800 1000 1200 1400 additional melting above 1000°C. Temperature [°C] Fig. 9: Melting curves for bottom ashes from Slagelse CHP

To test the usefulness of the diagram in Fig. 7 as a basis for classification into more or less fusible ashes, melt fractions extracted from HTLM melting curves are assigned to the data points. The numbers in square frames beside the data points represent melt fractions at 880°C. A consistent trend is evident for the bottom ashes towards lower values of ‘melt fraction at 880°C’ with an increasing content of ‘quartz and alumino silicates’. This is in accordance with the more refractory nature of quartz and most alumino silicates compared to K and Ca silicates. For the fly ashes a trend is seen of lower melt fractions for a given temperature (data labels show melt fractions at 800°C, at which temperature the salts are assumed to be melted) with an increase in silicate content relative to KC1 content. This is in accordance with the low melting point of KC1 compared to most silicates. In conclusion, it is found that the proposed triangular diagram may be useful to classify the relative fusibility of biomass ashes, and the figure furthermore illustrates the logical relation between the chemical composition of the ashes and their fusion characteristics.

Comparison to standard AFT For all fly and bottom ashes, the fusion was additionally characterised by the standard ash fusion test (ISO540, 1981). This test implies heating up of a cube of ash and noting the temperatures corresponding to 1) rounding of the cube comers, at the Initial Deformation Temperature, IDT, 2) the shape change from cubic to hemispherical, at the Hemispherical Temperature, HT, and 3) flowing of the ash, at the Fluid Temperature, FT.

Table 4: Relation between AFT temperatures and melt fractions determined by HTLM and STA, respectively

IDT HT FT IDT AT HTLM STA HT HTLM STA FT HTLM STA [°C] [°C] [% melt] [% melt] [°C] [%melt] [% melt] t°C] [% melt] [% melt] SLA1FA 750 -106 29.9 51.6 930 46.6 54.9 1170 77.4 65.1

SLA1BA 910 79 13.0 0 1190 - 53.6 1430 - - SLA3FA 650 -50 51.5 70.2 660 52.7 70.6 750 67.2 71.3

SLA3BA 870 -45 22.5 2.8 970 (15.6) 13.6 1090 81.1 41.7

SLA6FA 710 -60 46.1 54.0 740 55.2 68.5 830 74.2 82.3

-68 1110 SLA6BA 830 10.5 7.4 56.5 58.2 1440 - -

SLA7FA 710 -65 56.5 54.4 770 66.4 64.3 1010 79.1 73.1

SLA7BA 900 -95 16.8 7.8 1150 22.6 82.0 1410 - -

SLA8 FA 790 -270 37.5 53.8 800 37.7 53.8 1490 - -

0 8.0 0.1 SLA8 BA 830 1150 49.9 72.7 1390 - -

HASFA 950 -298 53.0 28.6 1110 72.2 60.5 1380 - -

HASBA 970 -53 22.4 3.6 1240 - 95.0 1360 - - Ave. FA 45.8 ± 52.1 ± 55.1 ± 62.1 ± 74.5 ± 73.2 ± BA 10.2 13.4 12.7 6.9 5.3 7.1 15.5 ± 3.6 ± 41.3 ± 62.5 ± - - 6.1 3.4 17.2 28.4

Table 4 shows the melting onset as determined by the STA and the fractions of melt present in the various ashes at the IDT, HT and FT as determined by HTLM and STA, respectively. It is seen that for the fly ashes, the melt formation starts at temperatures well below the IDT, ranging from 298°C below for the fly ash from Haslev to 50°C below for the fly ash from experiment 3 at Slagelse. For the bottom ashes, melt formation is detected between 0 and 95°C below the IDT, excepting the bottom ash from experiment 1 at Slagelse, for which melt is detected 79°C above the IDT. For the fly ashes, substantial quantities of melt have been formed at the IDT, averages of 45.8 % and 49.4% melt are found by the HTLM and the STA, respectively. However, the melt fractions detected for the bottom ashes at the IDT is much lower, respectively 15.5 % and 2.9% on average. The reason for the STA to find no melt at the IDT for two silicate rich bottom ashes might be that the rounding of the comers has been due to sintering of the ash in stead of melting. At the Hemispherical Temperatures melt fractions are generally between 35 and 70% melt, the bottom ashes generally showing smaller melt fractions than the fly ashes, and the STA generally detecting slightly higher melt fractions than the HTLM. At the fluid temperature melt fractions measured generally lie between 60 and 80%, the averages found to 74.5 and 72.6% melt.

These findings support previous criticism [Gerald et al., 1981; Huggins et al., 1981; Coin et al., 1996; Wall et al., 1996] stating that the IDT does not indicate the melting onset. The IDT has been known to indicate the temperature, at which ash particles in an operating furnace have cooled sufficiently to have only a slight tendency to stick together [Bott, 1991], but with melt fractions of 30 to 60% the stickiness of the ash may be somewhat higher than only ‘slight’, meaning that keeping furnace exit gas temperatures below the DDT does not guarantee avoidance of deposit problems in the boiler convective pass.

SIMPLE DEPOSIT MODELING

The quantity of melt in an fly ash is known to increase its tendency for adhering to surfaces after impaction [Srinivasachar et al., 1990; Walsh et al., 1990; Benson et al., 1993; Richards et al., 1993]. In this study simple calculations have been made to estimate the deposit formation rate as it would occur if only condensation and inertial impaction of the collected fly ashes were important. This means that thermophoresis has not been taken into consideration. In the model, the build up of deposit is assumed to be initiated by condensation of gaseous KC1. After a layer of 20 pm has formed, the condensation process is assumed to end/stop, and the deposition is continued by impaction of ash particles.

The ash deposit formation rate, N^p (g/m2/h), caused by inertial impaction has been calculated as [Frandsen, 1997]:

N. U imp 8 FA imp stick (i) where ug (m/s) is the bulk gas velocity, cFA (g/m3) is the mass concentration of fly ash particles in the flue gas, f^, is the weight fraction of fly ash particles that impacts the tube, and f^k is the fraction of those particles impacting the tube that sticks to it. In the equation, the product ug-CpA represents the total flux (g/(m2h)) of fly ash particles in the gas. (For calculating the deposition flux, the total flux is corrected for the facts that not all particles impact the tube (multiplication by fimp), and not all impacting particles adheres to the tube (multiplication by fstick ).) The mass concentration of fly ash particles in the flue gas is estimated based on the mass flow of fly ash compared to the mass flow of fuel ash introduced into the furnace, and the prevailing air excess number. The velocity of the gases have been measured to 3.1-3.6 m/s in the furnace and 9.1-10.3 m/s in the convective pass.

Estimation of the impaction coefficient, fimp, may be carried out based on the Stokes number, Stk, for potential flow around a tubular cylinder [Wessel and Righi, 1988], using the following relationship [Israel and Rosner, 1983]:

fimp = D +be-l-ce-2 +de~2y' (2) where e = Stk - a, and the coefficients applied are those given by [Israel and Rosner, 1983].

The sticking coefficient of the fly ash particles has been estimated based on the presented STA ash fusion curves. The sticking coefficient was estimated to simply equal the fraction of melt in the fly ash at the measured gas temperature, after having corrected the melt fraction for the fraction of melted ash that has been evaporated at the gas temperature. The sticking coefficient of the tube is estimated to zero, since melt only occurs at temperatures above 610°C, and the deposit surface structure did not show any sign of melting. Based on these figures the ash formation rates presented in Fig. 10 were found.

As seen in the figure, the model overestimates the deposition rate for all experiments. The ratio between calculated rates and measured rates varies between 1.4 and 11.2 except for one extreme case, where the model overestimates the deposition rate by a factor of 47.2. The average ratio is 10.3 including all experiments, and 6.6 excluding experiment SLA3, OH. f c o Regarding the deposition 1 measurements, it is known that O E "U .O) deposit typically detaches the "O probe, both during experiment, 2 but particularly during probe til3 withdrawal. This means that 8 the mass of ‘loose’ deposit measured is probably lower 200 400 600 800 than the ‘actual’ value. On the Calculated deposition rate [g/m2h] other hand there is no doubt, that the model in addition Fig. 10: Calculated vs. measured deposition rates overestimates the deposition rates. Several reasons for this can be given: - the applied model only accounts for deposit build up, that is, no deposit removal mechanisms are taken into account. In reality, erosion of the deposit due to the impaction of dry fly ash particles may be severe. This causes unrealistically high deposition rates. - taking the melt fraction directly as the sticking probability is of course a rough simplification. On its route through the convective pass of the boiler, the fly ash particles may be enriched in condensible alkali species. Since the condensible species lower the melting point, probably the fly ash collected in the particulate removal device and thus the one that has been used as basis for calculations, show larger melt fractions at low temperatures (< 800°C) than the ash present in the flue gas passing the deposit probes. This causes that the sticking probability used is too high. - the rate of condensation in the convective pass is probably overestimated, since the concentration of gaseous KC1 has been estimated as if all sampled aerosols were present in the gas phase. Taking the prevailing temperatures into account, a large fraction of the KC1 has probably already homogeneously condensed before the gas enters the superheater probe location. The KC1 is therefore probably present as aerosols which will be transported to the probe surfaces by thermophoresis instead of gas diffusion. The rate of condensation may not equal the rate of thermophoresis, and the rate of condensation may therefore be overestimated. - finally, the modeling of the condensation phenomena is greatly simplified. A layer of equal thickness present for all experiment conditions is not credible, and since the condensation process plays a large role in purely straw fired boilers, this part of the model should be refined to obtain an overall model in better agreement with the measured data.

In Fig. 10, the measured deposition rates are shown as a function of the calculated values. There is a clear correlation between measured and calculated deposition rates, when excluding two data points, the one from SLA8,SH (206,120) and the one from SLA3,SH (783, 16.6). The reason for these data points to distinguish them selves from the others is not quite clear yet. For the rest of the data points, the obtained linear correlation means that even though the model does not quantitatively agree with the measurements, there is a clear qualitatively agreement, indicating that the model may be useful for ranking the deposit formation propensities of one fly ash compared to an other from the same kind of boiler. Taking the great simplicity of the model into account, this is not a bad result.

CONCLUSIONS

The fusion of fly ashes, bottom ashes and deposits collected in the furnace and the convective pass of a straw fired boiler have been quantified. CCSEM of the ashes revealed that the fly ashes and deposits contained high quantities of salts, especially KC1, and varying quantities of potassium and calcium silicates, whereas the bottom ashes consisted of a mixture of quartz, potassium, calcium and alumina silicates. The compositional differences of the ashes were clearly reflected in the fusion, as the salt containing fly ashes and deposits showed significant melting in the temperature range from 6 00°C to 750°C, whereas the melt formation in the bottom ashes primarily occurred at temperatures between 1000°C and 1200°C. Furthermore, it was generally found that increasing contents of salts in the ashes implied a larger fraction of the ash fusion to occur below 800°C.

Comparison between results from the standard ash fusion test and the newly developed techniques revealed that melt formation was initiated at temperatures much lower than the IDT, the temperature difference being 142°C below for fly ashes and 30°C below for bottom ashes. Melt quantities of approximately 45-50% were formed in the fly ashes at the initial deformation temperature, which means that probably the tendency for the ashes to form deposits can not be neglected at the IDT. For the bottom ashes the melt fraction at the DDT was lower, around zero to fifteen percent The agreement between the standard ash fusion test and the new techniques is thus worse for salt rich compared to silicate rich ashes.

The chemical composition of the deposits as determined by CCSEM indicated that the formation of the initial part of the deposits was dominated by condensation of KC1, whereas inertial impaction of fly ash particles played a higher role in formation of the bulk sintered layer on top. Furthermore inertial impaction was indicated to play a higher role for the deposits in the furnace compared to deposits in the convective pass. A very simple model set up to estimate the deposition rates was found not to agree quantitatively with the deposit formation rates measured, since deposit rates were overestimated by a factor of 1.4 to 11.2. Never the less, a linear correlation between calculated and measured deposition rates was found, meaning that the simple model was capable of ranking deposition rates from the different fly ashes.

ACKNOWLEDGEMENT

This work has been financially supported mainly by the Danish Energy Research Programme (EFP), and partly by the Combustion and Harmful Emission Control (CHEC) research programme, ELSAM (the Jutland-Funen Electricity Consortium) and ELKRAFT. Peter Arendt Jensen and Michael Stenholm are acknowledged for giving the project participantsaccess to the collected ash samples.

REFERENCES

Benson, S.A., Jones, M.L., and Harb, J.N. (1993). “Ash Formation and Deposition”. In L. D. Smoot (Eds.), Fundamentalsof Coal Combustion for Clean and Efficient Use New York: Elsevier. Bott, (1991). “The Assessment of Fouling and Slagging Propensity in Combustion Systems”. In S. A. Benson (Eds.), Inorganic Transformations and Ash Depositionduring Combustion New York: ASME Coin, C.D.A., Kahraman, H., and Peifenstein, A.P. (1996). “An Improved Ash Fusion Test”. In L.L. Baxter and R. DeSolIar (Eds.) Proceedings of the Engineering Foundation Conference, 1995, New York and London: Plenum Press Couch,G. (1994). Understanding slagging and fouling inpf combustion. IEA Coal Research, Report No. IEACR/72. Frandsen, F.J. (1997). Estimation of Ash Deposition Fluxes in Utility Boilers. Internal Report, Department of Chemical Engineering, Technical University of Denmark. Gerald,P.H., Huggins,F.E., and Dunmyre,G.R. (1981). “Investigation of the High-Temperature Behaviour of Coal Ash in Reducing and Oxidizing Atmospheres”. Fuel (60) 585 Hansen, L.A., Frandsen, F.J. and Dam-Johansen, K. (1997) “Ash Fusion Quantification by Means of Thermal Analysis”. These conference proceedings. HjuIer,K. (1997) “Ash Fusibility Detection Using Image Analysis”. These conference proceedings. Huggins,F.E., Deborah,A.K., and Gerald,P.H. (1981). “Correlation between Ash-Fusion Temperatures and Ternary Equilibrium Phase Diagrams”. Fuel (60) 577 ISO 540 (1981). “Determination of Fusibility of Ash”. International Standard Organisation Jensen, P.A., Stenholm, M., and Hald, P. (1997) “Deposition Investigation in Straw Fired Boilers”, accepted for publication in Energy and Fuels. Michelsen, H.P., Larsen, O.H., Frandsen, F., and Dam-Johansen, K. (1996). “Deposition and High Temperature Corrosion in a 10 MW Straw Fired Boiler”. Presented at the Engineering Foundation Conference, April 28 - May 3, Snowbird, Utah. Miles, T.R, Miles, T.R Jr., Baxter, L.L., Biyers, R.W., Jenkins, B.M., Oden, L.L. (1994). “Alkali Deposits Found in Biomass Power Plants. A Preliminary Investigation of Their Extent and Nature”. National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, Colorado. Moza, A.K. and Austin, L.G. (1981). “Studies on Slag Deposit Formation in Pulverized Coal Combustors. 1. Results on the wetting and adherence of synthetic coal ash drops on steel”. Fuel (60) Richards,G.H., Slater,P.N., and Harb,J.N. (1993). “Simulation of Ash Deposit Growth in a Pulverised Coal- Fired Pilot Scale Reactor ”. Energy & Fuels (7) 774-781 Srinivasachar,S., Helble,J.J., Katz,C.B., and Boni,A.A. (1990). "Transformations and Stickiness of minerals during pulverised coal combustion". In R W. Bryers and K. Vorres (Eds.) Proc. of the Engineering Foundation Conference on Mineral Matter and Ash Deposition from Coal. New York:Engineering Foundation Stenholm, M., Jensen, P.A., and Hald, P. (1996). “Fuel and Firing Characteristics of Biomass - Combustion Trials”. An EFP-93 Project (in Danish), Journal No. 1323/93-0015, The Danish Energy Research Programme. Sorensen, H.S. (1997). “Computer Controlled Scanning Electron Microscopy (CCSEM) of Straw Ash”. These conference proceedings. Wall, T.F., Creelman, RA., Gupta, RP., Coin, C., and Lowe, A. (1996). “Coal Ash Fusion Temperatures: New Characterisation Techniques and Associations with Phase Equilibria”. In L.L. Baxter and R DeSollar, Proceedings of the Engineering Foundation Conference on Applications of Advanced Technology to Ash-Related Problems in Boilers. New York: Plenum Press Walsh,P.M., Sayre,A.N., Loehden,D.O., Monroe,L.S., Beer,J.M., and Sarofim,A.F. (1990). "Deposition of Bituminous Coal Ash on an Isolated Heat Exchanger Tube: Effects of Coal Properties on Deposit Growth" Prog. Energy Comb. ScL (16) 327-346 Appendix E:

SEM-lmages and SEM-Analysis of Fuel, Ash, and Deposits from Straw and Straw-Coal Combustion.

E SEM Images and SEM Analysis of Fuel,

Ash and Deposits from Straw - and

Straw-Coal Combustion

Henning Sund S0rensen

Geological Survey of Denmark and Greenland (GEUS)

O

GEUS 1. Introduction As a measure to meet the demand for reduction of C02 emissions the Danish Government has commited the Danish Power Companies to burn between 1 and 1.2 million tons of straw per year by the year 2000. Therefore a considerable research effort is put into solving problems that may arise by shifting from the presently dominating coal fuel to biomass. Biomass in general has a quite different chemical composition than coal with respect to the contents of both organic and inorganic constituents as well as heating value.

The main surplus of biomass in Denmark is straw and therefore the efforts are concentrated on this biomass fuel. One of the main problems involved in straw combustion is related to the behaviour and fate of the ash forming inorganic species in the fuel. These elements can end up in fly ash, bottom ash and may be incorporated in various types of deposits on boiler surfaces. The presence of deposits can reduce the heat transfer and increase the temperature level in the boiler. Other significant problems involve corrosion beneath deposits and large blocks falling down due to gravity forces thereby damaging the boiler.

Clearly the fate of the inorganic constituents during combustion and deposition depends on their mode of occurrence in the straw. The inorganic species in straw can occur either disseminated or ionically bound in the organic structure, be located in inorganic straw constituents or can be present as constituents of incorporated terrigeneuos soil particles. Additional factors, such as combustion temperature and boiler design, also influence the fate of inorganic material during combustion.

Knowledge of straw ash composition and melting behaviour is important to be able to predict, at least to some extent, the fate of the inorganic elements in terms of practical operation of boilers. The aim is to be able to predict and possibly avoid formation of troublesome deposits and potential corrosion. Additionally, knowledge of straw ash compositions and morphology is useful for possible utilization or disposal of the ash. An extensive study of straw ash has been performed at the Geological Survey of Denmark and Greenland (GEUS) as part of the EFP-95 project on Characterization of Biomass Ashes. In addition the study encompassed wheat straw and wheat grains as representatives of different fuels as well as various deposits to study how ash can contribute to deposit formation. The analyses at GEUS were performed by SEM and CCSEM on a Philips XL40 SEM equipped with a Voyager 2.7 EDX analysis software package. The main incentive was to develop a technique for straw ash characterization by Computer Controlled Scanning Electron Microscopy (CCSEM) based on the well-established technique for minerals in coal (Laursen, 1997a).

-2- This report firstly presents representative SEM-images of fuels, ashes and deposits that were obtained during the study for future reference. Secondly a presentation and discussion of CCSEM data is provided. A CCSEM study of straw ashes from the Haslev and Slagelse Combined Heat and Power Plants (CHP) is reported in Sorensen (1997) (Appendix A) and Hansen et al. (1997) (Appendix D). In Sorensen (1997) the modification of the CCSEM technique to be suited for straw ash characterization is discussed in detail, and comparison between CCSEM results and ash fusion as determined by STA and HTLM is discussed in Hansen et al. (1997). Therefore these subjects will not be treated in detail here.

2. Sample material The samples that are presented in this report were obtained from different sources. Bottom ash and fly ash were obtained from the straw-fired Haslev and Slagelse CHP’s. Deposits from straw combustion were obtained from two experiments: Slagelse 3 and Haslev 1, burning barley and wheat, respectively.

Full scale co-combustion samples were obtained from the pf-fired MKS1 boiler at Studstrupvaerket. Fuel, char, bottom ash and fly ash were analyzed by CCSEM and deposits by SEMPC. The results are treated together with fusion and sintering results in Appendix G.

Laboratory ashes were produced at dK-Teknik from wheat straw and wheat grains. Additionally, a series of laboratory ashes were produced from fuels used at MKS1 at Studstrupvaerket, namely a Columbian bituminous coal and danish wheat straw. Mixed ashes were produced both by mixing the fuels before ashing and by ashing the fuels separately followed by physical mixing.

3. SEM images This section contains the SEM images that were obtained during the project. Most of the images were obtained to give representative images of the various samples: wheat straw, wheat grain, straw chars, bottom ash, fly ash, laboratory ash and deposits. Each image is accompanied by a brief description of the sample and the observed features. The description is not meant to be complete and emphasis has been put on the inorganic components. The majority of the images were obtained on embedded polished samples with the Backscattered Electron Detector (BSE). In BSE images brightness is proportional to average atomic number, i.e. most inorganic particles will show up bright compared to organic substances such as carbonaceous matter and embedding material.

-3- Approximately 0.2-2 grams of material were used depending on the available amount. The sample were blended with Epofix or Carnauba Wax during thorough mixing and was left to harden. In the case of Epofix the hardened block were sectioned perpendicular to the direction of sedimentation and subsequently embedded with the new face pointing upwards. By analyzing the whole cross section bias due to gravity fractionation is avoided. In the case of Carnauba

Wax no perpendicular sectioning was performed since the wax hardens almost immediately leaving no time for sedimentation. The blocks were ground and polished with the last step being 0.25 micron diamond paste and subsequently coated with a thin carbon layer in a Polaron TB500 Coater. During handling of the samples no water or alcohol were utilized, since these compounds may dissolve salts. For lubrication and cooling during grinding and polishing either Buehler lapping oil or distilled petroleum were used. Images of straw chars were obtained on powder samples, i.e. samples that were distributed loosely on carbon tape mounted on a sample holder and subsequently coated with carbon.

3.1 Danish wheat straw and wheat grain. The samples were studied to evaluate mineral compositions and to assess the extent to which CCSEM analyses is suited to characterize and quantify inorganic constituents of biomass fuels. The CCSEM technique only measures inorganic particles whereas disseminated material is not detected.

Fig. 3.1. Representative images of biomass fuel examples. a) Milled wheat straw. The inorganic or inorganic-rich constituents show up bright compared to the organic material. Most conspicuous is an elongated rim covering the length of one thin straw fragment. This type of rim is a common feature in the straw and is mainly composed of Si (>90%). Additional inorganic particles consist mainly of terrigeneous derived clay or quartz particles. b) The image shows the margin of a wheat grain. The major part of the inorganic-rich particles are located close to the wheat grain boundary as small K, Mg and P rich spherical particles arranged in clusters. These are phytates which function as important storage-location of the nutrients K and P (Marschner, 1995).

- 4- There is an obvious difference in the composition of inorganic particles in the straw and the grain, the straw being dominated by silica whereas the grain is dominated by K and P rich particles. However, inorganic elements that are disseminated or bound in the organic structure is not readily discernible in the BSE-image. This means that, since CCSEM analysis is controlled by contrasts in BSE images, possibly occurring organically associated elements as for example K and Ca are not detected (see section 4.1).

3.2 Bottom ash and fly ash from straw fired plants: Haslev CHP and Slagelse CHP Bottom ash and fly ash were sampled at two danish straw fired boilers in due of an earlier biomass characterization project (Stenholm et al., 1996). Fly ash was collected in a bag-filter at Haslev CHP and in an electro-filter at Slagelse CHP. Bottom ash was collected from material that were automatically scraped off the grate in both cases. Four of the experiments utilized wheat straw, one used barley straw and one rape straw (Table 3.1). All ash samples were subjected to CCSEM analysis (Sorensen, 1997). The CCSEM results can be found in Frandsen et al. (1998). In this section representative BSE images of the analyzed samples are presented.

Table 3.1 Exoeriment Fuel Haslev 1 Wheat Slagelse 1 Wheat Slagelse 3 Barley Slagelse 6 Wheat Slagelse 7 Wheat Slaoelse 8 Race

-5- 3.2.1. Bottom ash The bottom ash samples were milled at dK-Teknik to be used for traditional fusion measurements. Therefore the form and size of the particles is not necessarily representative of original features.

Fig. 3.2. Representative BSE-images of bottom ash from Slagelse and Haslev CHP (opposite page) a) Bottom ash from Haslev 1 (HAS1BA). Ash particles range in size from around 1 micron up to 200 micron. Variable BSE reflection in individual particles evidence that the particles are composite. Commonly they are composed of a Si-rich area or core (medium grey) associated with a (probably) fused mass of K-Ca-(Fe-AI) rich silica material (white or light grey). An example of this type of zoning is illustrated in Fig.3.3. Unburnt straw (dark grey) is also present. b) Bottom ash from Slagelse 1 (SIA1 BA). Particle compositions are similar to the ones observed in a). The change in particle size from bottom to top in the image is an effect of sedimentation during sample preparation. In CCSEM analysis this fractionation is taken into account by analyzing a whole cross section. c) Bottom ash from Slagelse 3 (SLA3BA). The image is dominated by K-Ca-Si rich as well as Si- rich particles. d) Bottom ash from Slagelse 6 (SLA6BA). In addition to silica-rich particles several crystals of KCI up to 10 micron are observed. e) Bottom ash from Slagelse 7 (SLA7BA). The image is dominated by silica rich composite particles ranging in size up to around 300 micron. f) Bottom ash from Slagelse 8 (SLA8BA). The image is dominated by silica rich composite particles ranging in size from 1 micron up to 500 micron. There seems to be more fine (1-10 micron) material in this bottom ash than in the bottom ashes from the other experiments. 7 Fig.3.3 Potassium-enriched margin on . bottom ash particle a)

Fig.3.3. Compositional zoning in bottom ash particle from Haslev 1 (HAS1BA). a) Shows the outline of a 100 micron silica-rich bottom ash particle, b) location of line scan, c) line scan illustrating relative distribution of Si, K, Ca and BSE grey level. The K (and Ca) content increases whereas the Si content decreases gradually outwards over a distance of ca. 10 micron. The line scan is 58.25 microns long.

- 8- 3.2.2. Fly ash

During preparation of the fly ash samples they were thorougly mixed with epoxy to separate lumps of material that may have formed during sampling or in the sample container. However, as illustrated below much of the fine grained material appears in aggregates in several of the samples (Fig.3.4).

3.2.3. Comments to observations of bottom ash and fly ash from Haslev CHP and Slagelse CHP Generally bottom ashes are dominated by relatively large silica-rich particles, most of them containing K and subordinate Ca. Many of the particles are composite with domains of almost pure silicon. It appears that the silicon in the straw have reacted to varying degrees with loosely bound metals, especially K that were evaporated during combustion. The extent to which this process have occurred indicates that the process took place over some time, i.e. until the bottom ash were scraped off the grate. A line scan extending from a Si-rich core to the surrounding K and Ca enriched margin illustrated that the change in composition occurs gradually over a distance of approximately 10 microns (Fig.3.3). The shape and size of the gradient makes is likely that the reaction between K and the Si-particle is diffusion controlled. In summary it appears that K is bound to a significant extent to silica in the bottom ash.

The fly ash is dominated by submicron or micron-sized particles of primarily KCI with subordinate K2S04 and silica-rich compounds (Fig.3.4). These particles are commonly present in clusters or aggregates that can reach a size of up to 0.5 mm. These large masses probably formed during fly ash sampling or handling. However, minor attachement of fly ash particles to each other or to larger silica-rich particles is common and is believed to have formed in the boiler. The fly ash formed in rape combustion (SLA8FA) is clearly different from fly ash from wheat and barley combustion, especially in terms of a higher content of Ca and S rich particles. However, the bottom ash formed during rape combustion (SLA8BA) does not distinguish itself significantly, probably as a result of carry over of material from the former experiment that utilized wheat as fuel. Fig.3.4. Representative BSE-images of fly ash samples from Haslev and Slagelse CHP

(opposite page). a) Fly ash from Haslev 1 (HAS1FA). The fly ash contains much fine grained (submicron to micron) material and several larger (circa 20-150 micron) particles. Individual submicron particles are too small to analyze, but qualitative SEM/EDX show that they mainly consist of KCI with subordinate K2S04 and small silica-rich particles. In CCSEM analyses these masses of fine­ grained material are commonly analyzed as one big particle because they can not be separated on the basis of a backscatter image. Thin elongated Si-K rich rims up to 200 micron long are typically associated with aggregates of fine material. Composite silica-rich particles similar to the ones observed in the bottom ashes are also present. b) Fly ash from Slagelse 1 (SLA1FA). This sample is similar to the Haslev 1 fly ash. Much of the submicron material masses is associated with Si-K rich rims. c) Fly ash from Slagelse 3 (SLA3FA). In this sample almost all of the submicron material occurs in rounded aggregates up to 200 micron across. Si-K rich rims are associated with such masses. d) Fly ash from Slagelse 6 (SLA6FA). This image shows quite large (up to 0.5 mm) masses of submicron particles. Bright appearing KCI crystals appear to be larger in this sample than in the other fly ashes. The lower right insert shows a spherical Si-K rich particle with a rim of KCI. e) Fly ash from Slagelse 7 (SLA7FA). Contains relatively large (50-200 micron) Si-rich particles with attached masses of KCI and minor K2S04. Also present are large particles consisting of K, Ca and Si. f) Fly ash from Slagelse 8 (SLA8FA). This sample is different from the remaining fly ash samples, reflecting that it originated from rape combustion. Rape has a contrasting composition compared to wheat and barley especially in terms of higher Ca and S and lower Si-content. Most of the particles or masses of particles are rich in Ca, and are probably mixtures of CaS04, Ca- oxide and K2SC4.

- 10- Fig.3.4 Haslev and Slagelse Fly ash

- 11- 3.3 Deposits from the Haslev 1 and the Slagelse 3 experiments. Deposits were collected in the Haslev and Slagelse CHP’s on water-cooled deposition probes.

Probes were placed both in the furnace and in the superheater region (Fig.3.5). After sampling the deposits were first brushed off with a soft brush (“outer” loose part of deposit), and afterwards with a hard brush (“inner” hard part of deposit). The deposit samples were embedded, polished and subjected to CCSEM analysis (section 4.3.2). Representative BSE images of the polished, embedded samples are shown below in Figure 3.6.

HASLEV CHP (23MWth) SLAGELSE CHP (31 MW*) FD FD t 4 SH1 SH1

'SD SH1 > SH2 H+H SH2 V SD

Fig. 3.5. Schematic illustrations of boiler designs for Haslev and Slagelse CHP. Arrows indicate location of deposit probes in the furnace (FD) and in the superheater region (SD) (modified after Hansen etal., 1997)

-12- Fig. 3.6. Representative BSE-images of deposits from Haslev and Slagelse CHP (opposite page) a) SLA3SD-2 (outer superheater deposit). The sample consists mainly of small (< 50 micron) KCI particles and subordinate K-Ca silicates (particles rich in K, Ca and Si, in varying proportions, probably amorphous). b) SLA3SD-3 (inner superheater deposit). The sample consists primarily of KCI with subordinate K2S04. Some Fe-oxide is present as well, probably being disrupted scales of the oxidized probe metal surface. c) SLA3FD-1 (outer furnace deposit). The deposit is dominated by silica-rich particles from 5 to 150 micron. Some KCI is also present mostly attached to silica particles. The silica-rich particles contain varying amounts of K and Ca. d) HAS1SD-1 (superheater deposit). The image shows KCI crystals associated with K2S04 and K-Ca silicate. e) HAS1FD-1 (outer furnace deposit). The image shows the presence of KCI and fused amorphous K-Ca silicate. Thin Si-K rich rims are also present. f) HAS1FD-3 (inner furnace deposit). The image shows a predominance of KCI and subordinate K-Ca silicates that are often spherical and with rims of KCI attached.

3.3.1. Comments to observations on deposits from Haslev 1 and Slagelse 1 Generally the furnace deposits are primarily composed of KCI, K-Ca silicates and K2S04. The silicates commonly appear fused either in the form of amorphous particles or as minor spherical fly ash particles. KCI occur either as submicron to micron sized particles commonly together with K2S04, as singular larger particles or as coatings on silicates.

- 13- 4 Fig.3.6 Haslev and Slagelse Deposits 3.4 Fuels from co-combustion at MKS1, Studstrupvaerket Fuels from the experiments 3, 4 and 5 at MKS1 at Studstrupvaerket were analyzed by CCSEM.

The coal utilized in all three cases were a South American high volatile bituminous coal whereas the feedstock straw was wheat. Representative images of the fuels are shown in Fig.3.7.

r \

' Acc.V Spot Magn Det WD I—■— 500 pm J60KV6Q 63x BSE 10 6 MKS6COAL

Fig.3.7. Representative images of fuels from MKS1. a-c) The BSE images show MKS1 coal from experiments 3, 4 and 5 (MKS3C, MKS4C and MKS5C). The coal particles appear light grey whereas the minerals appear bright. The mineral association is dominated by quartz (Q), various clay minerals (K: kaolinite and I: illite) as well as minor pyrite (P). d) Straw utilized in experiment 4 (MKS4S). This image is an example of the wheat straw used at MKS1. The straw was milled before sampling rendering fragments up to around 2 mm in size. The image shows the common presence of Si-rich margins partially outlining straw fragments.

- 15- 3.5 Straw chars sampled from the burner zone at MKS1 Combustion chars were sampled from the burner zone at MKS1 during three experiments with addition of 20% (exp.2), 10% (exp.4) and 0% straw (exp.9), respectively. Sampling was

performed with a suction pyrometer that extended into a mixing zone of coal and straw (Fig.3.10). Therefore both straw chars and coal chars were sampled. The chars were collected on a sintered copper filter with a nominal pore size of 40 micron (Appendix F). However, the filter-pores are progressively blocked during the sampling time of approximately 30 seconds meaning that smaller particles may be collected in the last phases of sampling. The char samples illustrated in Fig. 3.8 and Fig 3.9 were subjected to CCSEM analysis (see section 4.4).

3.5.1. Summary of chars from MKS1 The straw chars are typically covered or partially covered with an appearing^ fragile looking Si-K rich rim (Fig.3.8). Attached to the rims are small KCI and sometimes K2S04 crystals and a thin appearing K2S04 coating covers much of the rim. The observations indicate that KCI and K2SQ4 have condensed on the Si-K compound either in the flame zone or during sampling. During sampling the flue gas continually flows past the collected, cooled chars and through the filter. Therefore condensation on the chars is likely to occur during this process. Coal chars are typically smaller and rounded compared to the elongated straw chars. Occasionally coal derived fly ash particles are attached to the straw chars, but condensed KCI or K2S04 were not observed on coal chars.This feature indicates that some (perhaps the initial) condensation occurred in the flame zone, because otherwise one would expect coal and straw chars to exhibit similar amounts of KCI and K2S04.

-16 - Fig.3.8. Representative BSE images of char powder samples from experiment 2 (20% straw,

75% load) (opposite page) a) The straw chars are typically elongated and up to 2 mm long, retaining parts of the straw structure. Conspicuous Si-K rich rims are commonly present on the surface of the chars. Coal chars are typically much smaller (around 100 microns) and rounded. b) The image illustrates the distinct difference between the straw chars with partially broken Si-K rich rims and the rounded coal chars. c) Shows detail of Si-K rich rim on straw char. The rim contains a multitude of approximately 15 micron large round or oval Si-rich “lids". The origin of this structure is unknown, but it is speculated that it is determined by the original structure of the straw. Also seen are bright appearing KCI crystals attached to the rim. d) Shows detail of Si-K rich rim. The Si-rich lids are partially detached. Also seen are KCI crystals and a spherical coal derived alumino-silicate fly ash particle. On the Si-K rich rim a thin coating of K2S04 is observed. e) High magnification BSE image showing 2 micron KCI crystals and coating of K2S04 on a Si-K rich rim. f) Aggregate of KCI crystals and partial coating of K2S04 on Si-K rich rim. g) Interior of straw char showing the highly porous structure that is formed during degassing of the straw. Also seen is a broken off Si-K rim. h) Shows detail of char interior. Attached to the organic structure are K2S04, KCI and a spherical coal derived fly ash. IS- Fig.3.9 Chars nomMKSl a) b)

Acc.V Spo^Magn Det WD I------1 200 pm 15.0 kV 45 ro'Ox BSE 10.8 HKSl.koksO, c)

Fig.3.9. BSE images of embedded and polished chars. a) Char from exp.2: 20% straw, 75% load, b) Char from exp.4: 10% straw 100% load, c) char from exp.9: 100% coal, 100% load. The straw chars in a) and b) illustrate the Si-K rich rims as well as attached KCI and K2S04. c) The image shows only coal chars illustrating their smaller size and rounded morphology compared to straw chars.

- 19- A B C Z rA:A- FlattenI superheater (1st/2nd pass) B: Secondary superheater C: Reheater D: Upper primary superheater E: Lower primary superheater F: Economiser

/MEASURING POSITIONS: "N COAL 0: Sampling of char particles COAL & STRAW 1-5: Sampling of deposits, fly ash particles, flue COAL gas concentrations (including alkali) and temperature measurements 6: Sampling of aerosols V.

Fig.3.10. Schematic illustration of the MKS1 boiler at Studstrupvaerket. Filled circles indicate deposition probe locations (modified after Andersen etal., 1997).

3.6 Full scale ashes from MKS1, Studstrupvaerket Fly ash and bottom ash from MKS1, Studstrupvaerket were subjected to CCSEM analysis. The study was concentrated on bottom ash and fly ash from experiments 3, 4 and 5, all with 100% load and 20%, 10% and 0% straw addition, respectively (Appendix G). The bottom ash was collected dry once every hour, whereas the fly ash was collected with electrostatic filter. Fig.3.10 illustrates schematically the MKS1 boiler design. Representative images of bottom ash and fly ash samples are shown in Fig.3.11.

3.6.1. Comments to the bottom ash and fly ash images. The bottom ash and fly ash particles from MKS 1 are strongly dominated by various silica-rich particles (Fig.3.11). Of special interest is the observation of K-containing alumino-silicate particles or rims surrounding cores of almost pure silicon in the experiments with straw addition. Due to the fluxing effect of K on silica-rich compounds their presence will most likely have an adverse effect on deposit formation due to a lower melting point and thereby higher sticking propensity. Clearly, in the case of zoned particles the composition at the margin is quite important for the stickiness of a particle.

-20- Fig.3.11. Bottom ash and fly ash samples from MKS1 (opposite page). a) Bottom ash from experiment 3 (MKS3BA); b) Bottom ash from experiment 4 (MKS4BA); c) bottom ash from experiment 5 (MKS5BA); d) fly ash from experiment 3 (MKS3FA); e) fly ash from experiment 4 (MKS5FA); f) fly ash from experiment 5 (MKS5FA). The bottom ash samples all contain a high proportion of relatively large (up to 300 micron) particles that are mainly composed of silicon with various amounts of other elements such as Al,

K, Ca and Fe. The particles generally appear amorphous and fused. Commonly they contain vesicles and many are composite being composed of both pure silicon and various alumino­ silicates. In the samples from co-combustion (MKS3BA and MKS4BA)it is not uncommon to observe rims of alumino-silicates with a substantial content of K and Ca surrounding Si-rich cores. The fly ash samples (d-f) contain many spherical fused particles that are composed mainly of K- (Ca-Fe) bearing alumino-silicates. Additionally silica-rich particles are prominent and amorphous composite particles consisting of both phases are commonly observed. The alumino-silicates in the sample without straw addition (MKS5FA) contain less K than the two with 20 and 10% straw addition. Fig.3.11 MKS1 Full Scale Ashes 3.7 Deposits from MKS1, Studstrupvaerket Five deposits from MKS1 were analyzed by SEMPC to evaluate their composition as function of straw addition and location in the boiler and to compare them to fly ashes (Appendix G). Schematic location of the deposit probes is illustrated in Fig.3.10. The deposits and the associated straw fractions and load are presented in Table 3.2. The analyzed deposits were all exposed for 18 hours with a probe metal temperature of 540°C. All SEMPC analyses were performed on the upstream deposit, i.e. the position pointing towards the direction of gas flow.

Table 3.2. Deposit samples selected for SEMPC analysis.

Code Straw Load Position

MKS31D-1 20% 100% 1

MKS32D-1 20% 100% 2

MKS33D-1 20% 100% 3

MKS41D-1 10% 100% 1

MKS42D-1 10% 100% 2

MKS51D-1 0% 100% 1

After exposure the probes were embedded in epoxy and cross sectioned, ground and polished with the final step being 0.25 micron. Before SEM-analysis the block was coated with a thin layer of carbon. The SEMPC analysis was performed by setting up a number of adjoining frames at 200 times magnification covering part of the upstream deposit. In each of the frames 20 by 20 points were set up in a grid and analyses were performed at each point. Care were taken to analyse the whole thickness of the deposit. It was unavoidable to overlap a small part of the probe metal when setting up the areas (Fig.3.12). To avoid analyzing epoxy filled pores the analysis procedure was stopped if the total counts were below 600 after 1 second. To avoid inclusion of probe metal in the data treatment results with Cr203 >0.1 % (w/w) were omitted from the dataset. However, some of the oxidized metal-layer may be incorporated in the data treatment. The remaining X-ray spectra were quantified using theoretical standards included in the EDX-system.

-23- Figure 3.12 gives three examples of upstream deposits that were subjected to SEMPC. The thin bright zone at the top of each image shows the probe metal that is overlapped by analysis.

Fig. 3.12. Examples of sections used for SEMPC (opposite page). a) Deposit MKS32D-1, upstream (experiment 3, 20% straw, position 2, 540°C metal temperature, 18 hours exposure). The image is put together of 9 individual 200x magnification photomicrographs. The deposit is approximately 1.2 mm thick on average with an undulating surface. It appears that the presence of large particles coincide with “hills" whereas small particles coincide with “valleys" on the surface. b) Deposit MKS42D-1, upstream (experiment 4, 10% straw, position 2, 540°C metal temperature, 18 hours exposure). The image is put together of 8 individual 200x magnification photomicrographs. The deposit was formed in the same position as in a) but with only 10% straw addition compared to 20%. The deposit varies in thickness between 100 micron to 1.2 mm and is strongly undulating. c) Deposit MKS51D-1, upstream (experiment 5, 0% straw, position 1, 540°C metal temperature, 18 hours exposure). The image is put together of 9 individual 200x magnification photomicrographs. The deposit varies in thickness from 0 to 1.3 mm in a “finger-like” structure. Again it seems that the outward extending fingers coincide with relatively large particles.

The SEMPC technique yields no information of particle size or distribution. Therefore an Xray mapping of a selected superheater deposit was performed to illustrate these features (Fig.3.13).

The brightness on the map is proportional to the intensity of the X-ray energies assigned to each element. The X-ray mapping shows that K2S04 is largely present as a binding phase covering the surfaces of fly ash particles.

3.8 Laboratory ashes Fuels were ashed at dK-Teknik in order to study the relationship between laboratory ashes and full scale ashes and to assess the extent to which ashing may assist in predicting ash deposition problems. In this section a number of representative BSE-images of laboratory ashes are shown.

-24-

Fig.3.13. MKS32D-1, UP, X-RAY MAPPING BSE Si K S

Fig.3.13. X-ray mapping of MKS32D-1, upstream. The Si mapping illustrates the presence of silica-rich fly ash particles. Evidently fly ash particles with no or low Si are also present. These are mainly rich in Fe. The mapping of K and S shows a clear correlation between K and S in thin coatings and zones between fly ash particles, indicating that K2S04 constitutes a binding phase of this superheater deposit. The frame furthest to the left shows a backscatter image for reference. 3.8.1 Ashed danish wheat straw and wheat grain As examples of two biomass fuels of contrasting compositions wheat straw and wheat grain were ashed at dK-Teknik. The samples were ashed at both 550°C and 815°C to test the consequence of different ashing temperatures for the produced ash. Representative BSE images are shown in Fig.3.14. a) b)

Fig.3.14. Wheat straw and wheat grain laboratory ashes. a) Wheat straw ashed at 815°C (LAS2). The ash consists of strongly fused, vesicular amorphous particles up to millimeter size. The particles are mainly composed of K-Ca silicate with subordinate K-phosphates and K-sulphates. b) Wheat grains ashed at 815°C (LAST). The ash consists of fused particles reaching up to several mm in size (not shown).The particles contain various phosphate phases, the brightest being rich in K and Ca, whereas the darker areas are rich in K and Mg. Regular grain margins indicate that the slags have recrystallized into these compounds, c) Wheat straw ashed at 550°C (LAS1). The sample consists partly of sub­ micron to micron-size particles of KCI and K-phosphate and partly of larger K-Ca silicates either occurring as elongated rims or as equant grains. Morphologically the laboratory ash is quite similar to the fly ash samples shown in Fig.3.4. However, the laboratory ash typically contain a smaller amount of KCI than straw fly ash. This can partly be a result of evaporation of some Cl during the laboratory ashing even at this relatively low temperature, but is mainly a consequence of element fractionation between bottom ash and fly ash so that the fly ash is relatively enriched in KCI (see section 4.3.1). d) Wheat grains ashed at 550°C (LAS6). The ash particles vary in size from around 5 micron to 250 micron. As was the case for the 815°C ash, the particles consist mainly of various phosphates rich in K with subordinate Mg and Ca. 3.8.1.2. Comments to the straw and grain laboratory ashes The ashes that were produced at 815°C are strongly fused and consist of massive slags. Based on morphology they are therefore very unlike fly ash. Also taking into account that volatile components as for example Cl and K evaporates to a significant extent this temperature is not considered viable for producing ashes to be utilized for fusion or sintering experiments. In contrast ashing at 550°C produces particulate straw ash that morphologically appears similar to full scale fly ash as observed in the Haslev and Slagelse samples.

3.8.2 Ashed wheat straw, coal and fuel blends from MKS1 at Studstrupvaerket Fuels (coal and straw) and mixed fuels (80%coal, 20% straw on an energy basis) from MKS1 experiment 3 were ashed in the laboratory at dK-Teknik. Additionally, coal from experiments 4 and 5 were ashed. Coals were ashes at 815°C whereas straw and straw-coal mixtures were ashed at 550°C to minimize evaporation of alkalies. All laboratory ashes were analyzed by CCSEM and representative BSE images of the ashed samples are shown i Fig. 3.15.

3.8.2.1 Comments to MKS1 laboratory ashes The coal ash samples are dominated by relatively large particles of alumino-silicates, quartz, Ca (Mg) sulphates and minor Fe oxide (Fig.3.15). Fine-grained material is largely composed of silica-rich compounds. Straw ash on the other hand is mainly composed of elongated flakes of Si-K rich particles that morphologically resemble similar flakes observed in straw chars (Fig.3.9). Masses of fine-grained material are typically composed of mixtures of K, Ca, Mg, Cl, S and P. The particles in these masses are generally too small to be individually analyzed. Fine-grained material is also seen lying freely in the embedding epoxy. The physically mixed ash (MKS3POSTMIX) is dominated by the coal ash. From the presence of K2S04 in MKSPREMIX it appears that K preferentially reacts with S preferentially to Cl, the S being supplied by the coal.

-28- Fig.3.15. Laboratory ashes from MKS1 (opposite page). a) MKS3LAC. 815°C coal ash, experiment 3. In the image is observed various alumino-silicates, calcium (magnesium) sulphates and iron oxides. The sample is relatively coarse grained with particles extending up to around 200 micron. The fine-grained particles consist mainly of Al-Si rich material. b) MKS4LAC 815°C coal ash, experiment 4. The image is dominated by coarse particles of various alumino-silicates, Ca-Mg sulphates and iron oxide. The diffuse dark grey round structures that are present are pieces of an acrylic filler that were mixed in to keep ash particles from touching each other. c) MKS5LAC. 815°C coal ash from experiment 5. The image illustrates various alumino-silicates, Fe-oxide, Ca sulphate and Ca-Mg sulphate. d) MKS3LAS. 550°C ash of wheat straw from experiment 3. Observed are thin elongated particles composed primarily of K and Si. Attached to these are typically masses of fines composed of mixtures of K, Ca, Mg, S, Cl and P. These particles are too small to analyze individually. e) *MKS3POSTMIX. This sample is a blend of coal laboratory ash and straw laboratory ash (i.e. MKS3LAC and MKS3LAS). The ashes are mixed to represent 20% straw addition on an energy basis. The sample is dominated by the coal ash (alumino-silicates and sulphates) wheras only a few K-Si rich flakes are evident. The sample were prepared by physically mixing the two ashes after the ashing process. f) MKS3PREMIX. The sample was prepared by first blending straw and coal and subsequently ashing the mixture at 550°C. The most significant difference to MKS3POSTMIX is the presence of K2S04 and the apparent lack of KCI.

-29- Fig.3.15 MKS1 laboratory ashes

Fe-oxide _-.v' • CaMgS04

Acc.V Spot Magn Dot WD I------200 pm 15.0 kV 4.6 lOOx BSE 10.7 MKS4LAC c)

'CaMgSpj- SU' Al-Si

""" % Acc.V Spot Magn Dot WD I------1 200 pm 16.0 kV4.6 lOOx BSE 10.7 MKS5LAC 4. CCSEM results In this section the results from the CCSEM analyses are presented and discussed. As mentioned above, results from Haslev CHP and Siagelse CHP are treated in Sorensen (1997) (Appendix A) and Hansen et al. (1997) (Appendix D), and therefore only the main results will be discussed here. Results from MKS1 are reported in Appendix G. The full set of CCSEM and SEMPC data are presented in Frandsen et al. (1998).

4.1. CCSEM analysis Between 0.2 g and 1 grams of each sample were embedded in epoxy and after hardening

sectioned into two pieces parallel to the settling direction. The two pieces were embedded with

the new face pointing upwards to avoid density bias due to settling. During the work it was found that many particles settled to touch each other. This induces an adverse effect on the CCSEM analysis. Therefore some samples were mixed with a filler consisting of 10-50 microns diameter acrylic balls to keep ash particles from touching. The blocks were ground and polished using 1/4 mm diamond powder for the final step. No water or alcohol were used during preparation since both can dissolve various salts present in the sample material. Either Buehler Oil or distilled petroleum were used as lubricant and cooling during grinding and polishing.

The polished blocks were coated with a thin carbon layer in a Polaron TB500 coater. The SEM work was performed on a Philips XL 40 Scanning Electron Microscope with a Noran Instruments Voyager II X-ray analysis system attached. The use of CCSEM analysis for characterization of minerals in coal and coal ash at GEUS has recently been described in detail by Laursen (1997a; 1997b) and therefore only the basics of the technique will be mentioned here:

1. A number of points is set up randomly to cover a large part of the sample surface. At each point an analysis is performed at three different magnifications (25x, 100x and 500x) to achieve good resolution in the whole size range. 2. A Backscatter Electron (BSE) image with a good separation between inorganic and organic material is acquired. 3. An appropiate grey-level threshold value is set to create a binary image with only inorganic particles singled out. 4. The binary image is used to control the beam to perform a scan across each individual particle for five seconds. A particle is here defined as any cluster of connecting white pixels in the binary image. 5. An energy-dispersive X-ray-spectrum and morphological data are acquired during each raster scan yielding compositional and morphological data for each individual particle.

It shall be stressed that the CCSEM results are semi-quantitative for several reasons:

1. The EDX spectrum is not corrected for ZAP effects (Z: atomic number, A: absorbtion and F: fluorescence). 2. Particle size is based on cross sectional area and weight percent is estimated on the assumption that all particles are spherical. This involves an uncertainty for particles with more complicated shapes. 3. Particles with average diameters below 1 micron are omitted from the CCSEM analysis because much of the detected X-rays are emitted from a volume beyond the particle boundary, potentially leading to erroneous or even meaningless results. 4. Organically bound or otherwise disseminated elements are not detected by the CCSEM method which only “sees” particles with high BSE reflection, i.e. inorganic or inorganic-rich particles.

-31 - These aspects are, however, not critical for the CCSEM analysis because the X-ray spectra are primarily used to group particles into mineral categories with rather wide ranges of composition. The strength of a CCSEM analysis lies in information about element distribution between different particle types and different samples and in the fact that each particle is characterized with respect to both size and composition. However, it is emphasized that a CCSEM analysis is not directly comparable to a traditional bulk chemical analysis. The raw X-ray data are used to group the analyzed particles into a number of categories, commonly called mineral categories. However, the categories are only defined by composition and bears no information on crystallinity. The principles of data reduction are described in Sorensen (1997) and are summarized below:

The strategy for modifying the CCSEM procedures was to obtain an elaborate set of CCSEM raw data on straw ashes from Slagelse CHP and Haslev CHP and subsequently to utilize them to develop a new data reduction algorithm. In total 12 ashes were analyzed by CCSEM. Initially the raw data were reduced by using coal mineral characterization algorithms developed by Laursen (1997a). Due to the contrasting composition of straw ash and coal ash this resulted in a high number of unclassified particles (35-70%), i.e. particles that do not fit the criteria for any of the mineral categories. The compositions of the unclassified particles were subsequently studied in detail, and it was found that particles rich in chlorine and sulphur constituted a significant part pointing to the importance of chlorides and sulphates, compounds that have been reported present in biomass ashes elsewhere (e.g. Glanders and Steenari, 1994; Steenari and Langer, 1988; Bryers, 1994). In the case of wheat grains the unclassified fraction is dominated by particles rich in P (K, Mg and Ca rich phosphates).

On the basis of semi-quantitive SEM/EDX analysis and corresponding ratios between X-ray intensities a number of new categories were established in order to characterize the ash particles more complete. After each modification of the algorithms the datasets were categorized again and the amount of unclassified particles were taken as a measurement of the success of the algorithm.

Modifications of the algorithm included addition of a number of new categories. A composite category of KCI associated with silicates (KCI+silicate) were included, an association that is commonly observed in especially fly ash samples. These composite particles either represent Si-particles with condensated KCI or they result from close association between KCI and silicate induced by settling in the epoxy before it hardens (Fig.3.4). Such adjoined particles are seen as one single particle by an automatic CCSEM analysis and will result in an “average” composition ” with characteristics from both particles. A category of particles rich in K, Ca and Si called “K-Ca silicate ” were included in the algorithm. Characteristic for this category is a very low Al-content in contrast the case of clay-derived particles originating from soil. The K-Ca silicates can form by reaction between silica from the straw and vaporized inorganic species in the flue gas. Some of the K-Ca silicates, however, may have formed by reaction between volatilized inorganic species and terrigeneous derived quartz grains. Additionally categories for K2S04 and KCI were included in the algorithm as these phases constitute significant parts of especially the fine flyash fractions. A range of phosphates: K phosphate, Ca phosphate and K-Mg phosphate were included in order to encompass ash particles derived from wheat grain. The unclassified group were subdivided into unclassified phosphates, - sulphates, - chlorides and - silicates on the basis of their contents of P, S, Cl and Si, respectively, based on the assumption that particles in these groups will behave differently in terms of their response to high temperatures. Most of the categories from the coal mineral algorithms were retained and their proportions were practically unaltered compared to the initial data-reduction. This is important for future comparison between CCSEM analyses of coal ash and straw ash and especially for CCSEM analysis of ash from co-combustion of straw and coal.

The modification of the algorithms resulted in a decrease in the proportion of unclassified to between 4% and 25% for bottom ash from wheat and barley straw, whereas fly ash contained between 6% and 27% unclassified. In the case of rape straw, which were included in the analyses, unclassified still constituted 63% of the fly ash, underlining that it is problematic to encompass ash from biomass with contrasting compositions into one datareduction algorithm. The algorithm presently used is tuned to characterize ash from combustion of wheat and barley straw as well as ashes from co-combustion of straw and coal.

Table 4.1 shows the algorithm that is presently used for characterization of ash from straw and from co-combustion of straw and coal.

-33- Table 4.1. Data reduction algorithms for ashes and deposits from straw and straw+coal combustion quartz: A! <=10, Si >=80 ironoxide: Mg <= 5, Al <= 5, Si < 10, S <= 5, Fe >= 80 periclase: Mg >= 80, Ca <= 5, S <= 5 rutile: S <= 5, Ti >= 80 alumina: A! >=95 calcite: Mg <= 5, Al <= 5, Si <= 5, P <= 5, S < 10, Ca >= 80, Ti <= 5 dolomite: Mg >= 5, Ca > 10, Mg + Ca >= 80 ankerite: Mg < Fe, S < 15, Ca > 20, Fe > 20, Ca + Mg + Fe > 80 kaolinite (Ai-Si): Na <= 5, K <= 5, Ca <= 5, Fe <= 5, Al + Si >= 80, (0.8*Al) < Si, Si < (1.5*A1) montmorrillonite (Ca-Al-Si): Na <= 5, K <= 5, Ca <= 5, Fe <= 5, A1 + Si >= 80, (1.5*Al) < Si, Si < (2.5*Al), iliite (K-Ai-Si): Na <= 5, Ca<=5, Fe<=5, Al >= 15, Si > 20, K > 5, K + Al + Si >= 80, Fe-AI silicate: Na<= 5, Al >= 15, Si >20, K<= 5, Ca<= 5, Fe> 5, Fe +Al + Si >= 80 Ca-Al silicate: Na <= 5, A! >= 15, Si >20, K <= 5, Ca >= 5, Ca +Al + Si >= 80, Fe <= 5 Na-Al silicate: Na>=5, Al>=15, Si > 20, Na + Al + Si >= 80, K <= 5, Ca <= 5, Fe <= 5 aluminasilicate: Na <= 5, Al > 20, Si > 2, Si + Al >= 80, K <= 5, Ca <= 5, Fe <= 5 mixed silicate: Na< 10, Al >20, Si>20,K<10, Ca< 10,Fe< 10,Na+AI + Si + K + Ca+Fe>=80 Fe silicate: Na <= 5, Al <= 5, Si > 20, S <= 5, K <= 5, Ca <= 5, Fe > 10, Fe + Si >= 80 Ca silicate: Na <= 5, Al <= 5, Si > 20, S <= 5, K <= 5, Ca > 10, Ca + Si >= 80, Fe <= 5 Ca aluminate: Al > 15, Si <= 5, Ca > 20, Ca + Al >= 80 pyrite: S > 40, Ca < 10, Fe >= 10, Fe + S >= 80, Fe<=(S * 0.7) pyrrhotite: Fe +S >= 80,Fe>20, S>20, (0.7*S) 20, Ca >20, Ca + S> 80, Ti < 10, P <= 5 apatite: Al <= 5, P >= 20, S <= 5, Ca >= 20, Ca + P >= 80 Ca-Al phosphate: Al > 10, Si <= 5, P > 10, S <= 5, Ca > 10, Al + P + Ca>= 80 KC1: K >= 30, Cl >= 20, K + Cl >= 70 K2S04: K. >= 30, S >= 20, K + S >= 70 K phosphate: K >= 35, P >= 30, K + P >= 70 K-oxide/hydroxide: K>=90 Ca phosphate: Ca >= 30, P >= 30, Ca + P >= 70 K-Mg phosphate: K >= 25, Mg>= IS, P >=20, K+ Mg + P >= 80 NaCi: Na >= 30, Cl >= 30, Na+Cl >= 80 K-Ca silicate: Si >= 20, Si <= 80, K + Ca >= 25, Al<=2 Si-rich: Si >= 65, Si <= 80, S <= 10, P <= 10, Cl <= 10 Ca-rich: Al < 15, Ca >= 65, Ca <= 80, P <= 10, S <= 10, Cl <= 10 Ca-Si-rich: Si >= 20, Ca >= 20, Si + Ca >= 70, S <= 10, Cl <= 10, P <= 10 KCl+silicate: K >= 20, K >= (0.9 * Cl), K <= (1.7 * Cl), Si >= 15 unclassified silicate: Si > 20, S <= 5, P <= 5, Cl <= 5 unclassified sulphate: S >=20, S <= 5, P <= 5, Cl <= 5 unclassified chloride: Cl >=20 unclassified phosphate: P >=20 unclassified: remaining particles

4.2. Danish wheat straw, wheat grain and laboratory ash CCSEM analyses were performed on examples of two biomass fuels to assess the suitability of the technique for characterization of inorganic elements in biomass. The results are given on a mineral category basis together with results for 550°C laboratory ashes in Table 4.2.

CCSEM analyses of wheat straw and grains show that the inorganic fraction differs strongly from what is usually observed in coal; clay-minerals and pyrite, that are both important mineral constituents in coal only constitute insignificant proportions of the detected inorganic particles in straw and grain. The most important categories in the straw are quartz (mainly in the form of hydrated amorphous Si02 x nH20), various phosphorous - and chlorine rich phases as well as potassium-sulphate. In contrast, wheat grains have a distinctly different association of inorganic

particles, being dominated by phosphorous-rich phases with subordinate quartz (Table 4.2).

Table 4.2. CCSEM results on a weight percent category basis category Wheat straw Straw lab.ash Wheat grain Grain lab.ash 550°C (LAS1) 550°C (LAG1) Apatite (Ca-P) 1.6 2.1 k2so4 3.6 3.2 Ca silicate 2.0 3.2

Ca-rich 2.6 1.0 Calcite 2.4 2.8 Dolomite 1.2 K-phosphate 94.4 K-AI silicate 1.3 K-Ca silicate 39.3 K-Mg phosphate 42.7 KCI 14.5 KCI+silicate 2.6

Quartz 52.0 6.0 10.8 1.7

Si-rich 1.6 0.5 Unclassified chloride 2.5 Unclassified phosphate 6.7 1.9 23.1 1.0 Unclassified silicate 4.2 5.0 Unclassified sulphate 1.2 Unclassified 13.6 26.1 4.3 1.1 Sum 94.0 96.1 95.5 98.2

Only categories that constitute more than 1 % (w/w) are shown.

The level of K and Cl detected in the straw is quite low taken into account that these elements are important constituents of untreated straw. This is because the CCSEM technique only detects inorganic material that is present in distinct inorganic particles, meaning that low- concentration disseminated or organically bound elements are not analyzed. To try to assess the content of organically bound inorganic elements SEM/EDX analyses were performed on areas without any inorganic particles, revealing less than 0.2% K20 and below 0.25% Cl on a weight basis. Concentrations in this range are too low to allow quantification on SEM/EDX- results, meaning that dispersed elements in the organic structure can be detected but not quantified by SEM/EDX.

Wheat straw and wheat grains were ashed in the laboratory at 550°C. The results show that there is a marked compositional difference between the ash from straw and grains, respectively

(Table 4.2). The grain ash is dominated by K (Mg, Ca)-phosphates in accordance with the phosphorous-rich composition of the grains. In contrast, the straw ash is dominated by silicates,

KCI and smaller amounts of K2S04 and unclassified chlorides. The high content of silicates is

-35- expected due to the high content of silicates in the straw. The silicates in the ash, however, are

mainly K-Ca silicates with a low content of Al, in contrast to silica-rich ash particles formed

during ashing of coal, that (except for quartz) generally are rich in Al. The K-Ca silicate particles

are believed to have formed during the ashing by reaction between organically dispersed K and

Ca with remains of the straw-derived silica. This assumption is supported by the numbers

showing that the straw mineral fraction contains more than 50% of the quartz category and no K- Ca silicates, whereas the low temperature ash contains only 6% quartz and 39% K-Ca silicate

(Table 4.2). KCI constitutes 14% of the laboratory ash whereas this category was not detected in

the straw fuel, indicating that much of the KCI is mainly formed from dispersed K and Cl during

the ashing process. Alternatively, KCI crystals in the straw are too small ( < 1 micron) to be

detected by CCSEM analysis.

4.3. Haslev CHP and Slagelse CHP

The CCSEM study was focused on bottom ash and fly ash. Additionally, fuel from Haslev 1 and

deposits from Slagelse 3 and Haslev 1 were analyzed. A short description of the sampling

procedures are given in section 3.2 and BSE images are shown in section 3.2.1 and 3.2.2.

4.3.1. Bottom ash and fly ash from straw fired plants: Haslev CHP and Slagelse CHP

The CCSEM results of the bottom ash and fly ash from the 6 experiments are shown in Table

4.3 and illustrated in pie-diagrams in Fig. 4.1.

Table 4.3. CCSEM results of ashes from Haslev and Slagelse CHP presented as weight percent on a category basis Category HAS1BA SLA1BA SLA3BA SLA6BA SLA7BA SLA8BA HAS1FA SLA1FA SLA3FA SLA6FA SLA7FA SLA8FA

Fuel wheat wheat barley wheat wheat rape wheat wheat barley wheat wheat rape

KzSO, 1.0 1.1 1.7 1.0 8.3

Ca-rich 9.5

Ca-Si-rich 1.1 1.4

K-Al silicate 1.1 1.3 1.0 1.4 1.1

K-Ca silicate 34.1 30.0 83.8 79.3 58.7 67.9 14.8 4.2 2.2 1.0 4.0 10.9

KCI 2.3 1.8 26.9 39.2 62.1 85.0 58.3 2.5

KCi+silicate 8.4 19.9 6.5 11.4 1.1

Quartz 32.2 47.0 5.2 5.5 19.9 11.0 17.1 12.4 4.4

Si-rich 19.4 20.7 5.5 9.0 15.7 11.3 8.5 4.9 3.5 1.6

Unci, chloride 3.5 4.5 1.4 3.9

Und. silicate 1.3 1.2 1.2

Und. sulphate 1.0 1.7

Unclassified 5.1 2.1 1.7 1.3 4.3 13.8 11.6 26.9 3.7 10.4 60.3

Sum 96.5 97.7 98.9 97.2 96.9 97.0 97.8 96.7 92.9 97.7 98.0 95.8

Only categories that constitute more than i % (w/w) are shown. SLA1BA SLA1FA SLA7BA SLA7FA 4 r

SLA3FA SLA3BA SLA8BA SLA8FA

SLA6BA SLA6FA HAS1BA HAS1FA 00

I—ICa-Sirich gg KCI+silicate mca-rich I—I K-Ca silicate ■i unci, chloride I I Al-silicates l^guncl. sulphate ■ Quartz I—| unclassified ■ K2S04

Fig.4.1. Pie diagrams of fly ash and bottom ash from Haslev CHP and Slagelse CHP

K-Ca silicate

SLA3BA ■ Bottom ash SLAGBAd A Fly ash SLA8BA 0 A8RA SLA7BA

HAS1B. SLA1BAW HAS1 FA

Quartz +. aluminosilicates

Fig. 4.2. Fly ash and bottom ash samples from Haslev and Slagelse experiments

-37- The bottom ashes are typically rich in silica-rich categories, especially K-Ca silicate, quartz and

Si-rich. Only minor KOI is present and K2S04 is only detected in HAS1BA. The unclassified category constitutes less than 5% in the case of wheat and barley combustion and only 4% in the case of rape combustion. In contrast, the fly ashes are generally rich in KCI and KCI+silicate and have a rather low but variable content of K-Ca silicates. K2S04 is present in minor quantities except in the rape fly ash (SLA8FA) where it constitutes 8%, reflecting that rape is rich in sulphur compared to wheat and barley. The category of unclassified constitute between 4% and 14% in the case of wheat combustion, 27% for barley and 60% for rape combustion, illustrating that the data reduction algorithms are suited to characterize fly ash from wheat and barley combustion.

However, in the case of a fuel of contrasting composition such as rape, a high amount of unclassified particles remain.

To illustrate the variations in composition and in an attempt to correlate category composition with melting behaviour of straw ashes, the results are plotted in a triangular diagram of KCI, K-

Ca silicates and quartz+alumino-silicates (Fig.4.2). The group of alumino-silicates consists of

Ca-AI silicates, Fe-silicates, Fe-AI silicates, Si-rich, K-AI silicates, Al-silicates, mixed silicate and unclassified silicates and is believed mainly to be derived from terrigeneous dust. The quartz category may both be derived from the soil and from the straw. The three end-members constitute more than 90% of the bottom ashes and more than 60% of the fly ashes with the exception of rape fly ash (16%). The basis for creating this type of diagram is that it is expected that the melting of the ash will correlate with position due to the contrasting melting temperatures of the end-members. KCI melts at 770°C whereas K-Ca silicates may melt at temperatures as low as around 740°C if present in proportions close to the eutectic composition [Morey et al.,

1931]. However, most of the K-Ca silicate particles have compositions in the range of 5-15%

CaO, 70-85% Si02 and 10-25% K20 as illustrated in the three dimensional triangular diagram in

Fig. 4.3. These compositions will begin to melt at the eutectic temperature, but the major part will, under equilibrium conditions, not melt until higher temperatures have been reached. The group of “quartz+alumino-silicates ” will clearly have a wide range of melting temperatures, but it is assumed that in general the melting will occur at higher temperatures than for K-Ca silicates. This is especially true of the quartz category which on its own will not melt until 1713°C [Deer et al., 1966]. Separation between bottom ash and fly ash is distinct in the triangular diagram of KCI

- K-Ca silicates - quartz+alumino-silicates (Fig.4.2). Bottom ash is situated close to the silicate baseline, whereas fly ash range from close to the KCI apex roughly towards a ratio of 85%

“quarts+aluminosilicates" and 15% K-Ca silicates. This indicates that the silicate association in the fly ash generally has a lower content of material formed by reaction between silica and gas alkalies (K-Ca silicates) than what is observed in the bottom ash. The location of SLA8FA (rape fly ash) is not considered since only 16% of the total is represented in the diagram. By

-38- comparison with ash fusion tests by STA it was shown that ash fusion correlates with position in

the diagram with the lowest fusion temperatures close to the KCI apex and the highest close to

the quartz+alumino-silicates apex. A further discussion of position in the diagram in relation to

measured fusion temperatures is presented in Hansen et al. (1997) (Appendix D).

a) SLA7BA b) SLA7FA

Fig.4.3. Composition of K-Ca silicates in three-dimensional triangular diagrams. The relative heights of bars indicate the cumulative weight percent.

4.3.2. Deposits sampled at the Haslev 1 and Siagelse 3 experiments.

Deposits were sampled on water cooled deposition probes both in the furnace and the

superheater region of the Haslev and Siagelse boilers. For probe positions and sampling

procedures see section 3.3 and Fig. 3.6. The CCSEM results for the Haslev deposits are

presented together with a fly ash analysis for comparison in Table 4.4.

Table.4.4. CCSEM results for fly ash and deposits from Haslev 1. category HAS1FA HAS1SD-1 HAS1SD-2 HAS1FD-1 HAS1FD-2 HAS1FD-3

k2so4 1.1 1.8 Ironoxide 2.2 K-AI silicate 1.4 K-Ca silicate 14.8 3.8 1.1 28.9 19.8 2.3 KCI 26.9 76.0 80.8 61.1 71.2 76.6 KCI+silicate 8.4 3.0 6.1 1.6 2.6 7.7 Quartz 17.1 1.0 Si-rich 8.5 Unclassified chloride 3.5 4.4 6.6 1.3 4.6 Unclassified phosphate 2.7 Unclassified silicate 1.2 Unclassified sulphate 1.0 1.2 1.4

Unclassified 13.8 4.9 3.0 3.0 3.6 6.3 Sum 97.8 96.7 99.0 98.59 97.1 99.6

Only categories that constitute more than 1 % (w/w) are shown.

-39- Figure 4.4 shows the average composition with respect to chlorides (KCi+unclassified chloride),

KCI+silicate, K-Ca silicates and quartz+ Si-rich, of the respective deposit-types and fly ash. The

content of K-Ca silicate is significantly higher and chlorides a bit lower in the outer furnace

(loose) deposit than in the inner (hard) deposit. The iron-oxide in the inner part of the furnace

deposit (Table 4.4) is probably parts of the oxidized probe metal surface that were detached

during brushing off the deposit. The outer (loose) part of the superheater deposit is similar to the

inner part of the furnace deposit in terms of chloride and K-Ca silicate content. The results indicate that in the furnace KCI deposition by condensation is most important in formation of the

inner deposit, whereas fly ash particle impaction is a dominating process in the outer part.

Additionally, it appears that condensation of KCI is more important in the superheater region than in the furnace. When comparing the deposits in general to the fly ash it is seen that the deposits are all higher in chlorides and lower in quartz than the fly ash. Additionally, it is seen that the silicate fraction changes from being dominated by quartz to be dominated by K-Ca silicates when going from fly ash to deposits, indicating that reaction between K and silicates continues during and after deposit formation.

□Chloride

BKO+s&'cate □K-Ca sEcate ■Quartz+Skich

a 50,0

Ha1FA SH-outer F-outer F-inner

Fig.4.4. Average CCSEM category compositions of fly ash (HAS1FA), outer (loose) superheater deposit (SH-outer), outer (loose) furnace deposit (F-outer) and inner (hard) furnace deposit (F-inner).

CCSEM results for the analyzed Slagelse 3 deposits are shown together with a fly ash analysis for comparison in Table 4.5. Figure 4.5 shows the average composition with respect to chlorides

(KCI+unclassified chloride), KCI+silicate, K-Ca silicates and quartz+Si-rich, of the respective deposit-types and fly ash. The inner deposit from the furnace probe is depleted in K-Ca silicates and enriched in KCI as well as KCI +silicate compared to the outer (loose) bulk sintered deposit part.

-40- Table 4.5. CCSEM results for fly ash and deposits from Slagelse 3.

Category SLA3FA SLA3SD-1 SLA3SD-2 SLA3SD-3 SLA3FD-1 SLA3FD-2 SLA3FD-3

k2so4 1.7 1.4

Ironoxide 1.2

K-Ca silicate 2.2 4.6 2.1 68.1 68.1 28.0

KCI 62.1 96.1 89.1 83.4 18.1 2.5 35.6 KCI+silicate 1.6 5.0 Quartz 1.0

Unclassified chloride 1.1 6.2

Unclassified silicate 1.1

Unclassified 26.9 2.6 4.4 6.1 12.8 24.8 23.4

Sum 92.9 98.6 98.1 97.0 99.1 97.4 98.2

Only categories that constitute more than i % (w/w) are shown.

This indicates that the initial furnace deposit is formed primarily from condensation of KCI whereas for the outer part, which contains a large fraction of silicates, impaction of fly ash particles played a more dominant role. Both the inner and outer deposits from the superheater probe contain more chloride than does the inner furnace deposit, indicating that deposit formation by KCI condensation is more important in the superheater than in the furnace, substantiating the results from Haslev 1. The fly ash sample is notably low in silicate-rich particles and rich in KCI, being quite similar to the superheater deposits.

□ Chloride g KCH-silicate □ K-Ca silicate g Quartz+ Si-rich

SLA3FA SH-outer SH-inner F-outer F-inner

Fig.4.5. Average CCSEM category compositions of fly ash (SLA3FA), outer (loose) superheater deposit (SH-outer), inner (hard) superheater deposit (SH-inner), outer (loose) furnace deposit (F-outer) and inner (hard) furnace deposit (F- inner). 4.4 Co-combustion of straw and coal at MKS1, Studstrupvaerket, Char Samples

Various samples were obtained from MKS1 at Studstrupvaerket. The results for fuels, bottom,

ash, fly ash and deposits are treated in Appendix G and are therefore not discussed in this

section. This section deals mainly with CCSEM results for chars that were sampled in the burner zone where straw and coal were mixed. The objective was to acquire a snapshot of the

conditions present at an intermediate stage of combustion between raw fuel and ash. Chars were sampled at three experiments burning 20%, 10% and 0% straw, respectively (MKS2CH, MKS4CH and MKS9CH). BSE-images of chars are shown in Figs.3.8 and 3.9. Results of the

CCSEM analysis of chars are shown in Table 4.6. The samples consist of both straw chars and

coal chars so the results represent a mixing of inorganic components of both types.

There are quite large variation in the CCSEM composition between the three samples. The differences can to a large extent be accounted for by varying proportions of coal char and straw char in the sample. Therefore the main finding is the significant difference between char from coal combustion on one side (MKS9CH) and chars from straw + coal combustion on the other side (MKS2CH and MKS4CH) (Fig.4.6). The compositional variations between chars from the two straw+coal fired experiments probably mainly reflect local variations in the straw/coal ratio at the exact sampling location rather than variations in the overall straw fraction. Chars from coal combustion are rich in various Al-silicates (a grouping of all alumino-silicates except K-AI silicate) compared to co-combustion chars. Otherwise the main differences are that K2S04 and other sulphates (unclassified sulphates) as well as KCI are present in high amounts in co ­ combustion chars and not in coal chars. As discussed in section 3.5.1 these phases are to a large extent believed to be condensed on the chars during sampling. However, their presence substantiates that some K, Cl and S are vaporized during the initial degassing phase.

-42- Table 4.6. CCSEM results on a category basis for MKS1 chars.

Category MKCS2CH MKS4CH MKS9CH

Alumino-silicate 6.4 Ca-P 1.8

k2so4 11.9 23.2 Fe-AI silicate 6.8 K-AI silicate (lllite) 16.6 7.6 7.6 K-Ca silicate 2.3 1.6 Al-Si (kaolinite) 3.4 18.8 KCI 7.6 8.1 Ca-AI-Si (montmorrillonite) 1.6 16.3 Pyrite 1.2 Quartz 31.7 24.3 25.9 Si rich 3.9 5.5 Unclassified phosphate 2.0 'Unclassified silicate 4.2 1.9 10.0 'Unclassified sulphate 1.2 8.4 1.3 Unclassified 8.5 10.6 1.3

Sum 93.0 94.8 95.6

Only categories that constitute more than 1% (w/w) are shown.

eIVKCS2CH

Fig.4.6. Schematic illustration of composition by category for MKS1 chars. MKS2CH: 20% straw; MKS4CH: 10% straw;

MKS9CH: coal.

-43- 4.5. Ashed wheat straw, coal and mixtures of wheat straw and coal from MKS1 at Studstrupvaerket Laboratory ashes of MKS1 fuels were analyzed by CCSEM to evaluate what consequence “premixing" or “postmixing” has on resulting ash composition (see below). Five laboratory ashes were analyzed as schematized in Fig.4.7. BSE images of the ashes are shown in Fig.3.15.

MKS3LAC ashing----COAL

MKS3POSTMIX- mbo- ashing MKS3PREMIX MKS3LAS ashing — STRAW

Fig 4.7. Schematic illustration of the laboratory ashes produced from MKS1 experiment 3 fuels: South American high volatile bituminous coal and danish wheat.

MKS3POSTMIX was produced by first ashing the two fuels and thereafter mixing them in the appropiate proportions, based on heating value and ash content (Appendix F). MKS3PREMIX was produced by first mixing the fuels and subsequently ashing. All ashes were produced at 550°C to avoid large scale loss of volatile elements. All the ashes produced were particulate and was therefore suitable for CCSEM analysis. Selected results for the two mixed laboratory ashes are presented in Fig.4.8 and the full CCSEM results can be found in Frandsen et al. (1998).

BMKS3POSTMIX

30.0 •• □ MKS3PREMIX □ MKS3FA

25.0 • •

20.0 ••

15.0 ••

10.0 ■ •

Quartz lllite (K- Variuos K2S04 other Al-Si silicate sulphates

Fig.4.8. CCSEM data for mixed laboratory ashes from MKS1 experiment 3 on a category weight percent basis. For comparison is shown data for fly ash from the same experiment.

From Fig.4.8 it is seen that the two mixed laboratory ashes are quite alike when considering their contents of quartz and various alumino-silicates. However, it appears that while K-Ca silicate and KCI are present in the POSTMIX (separately ashed) these phases are not present in the PREMIX (ashed fuel mixtures), which on the other hand contains K2S04 and a higher content of

-44- other sulphates. In other words it appears that the K from the straw preferentially binds to sulphur from the coal when this is present. Additionally it is seen that K-AI silicate (illite) is lower in the PREMIX than in the POSTMIX. This is possibly because a significant part of the K present is bound to sulphate both in K2S04 and in the group of other sulphates, leaving less K available for reaction with alumino-silicates from the coal. Additionally, it appears that K-Ca silicates does only form when there are no significant alumino-silicates available for reaction (POSTMIX). This effect is substantiated by CCSEM analyses of fly ashes from co-firing experiments at MKS1 (MKS3FA and MKS4FA) in which no significant K-Ca silicates were observed (Appendix G).

When looking at bulk compositions based on CCSEM analyses, it is seen that the two mixed ashes are quite alike except for the fact that more S03 is present in the PREMIX whereas more Cl is present in the POSTMIX (Fig.4.9).

■MKS3POSTMIX

□MKS3PREMIX □MKS3FA

30 -

AI203 Fe203 CI207

Fig.4.9. Bulk composition (weight percent) calculated on the basis of CCSEM data for the mixed laboratory ashes.

The results of the ashing experiments indicate that whether the laboratory ash is formed by mixing the fuels before ashing or ashing them separately has only slight, but important, consequences for the resulting ash composition since reaction between straw K and coal S to form K2S04 occurs during the ashing. However, compared to fly ash from experiment 3, both ashing procedures yield compositions that are lower in Si02 and Al203 and higher Cl207 and S03 than the fly ash (Fig.4.9). The K20 content, however, is similar in the fly ash and in the laboratory ashes.

-45- 5. Summary and conclusions The SEM-study of samples from straw and straw+coal co-combustion have yielded a reference selection of representative images that will be useful for future comparison. The sample material encompassed potential fuels (wheat straw and grain), bottom ash, fly ash and deposits from straw combustion as well as fuels (coal and wheat straw), chars, bottom ash, fly ash and deposits from straw+coal co-combustion. Additionally, a variety of laboratory ashes were studied.

SEM and CCSEM analysis of the samples have given a broad view of the inorganic components of straw and of the distribution of elements between individual ash particles and deposits. The CCSEM technique does, however, not detect dispersed inorganic elements in biomass, so to get a more complete visualization of the distribution of inorganic elements additional analyses must be performed, for example progressive leaching. In contrast, the CCSEM technique is efficient in characterizing the distribution of elements in ash particles and between ash fractions and deposits.

The data for bottom and fly ashes have indicated that binding of potassium to silicates occurs to a significant extent. The silicates can either be in the form of alumino-silicates or quartz (in co ­ combustion) or as straw-derived amorphous silica (straw combustion). This process is important of two reasons. One is that potassium lowers the melting point of silica in the fly ash, potentially leading to troublesome deposits by particle impaction and sticking to heat transfer surfaces. The other is that the reaction between potassium and silica in the bottom ash binds part of the potassium meaning that it is not available for reaction with chlorine or sulphur to form KOI or K2S04. Both phases are potentially troublesome because they can condense on surfaces forming a sticky layer onto which fly ash particles can adhere and by inducing corrosion beneath the deposit.

The analyzed deposits from straw-fired CHP’s have shown that in the furnace condensation of KCI is important for formation of the initial layer, whereas building of the outer loose part of the deposit is dominated by particle impaction. Additionally, particle impaction is a more prominent process in the furnace compared to in the superheater region.

Potassium were also seen to react with sulphur preferentially to chlorine and silicate-compounds during low-temperature (550°C) laboratory ashing of mixtures of wheat straw and bituminous coal from MKS3. A similar effect was, however, not observed in full scale fly ashes from MKS3 in which potassium to a high extent have reacted with alumina silicates to form K-AI-silicates. In

-46- contrast potassium sulphates are present in deposits formed during co-combustion at MKS1 especially in the convective pass.

6. References Andersen, K.H., Frandsen, F.J., Hansen, P.F.B and Dam-Johansen, K. (1997). Full scale deposition trials at 150MWe PF-boiler co-firing coal and straw: summary of results. Presented at the Engineering Foundation Conference on Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7., 1997 Bryers, R.W. (1994). “Analysis of a suite of biomass samples. Foster Wheeler Development Corporation. FWC/FWDC/TR-94/03. Deer, W.A., Howie, R.A. and Zussman, J. (1966). ”An introduction to the rock-forming minerals.” Longman Group Ltd. Frandsen, F.J., Hansen, L.A., Sorensen, H.S. and Hjuler, K., 1998. EFP-95 project: Characterization of Biomasss Ashes - Data Compilation. CHEC Report No. 9804. Hansen, L.A., Frandsen, F.J., Dam-Johansen, K., Sorensen, H.S., Rosenberg, P. and Hjuler, K. (1997). “Ash Fusion and Deposit Formation at Straw Fired Boilers.” Proc. Eng. Found. Conf.: Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7,1997. Laursen, K. (1997a). “Characterization of Minerals in Coal and Interpretations of Ash Formation and Deposition in Pulverized Coal Fired Boilers.” Ph.D. Thesis, Geological Survey of Denmark and Greenland Report 1997/65. Laursen, K. (1997b). “Advanced Scanning Electron Microscope Analysis at GEUS.” Geological Survey of Denmark and Greenland Report 1997/1. Mgrchner, H. (1995) Mineral Nutrition of Higher Plants. 2 Ed., Academic Press (1995). Morey, G.W., Kracek, F.C. and Bowen, N.L.(1931). “The ternary system K20-Ca0-Si02.” Journal of the Society of Glass Technology, Vol.14, pp. 149-187. Glanders, B. and Steenari, B-M. (1995). “Characterization of ashes from wood and straw.” Biomass and Bioenergy, Vol.8, No.2, pp. 105-115. Steenari, B-M. and Langer, V. (1988). “Fasanalys av sintrade och osintrade halmaskor med och utan tillsats av kaolin respektive dolomit." Report OOK 88:12. Stenholm, M., Jensen, P.A. and Hald, P., (1996). “Biomasses brasndsels- og fyringskarakteristika. Fyringsforscg. EFP-93.” 1323/93-0015. (in Danish) Sorensen, H.S. Computer Controlled Scanning Electron Microscopy (CCSEM) analysis of straw ash Proc. Eng. Found. Conf.: Impact of Mineral Impurities in Solid Fuel Combustion, Kona, Hawaii, November 2-7,1997.

-47- Appendix F:

Various Notes on Apparatus, Equilibrium Calculations, Straw Leaching and Ash Mixing, and Samples from Coal/Straw Co-Firing.

F Appendix F

Contents

Pyrometer Design F-2

Electrical Resistivity Apparatus F-3

Effect of Leaching on Melting Behavior of Straw Lab. Ash F-4

HTLM of Samples from MKS1 F-6

HTLM of Coal and Straw Ash Mixtures F-7

Global Equilibrium Calculations F-9

by Klaus Hjuler, dk-TEKNIK Energy & Miljo

F-1 Pyrometer Design

A water cooled pyrometer with a total length of 5 meters was designed for sampling straw char from the burner zone during pulverized coal-straw co-combustion experiments. The maximal insertion depth was 3 meters, i.e. 2 meters was not subjected to heat from the burner. Accordingly, the pyrometer was constructed using standard Sandvik 253MA heat resistant steel.

The outer diameter is VA inch and the inner diameter is 3/8 inch. The gas velocity at the inlet was designed to be about 80 m/s at 1000°C, corresponding to a gas sampling flow of about 4.4 Nrrr/h.

The char was sampled on a 90 mm diameter filter made from sintered metallic (copper) spheres. The filter has a nominal pore diameter of 40 pm. A glass filter holder was constructed, Fig. F1, in which the filter was mounted using PTFE gaskets. The filter holder and the pyrometer was connected in-line, horisontally.

Conclusions from the tests: The pyrometer worked well for short periods of time, i.e. less than about 1A minute. After this the inner tube was blocked by a tar like substance. The pyrometer was then cleaned using pressurized air. The glass filter holder worked well and allowed the sampling of char to be observed visually.

m -

Fig. F1

F-2 Electrical Resistivity Apparatus

A simple apparatus was designed for the measurement of the electrical resistance of ashes vs. temperature in a controlled atmosphere at atmospheric pressure. The reactor was constructed on the basis of a quartz glass tube with a support for an alumina crucible. The ‘reactor 1 was placed vertically in a tube oven. The design is outlined in Fig. F2. About 0.5 g of sample is placed in an alumina oxide crucible between two Pt-Rh nets, acting as electrodes. The lower net is supported by porous ceramic paper, 2 mm in thickness, which allows gas flow through the sample. The crucible is placed in a quartz tube positioned in a vertical tube furnace (max. operating temperature 1340°C). Finally, an alumina oxide tube is placed loosely on the upper electrode in order to impose a constant pressure on the sandwich of electrodes and ash sample.

Glass connection

Dense Al.

Dense Al.

Pt-net

Porous Al.

Support

U

F-3 Effect of Leaching on Melting Behavior of Straw Lab. Ash

It is well known that main part of the potassium in straw can be leached by rain after harvesting, and full scale tests indicate that this decreases the deposition propensity of the straw ash. It was decided to perform a simple test of the rate of leaching, and to subject straw samples to leaching before ashing at 550°C. The melting behavior of ashes prepared from straw, leached straw, and mixtures of thereof were compared.

The leaching test was carried out on milled wheat straw, the max. particle size being about 0. 5.mm. The leaching was performed in a glass beaker at various intensities: boiling water for 30 and 15 min, and in stirred water at room temperature for 30,15,11, 9, 7, 5, 4, 3, 2, and 1 min. It was found that the concentration of dissolved potassium in the non-boiled liquid was the same after 1 min as after 30 min. The samples subjected to boiling water were filtered, and the potassium content of the eluate was determined. It was found that the potassium concentration in the eluate was lower than in the non-boiled liquid, probably due to capture of potassium in the filter cake e.g. by starch formed when boiling.

A straw sample from a moving grate fired, 6 MW heating plant (Hong) was divided in two samples, A and B. Sample A was milled and dryed. Sample B was milled, leached, and dryed. Afterwards six ash samples were prepared at 550°C from (percentages are weight percentage dry matter):

1. 100% A (unleached) 2. a mixture of 80% A and 20 % B 3: a mixture of 60% A and 40% B 4. a mixture of 40% A and 60% B 5. a mixture of 20% A and 80% B 6. 100% B (leached)

The resulting melting curves are shown in Fig. F3. Also shown for comparison are the melting curves from bottom ash, cyclone ash, and fly ash sampled from the Hong plant during combustion of the actual straw. Starting from above, it can be seen that the solid fraction of sample #6 is higher than of the other samples in the temperature range 600- 1200°C. Sample # 5 is very similar to # 6, but the solid fraction is significantly lower above 1050°C. Then, considering the curves below in Fig. F3, the behavior of the fly ash is typical of fly ashes from straw combustion due to a high content of KCI and K2S04, whereas the drop of the solid fraction at 800°C of bottom ash is untypical. The melting curves of the cyclone ash and samples # 1, 2, 3, and 4 are very similar.

From this study it can be concluded that the content of potassium in the straw influences the melting behavior of the ash, but probably as much as 80% of the straw has to be leached in order to lower the melt fraction of the laboratory ash significantly in the range 600-1200°C.

F-4 Hang Straw Fired Heating Plant (LA, BA, FA, CA)

Temperature (deg.C)

Fig. F3.

Studstrupvaerket (FA, BA, Dep.)

FA3 XFA5 XFA3 XFA3 XFA5 XBA3 e so + XBA5 BAS "o XBA4 (0 40-' xE3Dep1 xE3Dep3 30 xE3Dep2 X E3Dep3

20 * -

10..

700 750 800 850 900 950 1000 1050 1100 1150 1200

Temperature (deg.C)

Fig. F4.

F-5 HTLM of Samples from MKS1

During full-scale experiments with coal-straw pf co-combustion various samples were taken, among others fly ash (from the electrostatic separator), bottom ash (sampled dry), and deposits from deposition probes.

Fig. F4 shows the melting curves of selected samples from these experiments:

FAS: fly ash from a straw share of 20% (energy basis) FAS: fly ash from coal firing (no straw added) BAS: bottom ash from a straw share of 20% BA4: bottom ash from a straw share of 10% BAS: bottom ash from coal firing (no straw added) ESDepI : deposit from a straw share of 20%, high temperature probe E3Dep2: deposit from a straw share of 20%, medium temperature probe ESDepS: deposit from a straw share of 20%, low temperature probe

As can be seen from Fig. F4, all samples except BA4, ESDepI , and E3Dep2 have been run two or three times. From these runs it can be concluded, that sample ESDepS is the only one that significantly separates from the others in the temperature range 700-1200°C by having a higher melt fraction. At about 1150°C, heavy evaporation from the sample takes place which deposited on the top glass seal of the heating stage, resulting in significant lowering of the light transparency. The automatic routine for threshold calculation attemps to correct for this, and thats why the solid fraction suddenly goes above 100%. This behavior was not observed in any of the other samples. The deposit on the top glass seal was subsequently analysed by CCS EM, and it was found that the deposit was pure K2S04.

It can be concluded that a 20% straw share on an energy basis does not affect the melting behavior of bottom and fly ashes significantly below 1200°C. However, the low temperature deposit has a higher melt fraction below 1200°C and is enriched on K2S04, which evaporates at around 1100°C.

F-6 HTLM of Coal and Straw Mixtures

In order to compare the melting behavior of ashes from full scale co-firing of coal and straw with laboratory prepared mixed ashes, four ash samples were prepared at 550°C from a ‘premix ’ of (on a heat value basis):

1. 80% coal and 20% straw (as full scale exp. 3) 2.507 g dry coal/g dry straw 2. 60% coal and 40% straw 0.940 do. 3. 40% coal and 60% straw 0.418 do. 4. 20% coal and 80% straw 0.157 do.

Heat value of dry coal: 27.07 MJ/kg, of dry straw: 16.97 MJ/kg.

Moreover, the following ash samples were prepared:

5. 100% coal (from full scale exp. 3) ashed at 815°C 6. 100% straw (from full scale exp. 3) ashed at 550°C 7. 100% straw (from full scale exp. 4) ashed at 550°C 8. Leached straw (from full scale exp. 3) ashed at 550°C 9. Leached straw (from full scale exp. 4) ashed at 550°C 10. ‘Postmix ’ of straw ash and coal ash in the ratio as it should be present in sample # 1, i.e. mixture prepared after ashing from samples # 5 and # 6. The ash content of the coal and the straw was 13.5% a.r. and 6.2% on a dry basis, respectively. The postmix was blended from 0.378 g of coal ash and 0.062 g of straw ash.

The resulting melting curves are shown in Fig. F5. Starting from above, the coal ash (# 5) has the lowest melt fraction in the temperature range 600-1200°C. Admixture of straw ash increases the melt fraction in this range. The behavior of samples # 10 and #1 (reproduced twice) do not differ significantly. That is, apparently the ash mixture may be prepared as an ashed fuel mixture (premix) or as a blend of individually ashed fuels (postmix) with similar result. The behavior of samples #1,2, and 3 do not differ either. That is, the blends with 80, 60, and 40% coal result in ashes of comparable melting behavior. The behavior of the 20% coal blend (# 4), however, differ significantly from the others and is more like the pure straw characteristic (#6). The effect of leaching this straw (# 8) is somewhat surprising, because the melt fraction of the ash is higher than that of the unleached in the range 875- 975°C. In the range 975-1 200°C, the melt fraction does not increase any further. A similar behavior is not observed when comparing straw ash (# 7) and leached straw ash (# 9) from full scale experiment 4 (10% straw on a heat value basis). In this case, the melt fraction of the leached straw is significantly lower in the range 600-1200°C. Furthermore, the melting behaviors of the two straw ashes # 6 and 7 are significantly different.

It can be concluded that the melt fraction is not proportional to the content of potassium in the ash. The admixture of straw ash increases the melt fraction, and the behavior of straw ash is approached at more than 80% straw on an energy basis.

F-7 100

20 Solid Fraction (%) 30 70 80 90 10 o4- 600 ' — -a- -a- x , x x a x x x x x x x o ■: - .

—

Straw+Coal Straw+Coal Straw+Coal Washed Washed Washed Straw* Straw Coal Coal Straw Straw Straw Straw Straw Straw Straw Straw

LA3 LAS LA3 LA4 LAS+Coal LA3+Coal LA4 LAS LAS LA4(1) LAS

Coal

— Straw Straw Straw

700

(ISO540)

t (2) (3) (2) (ISO540) (ISO540) (1) —

LA LA LA LA

LA4 LAS LAS 60/40 40/60 20/80 80/20

LAS LAS

(2) (1)

(2) (1)

Studstrup Fig. F-8

F5.

LA — 1000 i — — 1100 I ------

1200 Global Equilibrium Calculations

Data from straw grate firing experiments at Haslev and Slagelse CHP were used as input for chemical equilibrium calculations in order to compare calculated and measured emissions and to obtain a better understanding of the chemistry. The calculations were performed using EquiTherm ver. 4.02.

Detailed bulk chemical composition of the straw, the combustion zone temperature, and the air/fuel ratio were the input data. Seven different straw qualities were fired. The input data are listed in Table F1. The programme requires Cl to be input as a stable specie such as NH4CI at the fuel input temperature, that is, some N and H were subtracted from the original fuel content. This does not influence the results. Due to the uncertainty of measuring the (not uniform) combustion chamber temperature and the (not uniform) air to fuel ratio, estimated values were applied. Measurements indicate that the combustion temperature in the SIS experiment increased about 100°C. In this case rape straw was fired, whereas wheat or barley was fired in the other experiments.

Tables F2, F3, and F4 show some of the results obtained when K2S04 was omitted from the list of possible stable components in the ash phase. When K2S04 was included, the calculated sulphur capture was much greater than measured due to the stability of this specie at the actual temperature. No K2S04 was found by CCSEM analysis of the bottom ashes from Slagelse and Haslev.

Parameter Experiment % w/w a.r. SI1 SIS SI6 SI7 SIS Ha1 Ha2 Humidity 15.0 11.6 13.2 16.1 14.0 15.1 13.5 C 39.3 39.3 39.1 39.3 39.1 41.1 40.5 H 5.19 5.37 5.01 5.09 5.07 4.57 5.15 N 0.38 0.29 0.01 0.35 0.39 0.40 0.22 S 0.03 0.02 0.05 0.02 0.37 0.01 0.03

NH4CI 0.08 0.42 1.13 0.18 0.42 0.38 0.68 Si 1.67 0.47 1.53 1.30 0.14 .82 0.77 Al 0.02 0.01 0.01 0.05 0.01 0.01 0.01 Fe 0.01 0.01 0.01 0.03 0.01 0.01 0.01 Ca 0.40 0.37 0.36 0.38 1.47 0.28 0.27 Mg 0.04 0.03 0.05 0.05 0.05 0.04 0.04 Na 0.01 0.05 0.01 0.02 0.09 0.01 0.01 K 0.34 1.53 1.83 0.54 1.47 0.58 1.08 P 0.02 0.03 0.06 0.06 0.11 0.05 0.03

Air/Fuel, Nm'Vkg 5.8 5.8 5.8 5.8 5.8 5.8 5.8 Comb, temp., °C 800 800 800 800 900 800 800

Table F1. Input parameters and data for equilibrium calculations.

F-9 Parameter Experiment % w/w in ash SI1 SIS SI6 SI7 SIS Ha1 Ha2

Ca 2Si04 - 22.4 - - 18.4 - -

CaSi0 3 23.6 - 16.3 26.2 - 29.6 23.2

CaS0 4 1.8 2.6 3.1 1.0 33.8 - 2.6

CaO - - - - 17.9 - -

K2Si03 - 52.6 - - - - -

K2Si205 16.9 - 66.4 27.5 - 43.1 71.9

KAISi308 4.5 - 0.9 12.7 - 3.5 1.4

k2co3 - 16.4 - - 28.9 - -

Si02 52.4 - 13.0 30.6 - 23.3 0.6

Sum 99.2 94.0 99.7 98.0 99.0 99.5 99.7

Table F2. Some major species in the ash phase.

Parameter Experiment ppm in gas SI1 SIS SI6 SI7 SIS Ha1 Ha2 KCI 22.4 205 303 49.3 213 112 193 5 CM 0.35 29.2 63.4 - 5.5 8.6 25.7 - KOH 0.05 37.6 - 0.05 405 0.05 0.05 NaCI 6.8 7.0 14.5 15.0 47.0 11.8 10.3

NaOH 0.01 1.2 - 0.01 88.0 - - HCI 21.7 0.26 288 48.0 0.09 104 185 S02 9.4 0.02 9.2 9.4 0.04 0.91 9.6 NO 37 38 37 37 89 37 36

Table F3. Species present in the gas phase at ppm-level.

Distribution Experiment % in gas SI1 SIS SI6 SI7 SIS Ha1 Ha2 K 7.7 22.3 26.5 11.1 48.2 24.9 25.5 S 34.6 0.10 20.4 52.0 0.10 100 36.3 Na 46.2 11.2 100 51.9 100 100 100 +100°C K 8.6 39.9 31.7 12.6 100 31.9 31.7 S 100 3.5 100 100 4.9 100 100 Na 100 36.0 100 100 100 100 100

Table F4. Calculated percentages of the fuel elements K, S, and Na present in the gas phase. The rest is captured in the ash phase. +100°C means that the input combustion temperature was increased 100°C to study the effect on the distribution. Appendix G:

Fusion, Sintering, and CCSEM Analyses of Ashes from Coal/Straw Co- Firing.

G EFP-95 Project: Characterisation of Ashes from Biofuels

Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing

Lone A. Hansen Henning S. Sorensen Flemming J. Frandsen

1998 Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 1

Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing

This report presents the results of characterisation of ashes collected at a PF- fired boiler co-fired with straw and coal in varying energy ratios. As for the ashes from the two straw fired boilers (App. D), the melting behaviour of the ashes as determined by STA will be interpreted in terms of their chemical and mineralogicai composition as determined by CCSEM, and compared to results from standard ash fusion tests. The ash melting behaviour results will be used for evaluation of sintering test results and additionally be used as input to a simple mathematical model estimating ash deposit formation rates, the results of which will be compared to the actual ash deposit formation rates measured in the plant.

1 Midtkraft Power Station, Unit 1 In 1996, a very comprehensive measurement campaign was conducted at the PF-fired unit 1 at the Studstrup Power Station. The main purpose of the campaign was to evaluate whether the concept of coal-straw co-firing was competitive to other technologies, with respect to economical and technical feasibility. The measurement programme included experiments of varying straw share (0-20% on an energy basis) and load (50-100%) to demonstrate plant operational data and operating costs when co-firing coal and straw, as well as experiments to clarity the fuel and process chemistry. The programme included, among other things, analyses of fuel, fly ash, bottom ash, deposits and flue gas. Corrosion experiments were conducted and a number of in-situ gas, temperature and aerosol measurements were performed. In this report, the focus will be drawn to a thorough investigation of three sets of fly ash, bottom ash and deposits. For details concerning the measurement campaigns are referred to Andersen et al., (1996) and Hansen et al., (1996).

1.1 Description of the Plant and Measurement Locations The Studstrup Power Station Unit 1 (MKS1) is a 150 MWe wall-fired unit with 12 burners arranged in three burner rows, and was first commissioned in 1968. For the coal/straw co-firing experiments, the four burners in the middle burner row was converted to coal/straw co-firing with up to 50% straw (energy basis). The platen superheater and the secondary superheater are located in the top of the boiler, the reheater is located at the entrance to the second pass, further into which the primary superheater is located. In Table 1, the major plant data are given, and Figure 1 shows a schematic drawing of the boiler with indication of sampling positions. Deposits described in this report were collected at position 1, 2, and 3. Flue gas temperatures at the three positions varied between 1220 and 920°C, as a function of position and load (Andersen, 1997c). Deposition, in-situ experiments and mass balance closures were performed Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 2

A B C

HEAT TRANSFER SURFACES

A: Platens Super Heaters (1st/2nd pass) B: Secondary Super Heater C: Reheater D: Upper Primary Super Heater E: Lower Primary super Heater F: Economiser

Sampling positions

COALN Z COAL & STRAW COAL

Fig. 1: Schematic drawing of the MKS1 boiler (modified from Andersen, 1996)

Table 1: Plant and operational data for MKS1 (Hansen et al., 1996)

Unit / electric capacity (th/e) 380 MW/150 MW

Straw capacity (energy basis) Max. 20 % ~ 5.55 kg/sec

Firing Method Wall fired

Burners 12 total, 4 coal/straw burners

Bottom ash condition Dry

Steam data: superheater 540°C, 143 bar

reheater 540°C, 41 bar

Steam flow 139 kg/sec

Furnace geometry: height 29.5 m

width 11.0 m

depth 8.1 m at several operational conditions, reflected in a number of experiments of which some are given in Table 2. In this chapter, focus will be drawn to ashes collected in experiments 3, 4, and 5, and for the sintering experiments (section 3) additionally the fly ashes from experiment 1 and 2 have been examined. Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 3

Table 2: Experiment no. as a function of operational conditions

Exp. no. 1 2 3 4 5

Straw share 20 20 20 10 0

Loadr%l 50 75 100 100 100

1.2 Fuel Composition The coal burned in experiments 1-5 was a South American high volatile bituminous coal, commonly used for power production in Denmark. The straw burned was primarily Danish wheat. Chemical compositions of fuels and the laboratory ashes are shown in Tables 3 and 4. For the straw burned in experiments 3 and 4, a total of four samples were ashed and analysed. A quite high deviation was found between analysis results due to the highly inhomogeneous nature of straw, and for this reason both a minimum and a maximum are given for the ash analyses. Typically, the straw is characterised by a lower content of carbon, sulphur, and ash, and a higher content of volatiles than the coal; for the ashes, the content of Al203 and Fe 203 is much lower, and the contents of CaO, K20, Cl, S03, and P205 is higher in the straw ash compared to the coal ash.

Table 3: Chemical composition of fuels [%w/w] (modified from Hansen, 1997)

C O H S N Ash Water Vol. LHV1

Coal 2 61.3 6.2 4.2 1.0 1.3 13.8 10.5 31.6 24.0 00 0 1

Straw 3 45.6 6.0 0.1 0.6 6.6 6.4 17.0 1 [MJ/kg] 2 Basis: As received 3 Dry basis

Table 4: Chemical composition of fuel ashes [%w/w] (modified from Hansen, 1997)

Si02 Al203 P205 so 3 CaO Fe 203 MgO Na 20 K20 Cl

Coal 59.8 19.1 0.2 2.1 2.0 8.1 1.7 0.6 2.2 <0.1

Straw 19.7 0.24 2.45 3.40 6.35 0.13 1.50 0.29 28.7 4.55 min

max 38.9 0.52 3.00 5.00 8.45 0.19 1.90 1.00 34.60 7.06

The mineral contents of the coal utilised in experiment 3, 4, and 5 were analysed by Computer Controlled Scanning Electron Microscopy, CCSEM. The results from the analysis are shown in Table 5 on a mineral category basis (%w/w). Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 4

Table 5 Mineral composition (% (w/w)) of coal

Mineral Category MKS3C MKS4C MKS5C

Quartz 23.8 15.3 23.5

Illite (K-AI-Si) 13.7 16.1 13.7

Montmorrilonite (Ca-AI-Si) 14.5 9.8 10.5

Kaolinite (Al-Si) 11.2 26.5 18.0 Aluminasilicate 8.3 5.1 7.4

Pyrite 2.9 5.6 5.1

Si rich 1.7 1.1 0.6 Fe-AI silicate 2.5 1.5 0.6

Clay/pyrite 1.7 0.8 1.0 Gypsum 1.7 2.6 2.1

Na-AI silicate 1.5 1.4 1.8 Dolomite 1.2 1.0 1.7 Unclassified silicate 6.1 5.8 4.4

Unclassified 3.9 2.4 3.2 Sum 94.9 94.7 93.6

The minerals of the coal are dominated by quartz and the clay minerals illite, montmorrilonite, and kaolinite, which makes up 63 - 68 % (w/w) in total. Silicates constitute 80 - 84 % (w/w) of the ash, whereas the pyrite content varies between 2.9 and 5.6 % (w/w). The quite large variations in CCSEM results may be ascribed to differences in the coal mineral compositions as sampled in the three experiments.

2 Ash Chemistry and Fusion

2.1 Samples The samples investigated include fly ash, bottom ash and upstream deposits from positions 1, 2, and 3, from respectively experiment 3, 4, and 5. Furthermore the fly ashes from experiment 1 and 2 have been examined for melting behaviour since they have been used for sintering tests. Upstream deposits are termed by the following code: MKS‘x”y’-‘z’, where MKS tells that the deposit was collected at the MKS1 boiler, ‘x’ indicates experiment number (3,4, or 5), y indicates position number, and ‘z’ indicates the metal temperature at the probe: T indicates low temperature (540°C) and ‘2’ indicates high temperature (580°C). All deposits investigated in this project have been Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 5 exposed for 18 hours in the boiler. E.g. deposit MKS32-2 means the deposit collected during 18 hours exposure at position 2 during experiment 3 with a probe metal temperature of 580°C.

2.2 Chemical Composition and Association of Elements in Ashes Bulk chemical analysis of the deposits has only been carried out for three deposits: deposit MKS32-1, MKS42-1, and MKS51-2. The reason for this is the very small amount of material collected for the other deposit measurements. The chemical analyses for the three deposits and the three fly ashes and bottom ashes from experiments 3,4, and 5 are shown in Table 6. Six deposits were analysed by Scanning Electron Microscopy Point Counting, SEMPC: MKS31-1, MKS32-1, MKS33-1, MKS41- 1, MKS42-1, and MKS51-1. The SEMPC analyses were performed on the upstream part of embedded, cross sectioned, and polished deposits (App. E). The chemical compositions as determined by SEMPC are shown in Table 7.

Table 6 : Chemical composition of ashes [%(w/w)] from experiments 3,4, and 5 (modified from Hansen, 1997)

Si02 ai 2o3 Fe 203 CaO MgO Na zO k 2o so 3 p 2o5 Cl MKS3FA1) 1.03 0.97 1.13 1.93 1.33 1.69 2.80 1.44 3.13 >40

MKS4FA1> 1.00 0.86 0.98 1.53 1.00 1.12 1.70 1.12 3.07 >20

MKS5FA1) 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1

MKS3BA1) 0.98 0.93 0.81 2.26 1.38 1.62 3.63 1.80 7.58 >110

MKS4BA1) 1.02 0.95 0.84 1.68 1.13 1.17 2.25 1.90 5.42 >100

MKS5BA1) 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1

MKS32-1 41.4 8.2 11.6 13.3 5.0 0.96 8.3 4.5 3.9 -

MKS42-1 32.0 5.9 16.8 17.7 6.9 0.61 5.3 - 6.6 -

MKS51-2 29.5 10.3 45.1 6.6 3.5 0.25 0.65 2.5 0.7 _ 1) Due to confidentiality of the data, only relative compositions are given Nomenclature: MKS3FA: Fly ash from exp. 3; MKS3BA: Bottom ash from exp. 3

Comparing the results for the fly ashes, it is seen, that the ashes are quite alike on an oxide basis. Systematic differences are seen, though, for the contents of CaO, Na 20, K20, S03, P205, and Cl, which all increase between 44 and 213% (except for chlorine, the increase of which is very large due to the very low content of chlorine in the ashes from coal combustion) in the fly ash as a function of straw fraction in the fuel (but still for MKS3FA these components constitutes only 10.4 % (w/w) in total). The observed trend correlates with the fuel ash compositions, where the above mentioned components are all found in larger amounts in the straw ash compared to the coal ash. The same trend is observed when comparing analyses for the bottom ashes: The Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 6

Table 7: Chemical compositions of deposits as determined by SEMPC

Si02 Al203 Fe 203 CaO MgO Na 20 K20 so 3 P2O5 Cl

MKS31-1 47.6 9.9 10.0 11.4 3.2 1.3 8.1 2.5 4.1 0.2

MKS32-1 46.7 7.5 8.1 15.3 4.1 1.2 6.3 3.5 5.5 0.1

MKS33-1 38.4 18.0 5.2 1.9 1.3 1.6 14.0 16.3 1.7 0.2

MKS41-1 42.3 9.3 21.3 10.4 4.0 1.2 4.1 2.5 2.8 0.2

MKS42-1 41.7 7.1 12.0 15.2 4.2 1.3 6.6 3.1 6.3 0.2

MKS51-1 38.7 10.7 32.0 7.4 3.9 1.5 0.8 1.2 0.8 0.1 overall composition does not differ significantly, but the contents of CaO, Na zO, K20, S03, P205, and Cl all increase (62 - 658%) as a function of straw fraction in the fuel (still < 12% w/w). The (surprisingly) high Cl content in MKS3BA and MKS4BA is supposedly caused by a relatively high content of unburned carbon in the ash (Hansen, 1997).

For the deposit compositions, larger differences are seen. When comparing SEMPC bulk chemical analyses for deposits collected in position 1 from experiment 3, 4, and 5, it is seen that the content of Si02, CaO, K20 and P205 increases with increasing straw share (from experiment 5 to 3); whereas the content of Fe 203 decreases, probably due to dilution. As for the fly ash analyses, this is not surprising, since the data reflects the change in fuel ash composition, except for Si02. Comparing results for different positions (pos. 1 vs. pos. 2), it is seen that the content of Si02, Al203, and Fe 203 generally decreases, whereas the content of CaO, MgO and S03 increases from position 1 to position 2. The results from position 3 in experiment 3 shows that the trend continues with decreasing Si02 and Fe 203 content in the deposit relative to position 2. KzO and S03 increases drastically whereas P205 and CaO decreases. These data mainly reflect the increasing content of potassium sulphate in the deposit at position 3.

Comparing the chemical composition data for deposits to those for fly ash, it is seen that, generally, the deposits are depleted in Si02 and Al203, and enriched in Fe 203, CaO, MgO, Na 20, S03, P205, and Cl. For K20, the partly straw fired experiments show deposit enrichment, whereas in the coal-fired experiment deposits show K20 depletion.

The mineralogical characteristics of the samples as determined by CCSEM, is shown in Figures 2, 3, and 4. The full data set is provided in Frandsen et al., 1998. In Fig. 2, the compositions of laboratory prepared ashes of the two fuels are shown, and a distinct difference between fuel ash minerals is clearly seen. The coal ash is characterised by high contents of quartz, K-AI-silicates, Ca-AI-silicates, Fe-AI-silicates and alumino silicates, whereas the dominant minerals in the straw ash are K-silicate, Ca-silicate and KCI. For the straw ash also the high quantity of unclassified particles should be noticed. These consist predominantly of K20 (~35 %), CaO (-20%), Si02 (-15%), Cl207 (-11%), and S03 (-8%) (Frandsen et al., 1998). Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 7

For the fly ashes (Fig. 3), the ashes from experiment 3 (MKS3FA) and 4 (MKS4FA) are quite alike mineralogically and they distinguish themselves from the fly ash from experiment 5 (MKS5FA). Comparing MKS3FA and MKS4FA to MKS5FA, the latter is seen to be characterised by: a relatively low content of quartz and K-AI-silicate and a relatively high content of Ca-AI-silicate, Fe-AI-silicate and Al-silicates. Furthermore it is seen that for all the mentioned categories, the trend is, that if a larger amount of a given species is present in MKS3FA and MKS4FA compared to MKS5FA, this species content is higher in MKS3FA compared to MKS4FA, and vice versa.These data show

40

35 ■ MKS3-coal lab. ash 30 □ MKS3-straw lab. ash 25

20

15

10

5 0

Fig. 2: Chemical composition of laboratory ashes as determined by CCSEM

35

30 DMKS3FA 25 □ MKS4FA I 20 ■ MKS5FA

o O

AJJl.rU.n. i lj (%: £ £ £ £ m o CO CO CO o O E S S ‘to "to to .2o w *a *o CO o ? 9 XCD c E 42 oCO CO 3 3 E c to

Fig. 3: Chemical composition of fly ashes as determined by CCSEM Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 8 that, when K is present, K-AI-silicate is the stable form of the different alumina-silicates. This means that when potassium is available for reaction (i.e. when straw is burned) a large fraction of all the present types of alumina-silicates react with it, increasing the amount of K-AI-silicates and decreasing the amounts of all other alumina-silicates in the fly ash.

Fly ash from experiments 3, 4, and 5 are plotted in three-dimensional triangular diagrams of Si02, K20 and Al203 in Fig. 4. The height of the columns are proportional to the relative weight percent of particles in the compositional range defined at the basal plane. In pure coal combustion (MKS5FA), two major peaks are noticed, one composed of virtually pure Si02 and one located at the Si02-Al203 line. The first peak represents quartz-derived particles whereas the second represents clay-derived particles. In both cases of straw addition (MKS4FA and MKS3FA) there is a relative increase in the Si02 peak, indicating that straw derived Si02 is preferentially incorporated in the fly ash. Noticeable the clay derived peaks in experiments 3 and 4 are dragged towards the K20 apex of the diagram underlining that K20 from the straw is incorporated in coal derived aiumino-silicates.

MKS3FA MKS4FA MKS5FA

A1203 A1203 A1203 Figure 4: Fly ash from experiment 3,4, and 5 plotted in triangular diagrams of Si02- K20-Al203 Concerning the quartz content, it does not, at a first glance, seem logical that it increases as a function of straw fraction, since more quartz is present in the coal laboratory ash compared to the straw laboratory ash (Fig. 2). The explanation to this phenomenon is that during the ashing of the straw in the laboratory, the inorganic matter in the straw undergoes chemical reaction. A CCSEM analysis of the corresponding straw (App. E) revealed a quartz content of 50-70% (w/w). During the ashing proces (8 h.) the quartz presumably reacts with the atomically dispersed K and Ca in the straw, which is not detected in the CCSEM analysis. The reaction results in a high content of K- and Ca-silicates in the straw laboratory ash. However, bulk chemical analysis of the fuels showed that the Si02 content of the straw was lower than that of coal (Table 4), indicating that this effect is not on its own responsible for the Si02 increase in the fly ash with straw addition. It is believed that the Si02 from the straw preferentially ends up in the fly ash since it upon dehydration and ashing is in the form of minute flakelike particles which are easily carried by the flue gas (App. E). Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 9

For the bottom ashes (Fig. 5) it is seen, that the contents of quartz are alike in all three ashes. As was the case for the fly ashes, the amount of K-AI-silicate is very low in MKS5BA compared to MKS4BA and MKS3BA, which is due to the low content of K20 in the coal ash. In contrast, the amounts of alumina-silicates, Ca-AI-silicate and Fe-AI- silicates are oppositely higher in MKS5BA compared to the two other bottom ashes. For the category ‘silicate rich' the amount increases sharply as a function of straw share. The average composition of the category tells that the category is principally composed of Si02 (80%), AI2Os (12%), K20 (4%), and CaO (3%). The increasing amount of this material type with straw share reflects basically the larger content of Si in the straw.

□ MKS3BA ? 20 QMKS4BA ■ MKS5BA

o 10-

Fig. 5: Chemical composition of bottom ashes as determined by CCSEM

SEMPC results are given for the examined deposits in Table 8. It is seen that the dominant species present in the deposits are mixed silicates, iron oxide, Fe-silicates, quartz, K-silicates, Ca-silicates (only in exp. 3 and 4, position 1 & 2), Fe-AI-silicates (only in exp. 5), iliite derived species (primarily exp. 3), and mixed phosphates (only for pos. 2 and 3). Comparing MKS31-1, MKS41-1, and MKS51-1, it is seen that the quantities of Fe-silicates and Fe oxide decrease with increasing straw fraction, which may reflect the low Fe content in the straw. Some of the Fe-oxide may originate from the oxidised surface of the probe, since a small degree of overlapping of the metal was unavoidable during the SEMPC analysis. However, this effect was minimised by omitting analysis points containing Cr203 (probe metal derived) from the data treatment (App. E). The quantity of mixed silicates and the total sum of silicate species increase as the straw fraction increases. No final explanation for this is known, but it may be due to the higher degree of association of K with Si in experiment 3 and 4, which may Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 10

Table 8: Results from SEMPC for deposits [%(w/w)] (from Hjuler et al., 1997) Category MKS31-1 MKS41-1 MKS51-1 MKS32-1 MKS42-1 MKS33-1

quartz 8 4 4 10 4 2

K silicate 3 - - - 1 -

Ca silicate 8 8 - 16 11 -

Fe silicate 3 10 19 3 7 -

Fe-AI silicate - - 13 - - -

illite derived 7 2 1 6 2 8

kaolinite - - - - 1 - derived mixed silicate 46 27 19 39 44 24

iron oxide 9 16 27 5 9 -

Ca phosphate 2 1 1 2 2 -

mixed - - - 3 2 1 phosphate

K sulfate 2 - - 1 1 6

K-Ca sulfate 1 - - 2 1 4

mixed - - - - - 6 sulphates unclassified 8 9 12 10 14 49 increase the collection efficiency for these (Si-rich) particles due to the fluxing effect of K.

Comparing deposit species for the same experiment but for different positions, it is seen that the content of Ca-silicates increases from position 1 to position 2. The content of Fe oxide decreases from position 1 to position 2 in experiments 3 and 4. The content of K- and Ca-sulphates increases slightly from position 1 to position 2, and phosphates are only present at position 2 and 3. These differences are all assumed to be caused by the lower gas temperature (1100-1125°C) at position 2 and 3 compared to position 1 (1225-1275°C). In position 3 in experiment 3 (MKS33-1) a high content of sulphates and a high proportion of unclassified is seen. The average composition of the unclassified category is 37% Si02, 17% Al p 3 14% K p, and 18% SO 6 so these analyses probably represent points in border regions between alumino-silicates and sulphates.

SEMPC results for deposits in position 1 in experiments 3, 4, and 5 are shown in Fig. 6. Similarly to the trend observed for fly ashes from the same experiments, there is a Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 11 trend of increasing K20 with increasing straw addition. The trend is, however, more marked than in the case of fly ash, i.e. the compositions extend further towards the Ka O apex. Two reasons can account for this: 1) the alumino-silicate particles may continue to react with K20 after deposition and 2) the more K20 rich compositions may actually represent border regions between the thin rims of potassium sulphate and alumino silicates, rendering signals from both phases as noted above.

Fig. 6: SEMPC results for upstream deposits from position 1 shown in three-dimensional triangular diagrams of Si02-K20-AI203 To further illustrate the incorporation of K20 in alumino silicate particles, Fig. 7 shows changes in the composition of Si02-K20-Al203 rich particles when going from coal over fly ash to position 1 deposits in experiment 3. The change from coal minerals to fly ash indicates that K20 from the straw reacts with coal derived alumina silicate particles dragging the clay-derived peak towards the K20 apex. The K20 content increases further from fly ash to position 1 deposit as a response of further reaction with K20 and due to the presence of thin potassium sulphate rims on fly ash particles. Such rims are best observed on SEM X-ray maps, as exemplified in Fig. 8, that shows part of the upstream deposit of 3B2LA. Thin rims rich in potassium and sulphur are seen to coat many of the rounded fly ash particles.

Fig. 7: SEMPC results of coal and corresponding fly ash and deposit from experiment 3 shown in triangular diagrams 2.3 Ash Fusion Quantification by STA All ashes from the MKS experiments consist solely of silicates, i.e. a variety of different metal silicates and quartz. The determination of melting curves for these ashes has Fusion,

Sintering, BSE

and

CCSEM

Analyses

of

Ashes

from

Coal/Straw

Co-Firing

Figure 8: x-ray mapping of MKS32-1 12 Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 13 been based only on a simple comparison of areas below the DSC curve, since it is not possible to determine specific peaks in the DSC signal. It has thus not been possible to determine whichspecies melt in which temperature ranges and thereby obtain useful melting enthalpies. The melt fractions obtained at the maximum temperature has been determined by combination of visual observations and SEM backscatter images of the surface structure.

2.3.1 Fly Ash Melting A typical set of STA curves are shown in Fig. 9 for MKS3FA. A typical feature is the long lasting DSC peak occurring at temperatures between 20°C to 600°C, which is believed to represent decomposition of clay minerals in the ash. Above those temperatures, nothing seems to occur until the DSC curve starts increasing significantly at 1060°C, which represents the melting onset. The DSC curve shows first one broad peak (1060 - 1265°C) and then another one which is not quite ended at the experiment termination at 1390°C.

0.4 <"

Melting onset1 I 1 0.2

o £ 0-1

— -96

Fig. 9: STA curves for MKS3FA Melting curves for the fly ashes from experiment 3, 4 and 5 are shown in Fig. 10. For each ash, several melting curves are shown to illustrate the method repeatability. As

0.6 -

0.4-

0.2-

0 400 800 1200 Temperature [°C] Fig. 10: Melting curves for fly ashes from experiment 3, 4, and 5 Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 14 can be seen, the melting onset is approximately 1180°C and differs within 30°C for fly ashes from experiment 4 and 5, whereas for experiment 3 the melting onset is app. 1040°C, with a difference obtained for repeating runs as high as 40°C. Neither of these repeatabilities is as good as the ones obtained for melting onset for fly ashes from straw combustion (Hansen, 1997). The lower repeatability for the silicate rich ashes is due to the very slow melting start for these ashes compared to the abrupt melting start for ashes containing a significant fraction of salts; the slow start renders a precise onset determination. For the remaining curve behaviour, MKS4FA and MKS5FA melts continuously and reach melt fractions of respectively 70 and 85% from 1200 to 1400°C. The repeatability is quite good for the MKS4FA and MKS5FA curves (< 5% melt), whereas for MKS3FA, the deviation between curves is quite high, resulting in a general deviation on less than 15% melt for a given temperature, but with as high as 25% difference at 1280-1 330°C. Nevertheless, all MKS3FA curves reflect a DSC curve with two significant peaks, one occurring from app. 1040-1250°C, and the next above 1340°C. The difference between curves reflects that the ratio between the areas/energies related to each of the two peaks vary for the different experiments. This indicates that the ratio between species melting in respectively the lower and the higher temperature range vary. The reason for the much higher uncertainty for MKS3FA compared to MKS4FA and MKS5FA is therefore believed to be due to higher degree of inhomogeneity of MKS3FA compared to MKS4FA and MKS5FA. It could be reasonably assumed that the co-firing experiments would produce a more inhomogeneous fly ash than one produced during combustion of one fuel; however, this does not explain why MKS3FA is so much more inhomogeneous than MKS4FA.

The curves also illustrate that MKS4FA and MKS5FA show quite similar melting behaviours, and that MKS3FA starts melting at lower temperatures and shows higher melt fractions at lower temperatures (1050-1300°C) than do MKS4FA and MKS5FA. Interpreting these melting curves based on the ash chemistry, this does not immediately seem to be logical. The chemical compositions of MKS3FA and MKS4FA (Fig. 3) are quite alike and contain higher amounts of the fluxing potassium than MKS5FA. Thus it would be believed that both MKS3FA and MKS4FA would start melting at approximately the same temperature and show larger melt fractions at lower temperatures than MKS5FA. Never the less, the relative behaviour of MKS3FA compared to MKS4FA and MKS5FA is as expected: For temperatures above 1200°, MKS3FA shows larger melt fractions than MKS4FA, which slow slightly higher melt fractions than MKS5FA.

A good theoretical estimation of melting behaviour would require use of multi-phase, multi-component equilibrium calculations. In this chapter, a simpler evaluation is made based on phase diagrams. The investigated fly ashes are shown in the phase diagram giving the best ternary representation in Fig. 11, in which each of the fly ash points represent between 85 and 88 % of the chemical composition of the ash. From the phase diagram the first melt is predicted to occur at 985°C for all three ashes. Comparison of these predictions to the measured melting curves should be carried out with great carefulness since the influence of Fe 203, CaO, and MgO on the ash melting Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 15 is neglected. Assuming no influence of these components is probably not a good assumption, since they are all known to have profound influence on melting in silicate systems. However, progressing with the comparison, it is seen that the measured onset for MKS3FA at 1040°C agrees reasonably well with the predicted 985°C, whereas for the other curves the melting onsets do not agree with the phase diagram. The melting is predicted to be complete for all three ashes at temperatures above 1500°C, with the

Si02

k 2o

Fig. 11: Part of phase diagram (Levin et al., 1964) with fly and bottom ashes indicated ending temperature being highest for MKS5FA and lowest for MKS3FA. Since the melt fraction obtained at 1390°C decreases in the order: MKS3FA, MKS4FA, MKS5FA, the relative positions of the melting curves seem to agree with the phase diagram.

2.3.2 Bottom Ash Melting Melting curves for bottom ashes from experiment 3,4, and 5 are shown in Figure 12. For these ashes only a single experiment has been made, except for MKS3BA. It is not possible to make a clear distinction between the bottom ash curves, which all start at app. 1050°C. Correlating the bottom ash melting behaviour to the chemical composition may again be done in the phase diagram shown in Fig. 11, in which each bottom ash composition point represents more than 90% of the ash. The diagram predicts no large difference in melting behaviour for the three ashes, but predicts MKS3BA to be slightly ‘lower melting ’ than MKS4BA and MKS5BA. The first melting point is predicted to be 985°C, and the melting to be completed at above 1600°C. The phase diagram predictions agree reasonably to the measurements, neglecting that the bottom ashes have reached almost complete melting at lower temperatures than predicted. This discrepancy could, as stated before, be due to the fluxing effect of the ash content of CaO and Fe 203, the effect of which is neglected when predicting melting behaviours based on Fig. 11.

The melting behaviour of respectively fly ash and bottom ash from the same/each experiment is compared in Figure 13. In the figure, full lines show bottom ash melting curves, and dotted lines show (reproductions of) fly ash melting curves. For experiment Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 16

-©- BA3 -e- BA3 -B- BA4 BAS

400 800 1200 1600 Temperature [°C] Fig. 12: Melting of bottom ashes from experiment 3, 4, and 5

BA3-1 FA3-2

0.6 -- FA3-3 FA3-4 0.4 -- BA3-2

FA4-1 0.8 - FA4-2 FA4-3

0.4 -

0.2

0.8 - FA5-1 FA5-2 0.6 -

0.4 • -

Temperature/°C Fig. 13: Melting of fly and bottom ashes from experiment 3,4, and 5 Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 17

3, the fly ash shows higher melt fractions than the bottom ash at temperatures below 1300°C, whereas for experiment 4 and 5, the fly ashes seem to show smaller melt fractions than the bottom ashes. Consulting again Fig. 11, this predicts that all bottom ashes should be slightly lower melting than the corresponding fly ash (still neglecting the influence of other ash components). Thus, agreement is obtained between phase diagram predictions and experiments for ashes from experiment 4 and 5, but not for experiment 3.

2.3.3 Deposit Melting It was intended to study a total of nine deposit samples. Unfortunately two of these (MKS41-1 and MKS51-1) were not accessible, and for two of the deposits from position 3: MKS43-1 and MKS53-1, the quantity of deposit collected was too low to obtain reliable results. To account for the missing deposits, an alternative set of deposits: MKS31-2, MKS41-2, and MKS51-2, was examined. Thus, a total of eight deposits were examined. These are presented in Table 9 together with an indication of available chemical composition /mineralogy data.

Table 9: Deposits examined for melting behaviour by STA Exp. 3 Exp. 4 Exp. 5 Deposit Analysis Deposit Analysis Deposit Analysis

Pos. 1 MKS31-1 STA, SEM MKS41-1 SEM MKS51-1 SEM MKS31-2 STA MKS41-2 STA MKS51-2 STA, BCA

Pos. 2 MKS32-1 STA, SEM, BCA MKS42-1 STA, SEM, BCA -

Pos. 3 MKS33-1 STA, SEM MKS43-1 STA - BCA: bulk chemical analysis

MKS31-1 MKS32-1 MKS33-1

MKS31-2 MKS41-2 MKS42-1 MKS43-1 #— MKS51-2

800 Temperature [°C] Fig. 14: Melting curves for deposits from MKS1 -experiments Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 18

Melting curves for the deposits are shown in Fig. 14, revealing typical melting in two distinct intervals, 1) a low temperature range from app. 1100°C to 1250°C, and 2) a high temperature range above 1350°C. Two deposits start melt formation at lower temperatures than the others: MKS31-1 and MKS43-1 with onsets at approximately 950°C, and two deposits extinguish them selves by showing quite large melt formation (-40%) in the high temperature range: MKS33-1 and MKS51-2.

Basing the interpretation on ternary phase diagrams, the following diagrams are relevant: Ca0-Al 203-Si02, K20-Al203-Si02, and Fe0-Al 203-Si02. None of these diagrams make a complete description of the deposits, since only 57-69 %(w/w), 50-70 %(w/w), and 60-81 %(w/w) of the deposit chemistry is represented in the respective diagrams. From these diagrams, the first melt is predicted to occur at 1170°C, 985°C, and approximately 1090°C. Since only a moderate part of the deposit chemistry is represented in the ternary diagrams, a precise prediction of ash fusion can not be expected. Nevertheless, the measured onset temperatures are in the range predicted from the phase diagrams. Furthermore, when plotting the deposits in the K20-Al203- Si02-diagram, two deposits distinguish themselves by having relatively high liquidus temperatures: MKS33-1 and MKS51-1 (this report App. A). This may be the reason for the large melt formation in the high temperature range for these deposits compared to the others. Another reason may be the high SiO^A^Og ratio for these ashes (compared to the others), since increasing Si02/Al203 ratios are known to increase fusion temperatures (Huggins et al., 1981). The reason for the low melting onset for MKS43-1 (for which the chemical composition is not known) and MKS31-1, is not known.

2.4 Comparison to standard AFT The results obtained from the standard ash fusion test (standard AFT) are shown in Table 10. As can be seen, the standard AFT results indicate almost similar (but slightly increasing onset for decreasing straw fraction) melting onset for the three fly ashes, but a difference is found for the Hemispherical Temperature ’s, which increases in the following order: MKS3FA, MKS4FA, MKS5FA. The measured Fluid Temperature ’s indicate, though, that MKS3FA may obtain complete melting at temperatures higher than MKS4FA and MKS5FA. For the bottom ashes three very similar melting behaviours are indicated by the standard AFT results, with the biggest deviation seen for the Fluid Temperature ’s , which indicate the melt completion to occur at the lowest temperatures for MKS5BA, and at the highest temperatures for MKS3BA.

Comparing standard AFT results to results from the STA, it is found that the STA detects melt formation at temperatures well below the IDT for all ashes. For the bottom ashes and MKS3FA, the temperature difference between melting onset as determined by STA and the Initial Deformation Temperature is quite high, between 110 and 160°C. The agreement is better for MKS4FA and MKS5FA for which the deviation in onset temperature is only 20-40°C. The fraction of melt present at IDT vary between 1 and 12%, except for MKS3FA which show 36% melt at the IDT. For the Hemispherical Temperature, melt fractions between 38 and 65% melt is detected. For the fly ashes, the standard AFT finds a relative order of melting curves (for T < 1400°C) of MKS3FA Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 19

Table 10: Standard ash fusion test results and comparison to melting curves

!DT/°C % melt ATTC HT/°C % melt FT/°C % melt

MKS3FA 1180 36.4 -142 1290 65.1 > 1500 -

MKS4FA 1190 1.1 -19 1330 43.3 1490 -

MKS5FA 1220 2.5 -35 1430 - 1490 -

- MKS3BA 1150 10.7 1 1270 49.1 1490

MKS4BA 1150 4.3 -153 1280 38.6 1460 -

MKS5BA 1150 11.0 -158 1300 54.2 1410 - * T0(STA) - IDT showing the highest degree of melting, MKS4FA showing the second highest and MKS5FA showing the lowest. This is equivalent with the STA results and predictions made based on ash chemistry. For the bottom ashes, the standard AFT predicts similar onsets, and indicates approximately the same melting behaviour below 1300°C, whereas the melting proces completion is indicated to occur at lower temperatures for MKS5BA compared to MKS4BA. This was not the result from the STA measuremnts. Overall, it is concluded, though, that a qualitative agreement between the standard AFT and the new STA-based test has been found.

3 Sintering of Ashes Sintering experiments were carried out in order to investigate the relationship between the melting behaviour of a certain ash and the corresponding densification/strength developing process. The conducted sintering experiments implied strength testing of heat treated ash pellets, the procedure of which is described in Skrifvars (1994) and Hansen (1997). The aim of the experiments was to investigate whether it is possible to predict the temperatures, at whichthe strength build-up process starts, based on the STA derived melting curves. For this purpose, the sintering behaviour of four fly ashes were determined and compared to the melting behaviour as determined by STA.

3.1 Sintering Strength Results The influence of the heat treatment temperature on the developed pellet strength for the fly ashes from experiment 1,2,3, and 5 is shown in Figure 15, for a heat treatment time of four hours (composition and fusion of MKS1FA and MKS2FA are shown in App. B of this report). For unsintered pellets or pellets sintered at temperatures below 850°C, the strength was so low that pellets literarily fell apart by the slightest touch. In Fig. 15, it is seen that for MKS1FA, MKS2FA and MKS3FA, the first strength development is detected at 950°C. Defining the sintering temperature as the highest temperature at which no strength increase is detected (Skrifvars, 1994) results in a sintering temperature of 900°C for MKS1 FA, MKS2FA, and MKS3FA, whereas for MKS5FA, the sintering temperature is 1000°C. In practice, this means, that if deposit surface Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 20

MKS1FA.

•S, 60 - MKS3FA I 50--

° 40 ■- MKS5FA I 30 -

° 20 MKS2FA

1000 Temperature/*C Fig. 15: Strength of pellets as a function of sintering temperature temperatures can be kept below the sintering temperature, no strength will be build up in the deposits, which will thus be quite easy to remove by sootblowing. At 1000°C, two of the ashes show a lower strength compared to the strength at 950°C, but for higher temperatures, the strength is generally seen to increase with increasing temperature. An exception to this are the last points at the curves for MKS2FA and MKS3FA, which exhibit lower strengths at 1150°C compared to 1100°C. Ash sintering/strength build-up is generally believed to be related to a porosity decrease. Since the pellet weight and dimensions were measured after heat treatment, an estimation of the pellet density could be made. Plotting the measured strengths as a function of the measured pellet densities (Fig. 16), it appears that a qualitative relation between pellet strength and density is found: When the density increases, so does the strength, and vice versa. A density decrease is thus observed for the two above mentioned decreases in strength, which is therefore believed to be caused by evaporation of material, leading to a higher pellet porosity.

MKS1FA MKS3FA

MKS5FA

MKS2FA

1.4 Density [g/cm3] Fig. 16: Relation between pellet strength and density Fusion, Sintering and CCS EM Analyses of Ashes from Coal/Straw Co-Firing 21

Furthermore, it can be seen at Fig. 15, that the strengths developed by the different ashes for a given temperature does not differ very much, but it appears that the strength decreases in the following order (strongest ash first): MKS1FA, MKS3FA, MKS5FA, MKS2FA, for temperatures between 1000°C and 1150°C. Taking the measurement uncertainty into consideration, it is not possible to distinguish between all curves, but MKS1FA does exhibit a significantly higher strength than MKS2FA, with MKS3FA and MKS5FA having intermediate strengths (for temperatures between 1000°C and 1150°C).

Sintering of MK52FA Sintering of MKS1 FA

§ 35 5, 50 -

2 40

§ 20 S 10 -*-7 = 1100*0

Time[h]

Fig. 17: Pellet strength as a function of sintering time

The influence of time on the strength development was also investigated. Fig. 17 shows the results for varying time of sintering at respectively 1050°C and 1100°C for all four fly ashes investigated. These temperatures were chosen based on the above mentioned experiments as sintering temperatures at which pellets of ‘medium ’ strength was obtained. The data show that strength development occurs very fast for the first half hour of heat treatment, during which the ashes generally obtain between 64 and 103 % of the finally obtained strength (~12 h heat treatment). For all ashes, the strength curves exhibit a bend between % and 4 hours heat treatment, the reason of which is not known at present. Strength measurement data was plotted as a function of pellet density, and it did appear that the density curves also exhibits a bend between a 34 and 4 hours, but no direct relation could be found. From 4 to 12 hours the strength of all ashes changed only slightly, with ratios between 0.79 to 1.17.

3.2 Relation between Ash Fusion and Sintering To investigate the relation between melting and strength build-up in the ashes, Fig. 18 shows matching values for strength and melt fraction in the ash at the different heat treatment temperatures. Generally, it is seen that strength build up is initiated at temperatures lower than the melting onset temperatures (T0) of the ashes. For the fly ash from 100 % coal combustion (MKS5FA), the full strength build up appears to occur without the presence of a melt phase, i.e. probably as a result of viscous flow sintering. Viscous flow sintering is normally the word used to describe strength build up in ashes through the appearence of a liquid phase of a high viscosity. In this case no liquid Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 22 phase has been detected in the ashes, but flow of viscous ash may occur (slowly) any way, since these ashes are glasses which do behave as liquids with a very high viscosity. For the ashes produced during co-firing (MKS1FA, MKS2FA, MKS3FA), it appears, that the initial strength build up occurs without the influence of a melted phase, but that strengths do not reach values above 4 N/mm2 without the presence of a liquid phase. The further behaviour shows no simple relation between melt fraction and strength obtained. The conclusion seems to be that the straw-firing produces ashes which sinter Strength vs. melt-%

MKS5FA MKS3FA MKS1FA MKS2FA

% melt in ash

Fig. 18: Relation between strength and melt fraction in pellet by a different mechanism than the pure coal-derived ashes and which needs the presence of melt to introduce high strengths. The straw share does not seem to have any systematic influence on the observed sintering behaviour, though, and no simple relationship between strength and melt fraction could be found. Probably the physical characteristics of the fly ash particles are dominant factors in the strength build up.

3.3 SEM of Sintered Ash Pellets In order to obtain information on the microstructure of the pellets at the different sintering stages, a selected number of pellets were cut through, and the cross sectional surface was studied in the SEM. The pellets studied were those produced by MKS2FA and MKS5FA, which had been heat treated for four hours at respectively 850, 950 (missing for MKS5FA), 1050,1100, and 1150°C. Pictures showing the pellet structure are shown in Fig. 19 and 20. The black areas represent epoxy filled voids and the different grey tones the density of the different ash particles. Both figures show that at 850°C the pellets consist of individual ash particles, which are not bonded together to any significant extent. Even the largest fraction of the smallest particles (which are most prone to sinter) seem to occur as discrete particles. Increasing the heat treatment temperature to 950°C, for which unfortunately no MKS5FA pellets were left, a quite high degree of bonding has occurred between the ash particles. At 950°C, far the largest fraction of the smallest particles have been attached to other particles and only a few of the original particles can still be seen as individual. Increasing the temperature further, to 1050, 1100, and 1150°C, the ash particle ‘agglomerates ’ grow larger in size Fusion, Sintering, and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 23

Fig. 19: SEM pictures showing the structure of sintered MKS5FA pellets Fusion, Sintering, and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 24

Pellet sintered at 850°C Pellet sintered at 950°C

Pellet sintered at 1100°C

Fig. 20: SEM pictures showing the structure of sintered MKS2FA pellets Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 25 and the pores present reduce in number and grow in size. This is the general trend seen for both ashes, and this reflects the measured strength increase.

Comparing the two ashes, it is seen that the MKS5FA pellets show a higher degree of particle attachment than MKS2FA at 1050°C. Consulting the strength curves it appears that no significant strength difference was found for pellet strengths. At 1100°C, no significant difference in microstructure can be seen for the two ashes, whereas the micro structure of MKS2FA sintered at 1150°C shows quite a higher attachment degree compared to MKS5FA sintered at 1130°C. Comparing the matching strength measurements, it is seen that the strength is significantly higher for the MKS5FA pellet compared to that of MKS2FA. Thus, based on the few cases studied here, no simple simple relation between the degree of particle attachment and pellet strength could be found.

The porosities of the pellets were estimated as the ratio between black and grey tone areas. Results from these area comparisons are not very accurate porosity determinations, and results should thus be interpreted with care. The obtained results are shown in Table 11, together with the matching pellet strengths. As can be seen there is a general trend, that porosity reduces as the heat treatment temperature (and pellet strength) increases. For MKS2FA, the data all follow this trend, except for the pellet sintered at 1150°C, which shows significantly higher porosity than the pellet sintered at 1100°C. The reason for this is not known, but the data correlate to a decrease in strength. For MKS5FA, the data are less unambiguous, the porosity decreases considerably from 850°C to 1050°C, but for 1100°C and 1130°C, the porosities are both higher than that measured at 1050°C. The matching strength measurements show a strength increase from 1050°C to 1100°C, even though the porosity increases slightly (10%). For the temperature increase from 1100°C to 1130°C still a slightly higher porosity is found, but the strength is not changed. The unchanged average strength at 1130°C is primarily due to one of the pellets showing a significantly higher strength than the other three pellets, though. Ignoring this one pellet, the average strength decreases. This means that the overall trend holds: pellet strength and porosity are related, causing pellet strength to increase when the porosity decreases, and vice versa.

Table 11: Strength and porosities (determined by SEM) for sintered ash pellets

850°C 950°C 1050°C 1100°C 1150°C

MKS2FA porosity 0.669 0.621 0.495 0.413 0.719

strength 0.04 3.7 19 36 22

MKS5FA porosity 0.605 - 0.443 0.505 0.547*

strength 0.1 0.2 0.8 49 48 * sintered at 1130°C Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 26

Especially for pellets of MKS2FA sintered at 950°C and 1050°C, it is possible to divide the structure into two parts: 1) original particles of unchanged form and 2) areas consisting of deformed material bonding together/connecting ash particles. The chemical composition of these two different material types was investigated by use of SEM-EDX, a result of which is shown in Fig. 21. At 950°C, the ‘binder phase ’ extinguishes itself from the ‘original particles of unchanged form ’, by a lower Si ratio, a higher Al, Fe and Mg content and a lower K content. Generally, a decreasing content of Si02 is known to decrease the viscosity of a glass, Al203 may influence viscosity both positively and negatively, whereas increased contents of Fe, Mg, and K decreases glass viscosity (Urbain et al., 1981). This differing chemical composition could thus indicate a ‘binder phase ’ with a lower viscosity than the surrounding particles, which explains the higher deformation tendency of this phase compared to the particles. Chemistry of MKS2FA at 950°C

Binder Part 1 Part. 2

Element no.

Fig. 21: Chemistry of ‘deformed ’ vs. undeformed particles

The chemistry of the two types of material was additionally characterised by an x-ray mapping (Fig. 22) of an area including both ‘original particles of unchanged form ’ and clearly deformed material. From the figure, it is seen that Si makes up the ‘sceleton ’ of the particle structure, and that very high Si concentrations are found in the undeformed particles. Al seems to ‘follow ’ the Si, except for a few pure Si02 particles. K is concentrated in areas/particles showing quite high deformation, and so is Ca, except for the Ca being present as phosphates. The Fe seems to be evenly distributed, and so does the Na. In conclusion, the deformed particles binding together the matrix seem to be enriched in K and/or Ca compared to the bulk particles, which is equivalent with the K/Ca-rich material to have a lower viscosity and thus a higher tendency to deform than very Si-rich or Al- and Si-rich material.

Significant strength build-up and a large degree of particle attachment/agglomeration was observed in the ash pellets at temperatures below the melting onset temperature (as determined by STA). Since no melt has been detected the particle attachment and the strength build up must occur due to flow of viscous material. The viscosity of the sinter tested ashes has thus been calculated. The model chosen for calculation was the following (Urbain et al., 1981): Fusion, Sintering, and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 27 950°C

at

sintered

pellet

ash

MKS2FA

of

section

cross

of

mapping

x-ray

22:

Figure Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 28

1.0 E+09

X X

1.0E+08 j ____ /

% f

: — —MKS5FA E, 1.0E+07 ' X — - MKS2FA if — - - MKS3FA g 1.0E+06 X.' V "X i MKS1FA 1.0 E+05

1.0E+04 800 850 900 950 1000 1050 1100 Temperature [°C] Fig. 23: Calculated viscosities [Pa s]

1000 B r| = A • T • exp( ) (1) where A and B are composition dependent variables (App. C in this report). This model was chosen between many available viscosity models, since this model have provided estimates being in good agreement with experimental values in a recent study of coal ash viscosities (Vargas et al., 1997). The calculated viscosities are shown in Fig. 23 and tabulated as a function of temperature in Table 12. As can be seen the ash viscosities decrease in the following order (high viscosity ash first): MKS5FA, MKS2FA, MKS3FA and MKS1FA, with the largest deviation between MKS5FA and MKS2FA values. Comparing this to the strength measurements, it was found that for temperatures above 1000°C, MKS1FA showed the highest strengths, followed by MKS3FA, then MKS5FA and finally MKS2FA. It thus appears that the ashes which show the lowest viscosity (presumably causing ash particles to have a higher deformation/attachment tendency) show the highest strength; the ash with the second lowest viscosity shows the second highest strength, but the two ashes with the highest viscosities have ‘switched ’ positions, so that the ash with the highest strength shows only the second lowest strength. It thus appears that a correlation between ash viscosity and strength build up exists, generally meaning that lower ash viscosities result in higher strength build up.

Table 12: Calculated viscosities [Pa-s] MKS1FA MKS2FA MKS3FA MKS5FA

800°C 2.8- 107 5.9-1 07 3.2-107 2.8-10 8

850°C 6.6-106 1.3-107 7.3-106 4.8-1 07

900°C 1.7-106 3.2-106 1.9-10 6 1.5-107

950°C 5.1-105 9.0-1 05 5.6-105 3.1-106 1000°C 1.7-10s 2.8- 10s 1.8-10 5 1.1-10s

1050°C 5.8-1 04 9.8-10" 6.4-10" 2.9- 10s Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 29

Investigating the data it also appears that the ash viscosity at the sintering temperature is almost alike for the four ashes: Ranging between 1.1-106 and 3.2-106 Pa-s. Relating this to estimates reported in the literature, this corresponds to rapid sintering (105 -107 Pa-s) as described by Raask, (1985).

In Table 13, gas temperature measurements conducted at position 1, 2, and 3 for the four experiments, for which fly ash sintering tests were conducted, are shown. Relating these temperatures to the sintering temperatures found from Figure 6.11 and considering worst case: A deposit surface temperature equal to the measured gas temperature, it is seen that sintering may happen for all experiments (1, 2, 3, 5) in positions 1 and 2, but not in position 3. This interpretation agrees with operational experience (Andersen, 1997). The most severe sintering is believed to occur for position 1 in experiments 2, 3, and 5, and position 2 in experiment 3.

Table 13: Measured gas temperatures [°C] (modified from Andersen, 1997b)

Exp. 1 Exp. 2 Exp. 3 Exp. 5

Position 1 1060 1190 1225 1266

Position 2 1030 954 1125 1099

Position 3 755 851 766 954

To evaluate the sintering tendency in the furnace, the viscosities of the bottom ashes were calculated and the temperature corresponding to a viscosity of 1 -3-10s Pa-s was found. These temperatures vary between 870 and 980°C, which is quite probable to occur at deposit surfaces in the furnace. Taking into consideration the melting onsets for the bottom ashes, very strongly deposits being melted at the surface are expected to form in the furnace. This is also in agreement with operational experience (Andersen, 1997).

4 Modelling Ash Deposit Formation The simple model presented for deposit formation calculations in Hansen et al. (1997) (App. D) was used to calculate deposit formation rates also in the MKS1 boiler, with the important model change, that condensation was not taken into account for the MKS1- boiler calculations. This change was made for two reasons: 1) condensation is not believed to play an important part in deposit build-up in coal-fired boilers, and 2) SEM images of the collected deposits showed no systematic evidence of an initial deposit layer consisting of condensated material (Andersen, 1997b). The model therefore only included impaction of fly ash particles. Assumptions were as earlier described: a) fly ash particles of uniform size (the average fly ash particle size) impacts on the probe according to the correlations presented by Wessel and Righi, (1988), Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 30

Table 14: Input parameters and results for deposit formation rate (DR) calculation

dp fentr 1> Vgas Tgas melt m 2) fstick *exp DR, caic meas. ratio pm % m/s °c % % % h g/m2/s g/m2/h

MKS31-1 33 92.4 17 1225 53.8 4.0 49.8 20.0 20.9 34.3 2196

MKS31-2 33 92.4 17 1225 53.8 4.0 49.8 19.6 20.9 14.7 5125

MKS32-1 33 92.4 6 1125 16.9 2.9 14.0 17.8 2.5 86.0 104

MKS33-1 33 92.4 6 766 0 0.6 0 20.0 0 11.0 -

MKS41-2 39 87.6 17 1275 15.7 8.3 7.4 17.8 2.7 78.1 123

MKS42-1 39 87.6 6 1108 0 3.5 0 18.2 0 6.5 -

MKS43-1 39 87.6 6 924 0 2.0 0 18.7 0 0.9 -

MKS51-2 28 91.5 17 1266 9.7 2.4 7.3 ? 2.9 46.6 223 1) the fraction of fuel ash being entrained as fly ash 2) mass of ash evaporated at gas temperature

and b) the fraction of impacting ash particles which stick to the probe is equivalent to the melt fraction in the fly ash as determined by the STA at the gas temperature prevailing. The input parameters used and the results obtained are presented in Table 14.

From Table 14 is seen, that the model overpredicts the deposit formation considerably (with a factor between 100 and 5100) for five measurement positions and underpredicts the deposit formation rate for three measurement positions. As can be seen, the three experiments showing no (calculated) deposit build-up are the three experiments showing the three significantly lowest (measured) deposit formation rates. However, the slightly higher melt fractions (7.3 and 7.4%) detected for ashes corresponding to MKS41-2 and MKS51-2 do not result in equally high measured deposition rates. The again higher melt fraction (14.0%) corresponding to MKS32-1, is equivalent to a slightly higher measured deposition rate, but the quite high (49.8%) melt fractions for MKS31-1 and MKS31-2 do not result in relatively higher deposition rates. Actually, far the most extreme overpredictions are found for the deposit measurement made in experiment 3, position 1, which indicates that the sticking coefficient is not proportional to the melt fraction in the equivalent fly ash. In conclusion, no linear correlation between melt fraction in fly ash and the rate of deposit formation can be found, and it seems that only small melt fractions need to be present in the fly ash to obtain quite high deposit formation rates.

The reason for the underpredicted deposit formation rates for deposit MKS33-1, MKS42-1, and MKS43-1, is that no melt is detected in the fly ash by the STA at the prevailing temperature. In the model, no melt detected results in a sticking coefficient equal to zero, which means that no deposit is formed at all. In a real boiler, one would expect some deposit to form, even if only dry ash particles impact the tube, since after Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 31 a while a very thin layer of small particles will probably create a slightly sticky layer, which may catch some of the incoming dry ash particles. Furthermore, the used sticking coefficient may be too low, due to the fact that hysteresis effects in melting behaviour displayed by silicate ashes have not been taken into account. The hysteresis effect may cause that even though no melt is detected by the STA (during ash heat-up), the real fly ash (which has been cooled on the route through the convective pass) may still contain a small fraction of melt, leading to a real sticking coefficient which is higher than zero.

Generally, there are three major reasons to the large deviation between measured and calculated deposit formation rates: 1) As was the case for the Haslev and Slagelse boilers, the measured deposit formation rates are (definitely) too low. In several experiments, large parts of the collected deposit fell off the probe during probe withdrawal (Andersen, 1997b) 2) In this model no initialisation time is included, which means that the fast deposit build-up by inertial impaction is started from time zero in the model. In reality, probably a thin layer of small particles need to be build up either by particle diffusion or thermophoresis before the particle impaction is initialised. If no initial layer is build up in reality, then probably at least the impaction rates are lower in the beginning of experiments, and then reaches a steady value after a certain initialisation time. 3) Finally, maybe the most important reason for the observed disagreement between calculated and measured deposition rates is the lack of including deposit removal mechanisms in the model. In a system like the presented with a very high particle loading in the flue gas (app. 90% of the fuel ash is entrained by the flue gas) erosion effects must be important. This is not at least true for the measurements in position 1, where the flue gas velocity is quite high (17 m/s). This explanation is supported by the measurements of the temperature signals from the deposition probes, which show systematic temperature changes, corresponding to the loss of deposit due to shedding (Andersen, 1997a).

The presented simple model is not able to describe the deposit formation rates reasonably. It must thus be concluded that the concept of using one (average) particle size and estimating the sticking propability directly as the melt fraction (using bulk ash chemistry in stead of particle-size-dependent chemistry) is not sufficient for modeling deposit build-up in coal-fired boilers. Nevertheless, it seems that there is a qualitative connection between the melt fraction and the deposit formation rate, which means that using a more complex non-linear correlation between melt fractions (of different size fractions of the fly ash) and the sticking coefficient may give a better agreement between calculated and measured deposit formation rates. A proper model should also include, though, an expression for deposit shedding. Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 32

5 Conclusions

CCSEM analyses of three sets of fly ashes, bottom ashes and deposits collected during co-combustion of respectively 0, 10, and 20% straw (on an energy basis) in a coal-/straw PF-fired boiler revealed that far the largest part of the ashes consisted of metal silicates. For the fly ashes, the chemical composition was quite alike on an oxide basis, but for the CCSEM data a clear trend was seen: when potassium was available for reaction (i.e. when straw was burned), a very large part of all present alumina- silicates had reacted to form potassium alumina silicates. Comparing fly ash compositions to deposit compositions, it was found that less different species were present in the deposits. Species being present (concentrated) in deposits included K-, Ca-, Fe-, and Fe-Al-silicates, iron oxide, quartz, illite derived species, and mixed phosphates.

All ashes examined showed melting in the temperature range from 1000°C to 1390°C, where the deposits typically showed a higher melt fraction at 1100°C than the matching fly and bottom ashes. The repeatability of the melting curves was investigated, and it was found that repeatabilities were approximately 30°C for melting onset and generally within 15% melt at a given temperature. The repeatability for these silicate rich ashes in general was lower than for the salt rich ashes presented in Chapter 5. This is primarily due to a very slow melting start for silicate ashes compared to the abrupt melting start for the salt rich ashes.

When comparing results from the STA melting quantification method to results from the standard ash fusion test, moderate quantities of melt (1-36%) was found at the IDT. Comparing the IDT to the onset of melting as determined by the STA, it was found that the first melting occurred as much as 150°C below the IDT. This stresses that the standard ash fusion method should be used with care when determining melting behaviour.

Sintering experiments were carried out to investigate the relationship between the melting behaviour of a certain fly ash and the corresponding densification/strength developing process. Strength was found to build up in all ashes at temperatures below the first melt appearance, probably due to the flow of viscous material. For the fly ash collected during pure coal combustion, very high strengths was build up with out any presence of a melted phase, whereas for the ashes produced during partly straw burning, strengths above 4 N/mm2 were not obtained without melt present. Viscosity calculations revealed that for all ashes the sintering onset was equivalent with a viscosity of (1 - 3) • 106 Poise. Relating sintering experiment results to real boiler performance, it was interpreted that severe strength build up could happen for deposits in probe position 1 and 2, but not in position 3. These interpretations were in agreement with operational experience.

A very simple model set up to estimate the deposition rates was found not to agree with the measured values. Thus, the model concept based on bulk ash chemistry Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 33 was too simple. Nevertheless, it did seem as if some connection between melt fractions and deposit formation rates was present, so for future calculations, a more complex correlation between sticking coefficients and ash chemistry must be employed, and deposit shedding need to be accounted for.

ACKNOWLEDGEMENT

This work was partly carried out as part of the CHEC Research Programme at the Technical University of Denmark, and partly at the Geological Survey of Greenland and Denmark. Specific funding from the Danish Energy Research Programme, Contract No. 1323/95-0007, is greatfully acknowledged. Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 34

Cited Literature

Andersen, K.H.A., (1997a) “Deposition Trials at Studstrup Power Station ” Presented at the Nordic Ash Workshop, Aabo Academy University, Turku, Finland, 1997

Andersen, K.H., (1997b) Personal Communication

Andersen, K.H., Frandsen, F.J., Hansen, P.F.B., and Dam-Johansen, K., (1997c) “Full Scale Deposition Trials at a 150 MWe PF-Boiler Co-Firing Coal and Straw: Summary of Results ” Presented at the Eng. Found. Conference on Impact of Mineral Impurities in Solid Fuel Combustion at Kona, Hawaii, November 2-7,1997

Andersen, K.H., Hansen, P.F.B., Wieck-Hansen, K., Frandsen, F., and Dam-Johansen, K., (1996) “Co-Firing Coal and Straw in a 150 MWe Utility Boiler: Deposition Propensities ” Presented to the 9th European Bioenergy Conference, Copenhagen, Denmark, June 24-2J, 1996

Frandsen, F.J., Hansen, L.A., Sorensen, H.S., and Hjuler, K., (1998) “EFP-95 Project: Characterisation of Biomass Ashes - Data Compilation ” CHEC Report No. 9805, Department of Chemical Engineering, Technical University of Denmark, 1998

Hansen L.A., (1997) “Melting and Sintering of Ashes” Ph.D. Thesis, Department of Chemical Engineering, Technical University of Denmark, 1997

Hansen,L.A., Frandsen, F.J., Sorensen, H.S., Rosenberg, P., Hjuler, K., and Dam- Johansen, K., (1997) “Ash Fusion and Deposit Formation at Straw Fired Boilers ” Presented at the Eng. Found. Conference on Impact of Mineral Impurities in Solid Fuel Combustion at Kona, Hawaii, November 2-7,1997

Hansen, P.F.B., (1997) “MKS1 Demoprogram. Stofbalancer og roggasemissioner ” In Danish. Internal Report from I/S Midtkraft, 1997

Hansen, P.F.B., Andersen, K.H., Wieck-Hansen, K., Overgaard, P., Rasmussen, I., Frandsen, F.J., Hansen, L.A., and Dam-Johansen, K., (1996) “Co-Firing Straw and Coal in a 150 MWe Utility Boiler: In-Situ Measurements ” Fusion, Sintering and CCSEM Analyses of Ashes from Coal/Straw Co-Firing 35

Presented at the Engineering Foundation Conference, Snowbird, Utah, April 28 - May 3, 1996.

Huggins, F.E., Kosmack, D.A., and Huffman, G.P., (1981) “Correlation between ash-fusion temperatures and ternary equilibrium phase diagrams ” Fuel, Vol. 60, pp. 577-584, 1981

Raask, E., (1985) "Mineral Impurities in Coal Combustion" Hemisphere Publishing Corporation, Springer-Verlag, 1985

SkrifVars,B.-J., (1994) “Sintering Tendency of Different Fuel Ashes in Combustion and Gasification Conditions ” Academic Dissertation, Combustion Chemistry Research Group, Abo Academy, Finland, 1994

Urbain.G., Cambier.F., Deletter,M., and Anseau.M.R., (1981) “Viscosity of Silicate Melts ” Trans. J. Br. Ceram. Soc., vol. 80, pp. 139-141, 1981

Vargas,S., Frandsen, F.J., and Dam-Johansen,K., (1997) “ELSAM-ldemitsu Kosan Cooperative Research Project: Performance of viscosity models for high-temperature coal ashes ” Internal CHEC Report no. 9719, Department of Chemical Engineering, Technical University of Denmark, August 1997. Appendix A: Positions of MKS deposits in K20-Al203-Si02 phase diagram

SiO, MKS5BA CristoboEJe ,/ MKS31-1, MKS32-1, /MKS3BA MKS41-1, MKS42-1' MKS4BA MKS51-1 MKS3FJ. MKS4FA MKS5FA

MKS33-1

2Si0g ~BSO"

Kp-Ay), Iy>-IIA1A Aifl, ~2020• Appendix B: Composition and melting curves for fly ashes from MKS, experiment 1 and 2

Relative chemical composition of fly ashes [%(w/w)] from experiments 1,2 and 5 (modified from Hansen, (1997))

Si02 AIA Fe 203 CaO MgO Na zO K20 so3 p 2o5 Cl MKS1FA 104 101 108 183 127 62 323 268 373 8000

MKS2FA 99 79 104 183 113 98 298 159 353 12200 MKS5FA 100 100 100 100 100 100 100 100 100 100

0.9 ----- MKS1FA 0.8 - - - MKS2FA c 0.7 - -

0.6 - .2 J).5 - - s 0.4 - - S 0.3

0.2 -

0.1 -

Temperature/°C

Melting curves for fly ashes from experiment 1 and 2 Appendix C: The Urbain Model (Urbain et al., 1981)

The content of species with a viscosity lowering effect are summed in a constant M: M = CaO + MgO + Na2 0 + K2 0 + FeO + MnO + NiO + 2 TiOz + 2 Zr02 where all species notations represent mole fractions; all Fe is assumed present as FeO

A constant a is calculated:

a =------—------M + a/2o,

A number of B,-constants are calculated on the basis of a:

B0 = 13.8 + 39.9355 • a- 44.049 • a 2 8 1 = 30.481- 117.1505 • a + 129.9978 • a 2 S2 = -40.9429 + 234.0486 • a - 300.04 • a 2 B3 = 60.7619 - 153.9276 + 211.1616 • a 2

The B constant used in the expression for viscosity is calculated as a sum of products of the B/s and the molar fraction of silica in the melt:

B = B0 + B1 • S/02 + B2 • Si02 + S3 • Si02

A, the second constant used in the expression for viscosity, is calculated from B: -In A = 0.2693 • B + 11.6725

Finally, viscosity, n (Poise), can be calculated as function of temperature, T (K), as:

1000 ■ B H = A • T • e T Appendix H:

Estimation of Ash Deposition Fluxes in Straw (Co-)Fired Utility Boilers.

H Estimation of Ash Deposition Fluxes in Straw (Co-)Fired Utility Boilers

Research Progress Report, February 1998.

Flemming J. Frandsen and Lone A. Hansen Table of Contents:

1. Introduction: Project Incitement 2

2. Residual Ash Transport: Inertial Impaction 3 2.1. Impaction of Particles on a Cylindrical Obstacle 4 2.2. Sticking of High-Si Fly Ash Particles 5 2.3. Sticking of Biomass Fly Ash Particles 8

3. Submicron Ash Transport: Thermoforesis 9 3.1. Models for the Thermoforetic Velocity, uT 10 3.2. An example: Calculation of uT 12

4. Ash Deposition in Straw-Fired Utility Boilers 13

5. Summary and Concluding Remarks 15

Acknowledgement 16

References 17 1. Introduction: Problem Incitement

The elements contained in fossil and bio fuels can be grouped in three concentration levels (Benson and Holm (1985) and Benson et al. (1993)): 1) the major elements, C, O, H, S, and N, forming the organic matrix of the fuel, 2) the ash forming elements, Al, Ca, Fe, K, Mg, Na, P, and Si, present in the concentration range of about 1000 ppmw to a few %(w/w) on a dry fuel base, and 3) trace elements, e.g. As, B, Cd, Cl, Cr, Hg, Ni, Pb, Se, and Zn, typically present in concentrations below 1000 ppmw on a dry fuel base.

The ash-forming elements occur in fuels as internal or external mineral grains, simple salts such as Na 2S04 and KCI or associated with the organic matrix of the fuel. Depending on the temperature and local stoichiometry during fuel particle heat-up, devolatilization and char burnout, the simple salts may vaporize, while the mineral grains may undergo phase transformations and approach each other forming residual fly ash. In the furnace, some of the fly ash will be removed as bottom ash while the rest will be entrained with the hot flue gas. The vaporized metal species may undergo several transformations: nucleation, coagulation, heterogeneous condensation and/or interactions with mineral grains/ash droplets in the char. Depending on the total specific surface area of the fly ash particles, the rate of cooling of the flue gas, the local stoichiometry, and mixing in the gas phase, local supersaturation with respect to certain chemical species such as Na 2S04(g), KCI(g), and/or K2S04(g) may lead to the formation of a fine mode aerosol by homogeneous nucleation and subsequent coagulation (Christensen (1995)). Vapors and fly ash particles may deposit on heat transfer surfaces in the boiler through a number of mechanisms (Baxter et al. (1992), Rosner et al. (1992), and Baxter (1993)): inertial impaction, thermoforesis, condensation, and eddy turbulence. Ash deposits may cause several operational problems, e.g. changes in the boiler heat uptake profile (Wall et al. (1994)), corrosion of heat transfer metal surfaces (Harb and Smith (1990), Jacobson et al. (1990), Ahila and Iyer (1992), and Michelsen et al. (1996)), and/or in extreme cases plugging of the convective pass of the boiler. This may cause unscheduled outages of the boiler and significant economical losses (Borio et al. (1992)).

Traditionally, three approaches of prediction of ash deposition propensities have been applied: 1) empirical prediction (based on bulk ash chemistry indices and plant/operational parameters), 2) thermochemical equilibrium calculations, and 3) mechanistic modeling of the ash formation and transport, and deposit build-up.

Empirical prediction of ash deposition propensities, based on ash chemistry indices and plant parameters, is easy to perform and requires a minimum of fuel (proximate, ultimate and bulk ash analyses) and boiler configurational (size of furnace, heat input, steam data and production rate, geometry of burner belt etc.) data. Anyhow, recent work, has revealed that the existing U.S. empirical data material is not always suitable for Danish coal-boiler combinations (Frandsen (1997a)).

2 A Global Equilibrium Analysis, GEA, based on minimization of the total Gibbs energy of a well-defined chemical system of known temperature, pressure and total elemental composition, may provide valuable flash-type partitioning data, i.e. information about the distribution of elements among and within phases in the system, see Frandsen (1995,1997b). The limitations in this type of analysis was briefly addressed by Frandsen et al. (1996).

Mechanistic prediction of ash and deposit formation requires detailed models for ash formation, ash transport and deposit build-up. The first basic need is a very detailed information about the abundance and association (mineral grains, organic association, simple salts) of the inorganic metal species in the fuel. Next, a model for fly ash formation is needed. This model should take into account char and mineral grain fragmentation, and coalescence of ash droplets. Then, a model for transport of ash species is needed, taking into account diffusion (Brownian, thermoforetic and eddy) and inertial impaction. Finally, a model for deposit build-up is needed.

Applying the simple theory outlined by Rosner (1986), Sarofim and Helble (1994) have reported the times, necessary to build up a uniform layer of deposit on a single tubular cylinder, considering each of the mechanisms, diffusion, thermoforesis, and inertial impaction, at a time, see Table 1.

Sarofim and Helble (1994) considered a coal with 10 %(w/w) ash, 2 %(w/w) of which forms a condensation aerosol with a diameter of 0.1 pm, and the remainder forming residual ash particles with a diameter of 20 pm. The 1200 K combustion products are assumed to approach a cylindrical heat transfer tube, 5 cm in diameter and 800 K, at a velocity of 10 m/s and a particle velocity of 10 m/s. The coal ash is assumed to contain 2 %(w/w) Na, 10 %(w/w) of which vaporizes and eventually forms a Na 2S04 deposit.

As seen in Table 1, inertial impaction is by far the fastest of the three transport mechanisms considered by Sarofim and Helble (1994). Thus, below, as the starting point, a simplified model for quantification of the inertial impaction mass flux, is outlined.

2. Residual Ash Transport: Inertial Impaction

Residual fly ash particles, i.e. particles with d P > app. 10 pm, may posses too much inertia to follow the gas streamlines around a tubular cylinder, e.g. a steam tube or a single cooled probe placed somewhere in the convective pass of a boiler. Thus, the residual ash particles may impact on the cylinder and possibly stick to this, see Figure 1.

An inertial impaction mass flux, Nimp (g/nfs), for particles of size dp, is given as:

^imp — f|rnp "^stick " ^ra (1) where u„ (m/s) is the bulk gas velocity, cra (g/m 3) is the mass concentration of residual fly ash particles, fimp is the fraction of residual fly ash particles that impacts the tube, and fstick is the fraction of those particles impacting the tube that sticks to it.

In eqn. (1), the product ujcra represents the total flux (g/m 2s) of residual fly ash particles, i.e. a mass (g) of ash transported through a unit area (m2) in a unit time (s). This flux is corrected for the fact that not all the particles impacts the tube (multiplication by fimp) and furthermore for the fact the not all impacting particles sticks to the surface of the cylinder (multiplication by fstick ), see Nielsen (1997). The product of the correction factors, fimp and fstick, is the capture coefficient, fca P, introduced by Benson et al. (1993).

2.1. Impaction of Particles on a Cylindrical Obstacle

The main forces acting on residual fly ash particles are inertial and drag forces. Inertial forces are proportional to the mass of the particle whereas drag forces are a function of the particle surface area. Whether the particles collide with an obstacle depends on the ratio of inertial and drag forces. The larger the inertial force, the greater the chance of collision with the obstacle. Thus, since the ratio of inertial-to-drag forces is dependent on the particle diameter, the proportion of impacting particles increases with increasing diameter (Huang et al. (1996)).

In order to express the ratio of inertial-to-drag forces, the Stokes number, Stkc, for flow around a cylindrical tube is introduced (Wessel and Righi (1988)): stkc = Pp_^Up (2) 9- IVd t

In eqn. (2), pP (g/m 3) is the particle density, d P (m) is the particle diameter, uP is the particle velocity, pg (kg/m/s) is the gas viscosity, and d t (m) is the tube diameter.

The Stokes number can be interpreted as the ratio of the characteristic stopping (relaxation) time of the particle to the charateristic flow time around the tube (collector). Israel and Rosner (1983) generalized the Stokes number by introducing an effective Stokes number, Stkceff , to include also non- Stokesian drag on particles:

Stk=ff =\|/-Stkc (3) where $ is a function of the particle Reynold number, Re P, defined as:

4 Pp' Vg-%, •dn Re p = (4) ^g In eqn. (3), iji is determined from:

24 rRe P.m=x d(Re p) V( Rep): (5) Rep Jo CD(Rep)-ReF

The drag coefficient, CD, in eqn. (5) may be calculated from the empirical correlation:

CD=^--(1 +0.158-Ref) (6)

i\©p suggested by Serafini (1954). Eqn. (6) is valid for Re p < 103.

Based on the Stokes number, Stkc, defined in eqn. (2), Israel and Rosner (1983) suggested the following formula for calculation of the impaction coefficent, fimp: fimP =1/[l + b-e" 1 -c-e -2 + d-e -3] (7) where: e = Stk^ - a (8)

Baxter and Hardesty (1992) fitted the coefficients, a, b, c, d, and e in eqn. (7- 8) to their own simulations, considering two cases: a) a solution including thermoforesis and b) a solution of the fundamental particle fluid dynamic equations neglecting thermoforesis.

In Table 2, the original coefficients suggested by Israel and Rosner (1983) and the modified coefficents found by Baxter and Hardesty (1992) are listed.

2.2. Sticking of High-Si Fly Ash Particles

As shown in Figure 1 and illustrated by Junker (1997) and Nielsen (1997) in a series of pictures from the Sandia Multi Fuel Combustor, not all the impacting particles will stick to the surface of the heat transfer tube. The sticking of a particle will depend on 1) its physical state, 2) its kinetic energy, 3) its angle of impact, and 4) the condition of the tube surface (Raask (1985)).

Imagine a particle with sticking probability pP approaching a surface with sticking probability, ps. As a first approach, the net mass fraction of impacting particles that sticks to a surface was expressed mathematically by Walsh et al. (1990):

5 fstiek = Pp (Tp )+ [1-Pp (Tp )]-Ps (Ts )-ke • [1-Pp Op )]• [1-Ps Os )] (9)

where TP and Ts are the particle and surface temperatures and kg is a factor describing the erosion effect in the particular system (Xie (1995)).

The first term in eqn. (9) represents the probability that in-coming particles are sticky and will adhere to the surface. The second term describes the probability of non-sticky particles to be caught by a sticky surface. The third term represents erosion, which may occur when non-sticky particles impact on a non-sticky surface (Xie (1995), Meuronen (1997)).

Richards et al. (1993) generalized eqn. (9) to apply for a particle population, by adding subscript i to the particle sticking probability, Pp, and then summing over all particle classes.

A key parameter in eqn. (9) is the sticking probability, p,-, for the particle(s) (i=P) and the surface (i=S). A particle approaching the surface of a cylinder in a flue gas channel will possess a well-defined kinetic energy, EkinP. The probability that the particle sticks upon impact depends on the dissipation of its kinetic energy in the surface of the cylinder (or the deposit). In case the kinetic energy of the particle is larger than a certain reference value, Ekin ref , the particle will rebound and be entrained by the flue gas passing the cylinder. Otherwise, in case Ekin P < Ekin-ref , the particle will stick to the surface.

Danecke (1971,1972), Davis et al. (1986) and Barnocky and Davis (1988a,b) carried out a number of pioneering experiments dropping small metal and plastic spheres upon a smooth quartz surface overlaid with a thin layer of viscous fluid. The parameters that were varied include the fluid layer thickness (10-70 pm) and viscosity (20 - 389 P), the ball diameter (0.8 - 3.2 mm), density (1.2 and 7.9 g/cm 3), and elastic properties (Youngs modulus of elasticity, E: 3.0 x 109 - 2.1 x 1011 N/m2, Poissons ratio: 0.28 - 0.4). Although, some of these parameters, and particularly the particle size is out-of-range compared to impaction of fly ash particles on a tube in a flue gas channel, the experiments clearly demonstrated the effect of the liquid coating on the quartz plate.

Walsh et al. (1990) related the dissipation of kinetic energy upon collision to the material viscosity and described the sticking probability as a function of temperature, T, and compositional dependence, x, ie.

(10) 1forq, CTlref

Thus, the value of the sticking probability depends on the speciation of a reference viscosity, below which the particles or deposit is assumed to be

6 perfectly sticky. There is an analogy between the reference viscosity and the reference kinetic energy of the colliding particle described above. The reference viscosity may either be used as an adjustable (fitting) parameter or may be determined from laboratory experiments with well-defined gas particle and surface temperatures (Srinivasachar et al. (1990)). The distinct advantage of eqn. (10) is its simplicity. However, the sticking probability is sensitive to the choice of reference viscosity, a choice which is far from being easy or clear. Walsh et al. (1990) used a reference viscosity of 80 P, Richards et al. (1993) used a value of 105 P, while Srinivasachar et al. (1990) measured the critical viscosity need for attachment of coal derived fly ash particles to 108 P. In order to describe sticking of fly ash from Danish coal- fired utility boilers, a value of 80 P is appropriate in eqn. (10).

Vargas et al. (1997) provided an overview of existing models for estimation of viscosity of Si-rich ashes. The models may be divided into three groups: 1) models based on tabulated values (Bottinga and Weill (1972)), 2) Urbain type viscosity models (Urbain et al. (1981), Streeter et al. (1984) and Kalmanovitch and Frank (1988)), and 3) Arrhenius type viscosity models (Watt and Fereday (1968) and Greenberg (1984)). The six models were tested on measured viscosity data from a suite of 12 worldwide high temperature coal ashes. The Kalmanovitch and Frank (1988) model provided good estimates of the measured viscosities for ashes with more than 50 %(w/w) Si02, while the Greenberg (1984) model provided good estimates of the measured viscosities for ashes with more than 25 %(w/w) Al203. The Urbain (1981) viscosity model provided satisfactory estimates for all the ashes included in the study.

The model for quantification of inertial impaction was tested on fly ash compositional data from a number of Danish and foreign power stations fired with coal, coal blends and co-fired with coal and straw. An example of output from the model is given in Figures 2 and 3. Based on proximate and ultimate analyses of the feedstock mixtures utilized in the Midtkraft-Studstrup Power Station, Unit 1, Demoprogramme, experiments no. 3-5 (covering 0 - 20 % straw on an energy basis) the amount and composition of the flue gas is calculated following the procedure outlined by Frandsen (1995). This gives a fixed flue gas composition as a function of temperature and pressure. A change in the flue gas composition as a function of temperature and pressure, can be introduced by assuming equilibrium in the gas phase as recommended by Frandsen (1994, 1995). The mass concentration of residual ash, cra (g/m 3), based on conversion of 1 kg of fuel is determined from:

C^IOOOii m) vfg where fash is the percentage of ash on an as received weight basis (from the proximate analysis), p is the fraction of ash ending up as residual ash, and Vfg is the volume (m3) of flue gas (temperature dependent) determined assuming validity of the ideal gas law.

7 The impaction probability and stickiness of the fly ash particles are calculated based on CCSEM analyses (see Frandsen et al. (1998)). For each size bin and chemistry in the CCSEM scheme, the impaction coefficient is calculated using eqn. (7) and the coefficients of Israel and Rosner (1983), while the sticking coefficient in eqn. (1) is calculated using the generalized version of eqn. (9) (Richards et al. (1993)). The sticking probabilities are calculated using eqn. (10) utilizing the viscosity model of Urbain (1981). The necessary operational data are provided in Table 3. It is assumed that the particles and the gas have equal temperatures and the deposit is assumed dry. fash is calculated from fuel compositional data (Andersen et al. (1997)) and it is assumed that 3 = 0.9, ie. 90 % of the ash introduced to the furnace is entrained as residual fly ash, the rest is removed as bottom ash (no significant vaporization occurs).

In Figure 2, it is seen that the flux of sticking particles by this model is equal at 0 and 20 % straw share. According to Andersen et al. (1997), the amount of deposit is increased when the straw fraction fired is increased from 0 to 20 %. The reason for this discrepancy may be that the deposit is assumed dry in all simulations shown in Figure 2. Deposit analyses performed at GEUS have indicated that the content of K is significantly increased in the 20 % straw deposits as compared to the 0 % straw deposits. This increase in K-content could very well induce a sticky (liquid) condition on the surface of the 20 % straw deposits compared to the deposits from coal-firing (0 % straw). A sticky surface of the deposit is reported to cause an increase in the deposit formation rate (Walsh et al. (1990)).

A calculated inertial impaction ash deposition flux of 55 mg/m 2/s correspond to approximately 200 g/m 2/h. Based on the deposition measurements at the Midtkraft-Studstrup Power Station, Unit 1, Andersen (1998) estimated deposition fluxes in the range 0.9 - 86 g/m 2/h. Considering the simplicity of the model for inertail impaction applied in this work, and the fact that shedding is not taken into account, theoretical and experimental results agree very well.

In Figure 3, the flux of sticking particles in different sections (PLSH - platen superheater, SSH - secondary superheater, RH - reheater and UPSH - upper primary superheater) of the MKS1 boiler is simulated. The reason for the drop in flux from the PLSH to the UPSH is that the viscosity of the fly ash particles decreases significantly during cooling of the flue gas between the PLSH and UPSH. The exponential nature of the Urbain (1981) high-Si ash viscosity model is clearly seen in Figure 3.

2.3. Sticking of Biomass Fly Ash Particles

The estimation of reliable viscosities is the major draw-back of eqn. (10) when applied on biomass ashes. All the existing models for ash viscosity, as a function of temperature and composition, is fitted to coal ash compositional data, ie. ashes with 35 - 55 %(w/w) Si02, 10 - 25 %(w/w) AI2Os and < 4 %(w/w) K20. For this reason, the viscosity models cannot be used to estimate

8 sticking probabilities of biomass ashes, containing only small amount of Al and 40 - 80 %(w/w) KCI (Michelsen (1996a,b)).

Thus, for the biomass ashes another concept for estimation sticking probabilities are needed. The fraction of melt may be determined from theory and/or experiments. In Figure 4, melt curves based on theory (the appropriate binary phase diagram) and Simultaneous Thermal Analysis of a mixture of KCI and K2S04 (44 mole-% KCI), are shown.

Based on the STA-curve Figure 4, app. 80 % of a particle composed of KCI- K2S04 (44 mole-% KCI) will be molten at 800 °C. Thus, pP(TP) = 0.80, in eqn. (9). A similar analysis could be made on a low or high temperature laboratory ash prepared from the fuel of interest or on a fly ash collected at the power station of interest. In case no information is available about the conditions (solid/liquid) at the surface of the tube, ps(Ts) = 0.0 may be applied in eqn. (9).

Hansen (1997) and Hansen et al. (1997) tried to apply the fraction of melt in a biomass ash directly as sticking coefficient, the basic idea being that the dissipation of the kinetic energy of the in-coming particles can be correlated to the fraction of melt in the ash particles. The conclusion of the work of Hansen (1997) and Hansen et al. (1997) is that a model based on inertial impaction alone or a serial combination of models for condensation (building up eg. a 20 pm thick layer of condensed KCI) and inertial impaction (based on bulk ash chemistry of the fly ash) over-predict the deposition flux measured in the superheater section of straw-fired utility boilers, as long as the fraction of melt in a biomass fly ash is applied directly as sticking coefficient.

As noticed above and seen in Table 1, inertial impaction is by far the fastest of the three transport mechanisms considered by Sarofim and Helble (1994). Anyhow, Christensen (1995) reported very high aerosol mass loadings, > 1000 mg/Nm 3, in Danish utility boilers, fired with wheat, barley and rape, which may lead to a significant deposit formation by thermoforesis.

Thus, it was considered important to be able to rank thermoforetic deposition fluxes in biomass-fired boilers. Below, as the starting point, a simplified model for quantification of the thermoforetic mass flux, is outlined.

3. Submicron Ash Transport: Thermoforesis

Thermal gradients either within particles or in the suspending gas can be responsible for motion of aerosol or submicron fly ash particles by creating forces acting on the individual particles. Consider molecular motion in a temperature gradient. The movements will be more vigorous at higher temperatures. When a particle is placed in this gradient, the momentum transferred to the 'hot' side of the particle exceeds that transferred to the 'cold' side of the particle, thereby causing a net force on, and movement of the particle. This effect may be very important for transport of submicron ash particles in the temperature gradient outside a steam tube or probe in the flue

9 gas channel of a utility boiler.

In order to determine the net force exerted by a temperature gradient on a particle, it is necessary to know the velocity distribution of the gas molecules at the particle surface. Depending on the value of the Knudsen number (see eg. Friedlander (1977)):

2-I Kn = —(12) dp

the particle itself can have very little, or significant, influence on the velocity distribution of nearby gas molecules. In eqn. (12), lg is the mean free path of the gas and d P is the diameter of the (spherical) particle. The mean free path, lg , can be calculated from kinetic theory of gases (Flagan and Seinfeld (1988)).

The thermoforetic mass deposition flux, NjheXg/rr^s), is defined as:

NTher = UT'Csa 0^) where uT (m/s) is the thermoforetic velocity and csa (g/m 3) is the mass concentration of submicron ash particles

In eqn. (13), the product uT-csa represents the total flux (g/m 2s) of submicron ash particles, i.e. a mass (g) of ash particles transported through a unit area (m2) in a unit time.

3.1. Models for the Thermoforetic Velocity, uT

The critical parameter in eqn. (13) is the thermoforetic velocity, uT. Several authors have suggested equations for calculation of uT. Consider a particle with diameter dp, in a gas at temperature Tg (K) outside a tubular cylinder with surface temperature Tc (K). The temperature gradient above the cylinder is VT (K/m).

For small particles with Kn » 1, Flagan and Seinfeld (1988) suggested:

VT uT = -Th-— (14) Pg Tc where the dimensionless group Th is predicted theoretically to lie between 0.42 and 1.5, with experimental measurements in a narrow range around 0.5. In eqn. (14), |ig (g/ms) is the gas viscosity and pg (g/m 3) is the gas density.

Epstein (1929) used the slip formula proposed by Maxwell and suggested the following expression for the thermoforetic velocity, uT:

10 3 1 n VT uT= —4 14^72)'^ (15) where vg (m2/s) is the kinematic gas viscosity and C is the ratio of the thermal conductivity of the particle, XP, to that of the suspending gas, Xg , ie.

Xp C = (16) ^9

In eqn. (15), Cc is the Cunningham slip factor, that may be calculated from (Flagan and Seinfeld (1988)):

1.10 Cc = 1+Kn- 1.257+0.40-exp (17) " Kn

In eqn. (17), Kn is the Knudsen number as defined above.

Brock (1962) suggested:

VT uT =-B-vg -Cc (18) Tc where all the variables have been defined above, except for the constant, B: 1+Ct • Kn-C B = (19) 1+3-Cm -Kn 1+Ct-KnC+(C/2)

Ivchenko and Yalamov (1971) estimated Cs = 1.147, while Peterson et al. (1989) modified the original values of Cm and Ct suggested by Brock (1962) to: Cm = 1.146 and Ct = 2.20. The variables Cc and C used in eqn. (18-19) are defined above.

Derjaguin and Yalamov (1965) gave the following expression for the thermoforetic velocity, uT:

1 4+(C/2)+Ct-Kn-C VT 2 1+(C/2)+Ct Kn-C Vg C TC (20) where all the variables have been defined above.

Finally, for small particles, with Kn » 1, Friedlander (1977) gave for the thermoforetic velocity, uT:

n 3 v g VT uT (21) 4 7r-a t Tc 1+ 8

while Rosner (1986) suggested for the same Kn-range, ie. Kn » 1:

uT = (22)

3.2. An Example: Calculation of uT

Consider a 0.1 pm spherical particle of KOI placed close to the surface of a 800 K tubular cylinder. Assuming temperature gradients of 250, 500 and 1000 K/cm, and applying the data given in Table 4, calculate the thermoforetic velocity of the particle, using a number of the equations outlined above.

First, consider the Epstein (1929) equation:

3 1 n VT uT •VCc'^ (15) 4 1+(C/2) ■c

in which C = 410.71, from equation (16). The value of Cc is calculated from eqn. (17), Cc = 4.591. The value of the Knudsen number, Kn, is calculated from eqn. (12) (Kn = 2.38). Based on the data given above, at 500 K/cm, the thermophoretic velocity is estimated to: uT = 0.013 m/s.

Next, consider the Brock (1962) equation:

VT uT=-B-vg.C c~ (18) *c in which, Cs = 1.147 (Ivchenko and Yalamov (1971)), and Cm = 1.146 and Ct = 2.20 (Peterson et al. (1989)). The remaining parameters in the Brock equation is given above. At 500 K/cm, uT = 0.403 cm/s.

Using the Derjaguin and Yalamov (1965) equation:

1 4+(C/2)+Ct-Kn-C VT 2'l+(C/2)+Ct-Kn-c'V c>c (20)

it is found that, uT = 1.77 cm/s, at 500 K/cm.

For small particles with large Knudsen numbers, ie. Kn » 1, the Friedlander (1977) equation, eqn. (21), assuming a t = 0.92, as suggested by Peterson et al. (1989), gives uT = 0.424 cm/s, at 500 K/cm. The Rosner (1986) and Flagan and Seinfeld (1988) equations, eqn. (22) respectively eqn. (14), gives

12 0.578 cm/s respectively 0.385 cm/s, at 500 K/cm.

A summary of the estimated thermoforetic velocities are provided in Table 5.

As may be seen in Table 5, the Brock (eqn. (18)), Friedlander (eqn. (21)), Rosner (eqn. (22)) and Flagan and Seinfeld (eqn. (14)) equations give similar results, while the Epstein (eqn. (15)) and the Derjaguin and Yalamov (eqn. (20)) equations give significant lower respectively higher thermoforetic velocities, uT. This is not surprising, since the Brock, Friedlander, Rosner and Flagan and Seinfeld equations are similar. For the gas-particle interactions and temperature gradients occuring in utility boilers, the Brock (1962) and Friedlander (1977) equations will be used to determine thermoforetic velocities

4. Ash Deposition in Straw-Fired Utility Boilers

Stenholm et al. (1996) and Jensen et al. (1997) have investigated the combustion of twelve well-defined batches of different straws (wheat, barley, rape) at two small combined heat and power production (CHR) boilers in Slagelse and Haslev, Denmark.

Slagelse CHR is a 31 MWth (8 MW electricity and 20 MW heat) grate-fired unit producing 40 tons of steam per hour, at 65 bar and 440 °C. Bales of straw are shredded before entering the combustion chamber through a screw-feeder. The straw burns out on a sloping grate. Deposition measurements were carried out at two locations using air-cooled probes, in the upper half of the furnace and in the third pass, the latter probe being located between the primary and the secondary superheater. The mean gas temperature during the wheat and barley experiments was 873 °C for the furnace location and 647 °C for the superheater location, see Figure 5.

Haslev CHR is a 23 MWth (5 MW electricity and 13 MW heat) grate-fired unit producing 26 tons steam per hour at 67 bar and 440 °C. Bales of straw are fed to the boiler, which is equipped with four cigar burners implying that the straw bales burn on the front side, from one end to the other as the bales are fed into the combustion chamber. The final burnout takes place on a sloping grate. Deposition measurements were carried out at two locations using air ­ cooled probes: in the top of the furnace and at the entrance to the third pass, just in front of the superheaters. The mean gas temperatures at the two locations were approximately 835 °C and 650 °C, respectively, see Figure 6.

Each experiment at the Slagelse and Haslev CHPs lasted approximately eight hours, and the parameters measured were among others local gas temperatures, exit flue gas composition and aerosol particles in the flue gas. In addition, detailed analyses of the straw, fly and bottom ashes and deposit samples were conducted (Jensen et al. (1997)). For both wheat and barley straw, the extent of deposit formation on the inserted probes could be correlated to the content of potassium in the straw fuel. In all experiments

13 firing wheat and barley straw, a much faster build-up of hard deposit was seen in the furnace compared to the superheaters.

During the experiments it was found, that the total deposition flux (ie. hard + loose deposit) was reasonably independent of time, whereas the fraction of hard deposit increased with time. The fraction of hard deposit correlated to the sulphur content of the deposit, the gas temperature and the amount of potassium in the straw, which could indicate that the formation of K2S04 has a strong influence on the formation of hard deposit (Jensen et al. (1997)). The composition of superheater deposits were found to compare to the fly ash composition, whereas the furnace deposits were enriched in Si, Ca and Mg. Detailed Computer Controlled Scanning Electron Microscopy (CCSEM) analyses of fly and bottom ashes and deposits from Slagelse and Haslev CHPs are provided by Hansen et al. (1997) and Sorensen (1997). From Environmental Scanning Electron Microscopy (ESEM) combined with Energy Dispersive X-Ray (EDX) analysis, the inner and outer layer of the deposits were found to be alike, both consisting primarily of K, S and Cl which correlates with the aerosol findings in the experiments (Christensen (1995)).

The aerosol measurements have shown formation of high concentrations of submicron aerosols, 195 - 960 mg/Nm 3 during wheat-firing, and 1010 -1660 mg/Nm3 during barley-firing. In the rape experiment at Slagelse CHP, a submicron aerosol concentration of more than 2000 mg/Nm 3 was measured (Christensen (1995), Christensen et al. (1997)). The aerosols consisted almost solely of K, Cl and S.

Based the experimental data provided in Table 6, simple linear correlations between aerosol mass loading (mg aerosol/Nm 3) and the K-content on a dry fuel base can be set up for the wheat-fired experiments at Haslev and Slagelse: mg aerosol/Nm 3 in the Haslev CHP, Csa , during wheat-firing:

Csa = -167.7 + 876.5 • % K (23) mg aerosol/Nm 3 in the Slagelse CHP. Csa , during wheat firing:

Csa = - 29.85 + 347.9 • %K (24) where %K is the content of potassium in the wheat on a dry fuel base. Eqn. (23) for the Haslev boiler is fitted by use of conventional least squares minimization, using the three experimental data points corresponding to wheat-firing in the boiler, ie. experiments no. 1,2 and 4. Eqn. (24) for the Slagelse boiler, is fitted by use of five experimental points, ie. experiments no. 1,2, 5, 6 and 7.

Based on the aerosol mass loadings, the metal surface and mean gas temperatures (Table 6, Figures 5-6), a thermoforetic velocity (calculated using

14 one of the Brock (1962), Friedlander (1977) or Rosner (1986) equations) and eqn. (13), a thermoforetic mass flux of aerosols can be calculated and compared to the measured mass flux (Jensen et al. (1997)).

Calculated thermoforetic mass fluxes has been compared with measured total superheater deposition fluxes from the Haslev and Slagelse experiments. The thermoforetic velocity was determined for a 0.1 pm aerosol in a 500 K/cm and a 700 K/cm thermal gradient, using the Brock (1962) equation (eqn. (18)). The size of the aerosol (0.1 pm) is arbitrary chosen, while the choice of thermal gradients of 500 and 700 K/cm outside the tubular cylinder is based on evaluations made by Zdravkovich (1997) and Hvid and Walther (1998) as part of the Interflow D project at the Danish Maritime Institute.

Figure 7 shows data from the wheat-fired experiments at Haslev and Slagelse, ie. experiments no. 1,2 and 4 at Haslev, and experiment no. 1,2,5, 6 and 7 at Slagelse. This figure is based on aerosol mass loadings calculated from eqn. (23-24).

5. Summary and Concluding Remarks

A simple engineering model for thermoforetic deposition of submicron ash particles in biomass (straw) fired utility boilers has been set up. Notice in Figure 7, that for deposition fluxes below 15 g/m 2/h, this simple model give reasonable deposition fluxes compared to experimental data. In systems with higher deposition fluxes the thermoforetic model under-predicts the deposition flux. One reason for this may be that in these systems inertial impaction play a major role in the deposition of ash particles on heat transfer surfaces. As stated by Hansen (1997), a model based on pure impaction will highly over ­ predict the deposition flux. The ideal case may be a serial or parallel combination of models for thermoforetic and inertial impaction ash transport.

The thermoforesis model applied in this work need to be improved in many points. First, a model for estimation of the thermal gradient above a deposit coated heat transfer tube is needed. Such a model should focus on heat transfer to a tube from a hot surrounding gas and on the thickness of the laminar boundary layer above the tube (deposit) surface. A simple empirical model for this thickness as a function of the fluid dynamic (flow regime) in the system will be tested in near future.

Another point that need improvement is the correlation between the fuel composition and the composition and size of the aerosols. In this study experimental data from a previous work lead to a correlation between the K- content in the fuel and the aerosol mass loading in the Haslev and Slagelse boilers. A general model following the guidelines provided by Frandsen (1994) and Christensen (1995) may solve this problem.

15 Acknowledgement

This work was carried out as part of the CHEC Research Programme at the Department of Chemical Engineering, Technical University of Denmark. The CHEC Research Programme is co-funded by Elkraft, Elsam, The Danish and Nordic Energy Research Programmes, The Danish Technical Reserach Council and the European Union.

Specific funding from the Danish Energy Research Programme, Contract No. 1323/95-0007, is greatly acknowledged, as is many useful discussions on inertial impaction and thermoforetic transport of ash particles with Sean Allan and Dr. Tom Erickson, Energy and Environmental Research Center, University of North Dakota, USA.

16 References:

S. Ahila and S. R. Iyer: “Fireside Corrosion of Superheaters and Reheaters in Coal Fired Boilers ”; Tool & Alloy Steels, 26, 45, 1992.

K. H. Andersen . F. J. Frandsen, P. F. B. Hansen and K. Dam-Johansen, “Full- Scale Deposition Trials at a 150 MWe PF-Boiler Co-Firing Coal and Straw: Summary of Results ”, Proc. Eng. Found. Conf. ‘Impact of Mineral Impurities in Solid Fuel Combustion ’, Kona, Hawaii, November 2-7, 1997.

K. H. Andersen : “Deposit Formation in Coal-Straw Co-Combustion ”, Industrial Ph.D.-Thesis, Midtkraft I/S, J. No. EF-582, 1998.

G. Barnocky and R. H. Davis . J. Colloid Interface Sci., 121, 226, 1988a.

G. Barnockv and R. H. Davis . Physics of Fluids, 31(6), 1324, 1988b.

L. L. Baxter . M. F. Abbott and R. E. Douglas: “Dependence of Elemental Ash Deposit Composition on Coal Ash Chemistry and Combustor Environment ”; In: Inorganic Transformations and Ash Deposition During Combustion, S.A. Benson (Ed.), Proc. Eng. Found. Conf., Palm Coast, FA, USA, March 10-15, 1991; 1992.

L. L. Baxter and D. R. Hardestv : “The Fate of Mineral Matter during Pulverized Coal Combustion ”; In: “Coal Combustion Science ”, Sandia Quarterly Progress Report, April - June, 1992, SAND92-8227, 1992.

L. L. Baxter : “Ash Deposition During Biomass and Coal Combustion: A Mechanistic Approach ”; Biomass. Bioenergy, 4, 85,1993.

S. A. Benson and P. L. Holm : “Comparison of Inorganic Constituents in three low-rank coals ”, Ind. Eng. Chem. Proc. Res. Dev., 24, 145, 1985.

S. A. Benson . M. L. Jones and J. N. Harb: “Ash Formation and Deposition ”; In: Fundamentals of Coal Combustion - For Clean and Efficient Use, L. D. Smoot (Ed.); Coal Science and Technology, Vol. 20; Elsevier, Amsterdam, 1993.

R. W. Bono, A. A. Levasseur, 0. K. Chow, L. S. Miemic: “Ash Deposition: A Boiler Manufacturers Perspective ”, In: S. A. Benson (Ed.), Proc. Eng. Found. Conf. ‘Inorganic Transformations and Ash Deposition during Combustion ’, Palm Coast, FA, March 10-15, 1991, ISBN 0-791 8-0657-X, 1992.

Y. Bottinaa and D. F. Weill, Am. J. Sci., 272, 438, 1972.

J. R. Brock . J. Col. Sci., 17, 768, 1962.

17 K. A. Christensen : “The Formation of Submicron Particles from the Combustion of Straw ”; Ph.D. Thesis, Dept. Chem. Eng., Techn. Univ, of Denmark, 1995.

K. A. Christensen . M. Stenholm and H. Livbjerg, “The Formation of Submicron Aerosol Particles, HCI and S02 in Straw-Fired Boilers ”. To appear in J. Aerosol Sci., 1997.

B. Dahneke . J. Colloid Interface Sci., 37, 342,1971.

B. Dahneke . J. Colloid Interface Sci., 40, 1, 1972.

R. H. Davis . J.-M. Serayssol and E. J. Hinch, J. Fluid Mech., 163, 479, 1986.

B. V. Derjaguin and Y. Yalamov . J. Col. Sci., 20, 555, 1965.

P. Epstein . Z. Phys., 54, 537,1929.

R. C. Flagan and J. H. Seinfeld . “Fundamentals of Air Pollution Engineering ”, Prentice-Hall Inc., New Jersey, 1988.

F. J. Frandsen : “The Stability of Aerosols - Prediction of Homogeneous Nucleation ”, Research Progress Report, Energy and Environmental Research Center, University of North Dakota, Grand Forks, North Dakota, June 1994.

F. J. Frandsen : “Trace Elements from Coal Combustion ”, Ph.D.-Thesis, Dept. Chem. Eng., Techn. Univ. of Denmark, ISBN 87-90142-03-9, 1995.

F. J. Frandsen . L. A. Hansen, H. P. Michelsen, K. Dam-Johansen: “Alkali Metal Partitioning in Hot Flue Gases - Aspects of Equilibrium Modeling ”, Proc. 9th Eur. Bioenergy Conf., Copenhagen, Denmark, June 24 -27,1996.

F. J. Frandsen : “Empirical Prediction of Ash Deposition Propensities in Coal- Fired Utilities ”; CHEC. Rep. No. 9703, January 1997a.

F. J. Frandsen : “Ash Chemistry Mapping ”; EU Joule 3 Coal Blends Project 12- Monthly Progress Report, January 1997b.

F. J. Frandsen . L. A. Hansen, H. S. Sorensen and K. Hjuler: “EFP-95 Project: Characterization of Biomass Ashes - Data Compilation ”, CHEC Report No. 9804, February 1998.

S. K. Friedlander . “Smoke, Haze and Dust”, Wiley, New York, 1977.

S. Greenberg : “Viscosity of Synthetic and Natural Coal Slags ”, Project Review Report, Univ. IL, 1984.

18 L. A. Hansen : “Melting and Sintering of Ashes”; Ph.D.-Thesis, Dept. Chem. Eng., Techn. Univ. Denmark, 1997.

L. A. Hansen . F. J. Frandsen, H. S. Sorensen, P. Rosenberg, K. Hjuler and K. Dam-Johansen, K.: “Ash Fusion and Deposit Formation at Straw Fired Boilers ”, Proc. Eng. Found. Conf. ‘Impact of Mineral Impurities in Solid Fuel Combustion ’, Kona, Hawaii, November 2-7, 1997.

J. N. Harb and E. E. Smith: "Fireside Corrosion in PC-Fired Boilers ”; Prog. Energy Combust. Sci., 16, 169, 1990.

L. Y. Huang . J. S. Norman, M. Pourkashanian, A. Williams: “Prediction of Ash Deposition on Superheater Tubes from Pulverized Coal Combustion ”, FUEL, 75(3), 271, 1996.

S. L. Hvid and J. H. Walther : “Interflow Part D: Final Report ”, Danish Maritime Institute, 1998.

R. Israel and D. E. Rosner : “Use of a Generalized Stokes Number to determine the Aerodynamic Capture Efficiency of Non-Stokesian Particles from a Compressible Gas Flow ”; Aersosl Sci. and Techn.; 2; 45; 1983.

I. N. Ivchenko and Y. Yalamov . Russian J. Phys. Chem., 45, 317, 1971.

N. S. Jacobson . M. J. McNallan and E. R. Kreidler: “High-Temperature Reactions of Ceramics and Metals, with Chlorine and Oxygen ”; High. Temp. Sci., 27, 381, 1990.

P. A. Jensen . M. Stenholm and P. Hald “Deposition Investigation in Straw- Fired Boilers ”, Energy and Fuels, 11(5), 1048, 1997.

H. Junker . “Cofiring Biomass and Coal ”, Ph.D.-Thesis, Aalborg University, 1997.

P. D. Kalmanovitch and M. Frank : “An Effective Model of Viscosity for Ash Deposition Phenomena ”, In: R. W. Bryers and K. S. Vorres (Eds.), Proc. Eng. Found. Conf. on ‘Mineral Matter and Ash Deposition from Coal ’, Santa Barbara, CA, 1988.

V. Meuronen : “Ash Particle Erosion on Steam Boiler Convective Section ”, Lappeenranta University of Technology, Ph.D.-Thesis, 1997, ISBN-951 -764- 181-8.

H. P. Michelsen : “Results of Deposit Measurements at Rudkobing KW”, CHEC Rep. No. 9603, January 1996a.

H. P. Michelsen : “Results of Deposit Measurements during Straw Firing at the Kyndby Power Station ”, CHEC Rep. No. 9604, January 1996b.

19 H. P. Michelsen . O. H. Larsen, F. J. Frandsen, K. Dam-Johansen: “Deposition and High Temperature Corrosion in a 10 MW Straw Fired Boiler", Proc. Eng. Found. Conf. ‘Biomass Usage for Utility and Industrial Power", Snowbird, UT, April 28 - May 3,1996.

H. P. Nielsen : “Deposits from Straw Combustion - Experiments at Sandia ’s Multi Fuel Combustor", CHEC Report No. 9707, September 1997.

T. W. Peterson . F. Stratmann and H. Fissan, J. Aerosol. Sci., 20, 683,1989.

E. Raask : “Mineral Impurities in Coal Combustion ”; Hemisphere Publishing Corporation, Springer-Verlag, Berlin, 1985.

G. H. Richards . P. N. Slater and J. N. Had): “Simulation of Ash Deposit Growth in a Pulverized Coal-Fired Pilot Scale Reactor ”, 7,774,1993.

D. E. Rosner : “Transport Processes in Chemically Reacting Flow ”; Butterworth, London, 1986.

D. E. Rosner . A. G. Konstandopoulos, M. Tassopoulos and D. W. Mackowski: “Deposition Dynamics of Combustion-Generated Particles: Summary of Recent Studies of Particle Transport Mechanisms, Capture Rates and Resulting Deposit Microstructure/Properties ”; In: Inorganic Transformations and Ash Deposition During Combustion, S.A. Benson (Ed.), Proc. Eng. Found. Conf., Palm Coast, FA, USA, March 10-15, 1991; 1992.

A. F. Sarofim and J. J. Helble : “Mechanisms of Ash and Deposit Formation ”; In: The Impact of Ash Depositions on Coal Fired Power Plants, J. Williamson and F. Wigley (Eds.), Proc. Eng. Found. Conf., Solihull, UK, June 20-25, 1993; Taylor & Francis Ltd., London, UK, 1994.

J. S. Serafini . NACA Rep. 1159, 1954.

S. Srinivasachar . J. J. Helble, C. B. Katz and A. A. Boni: “Transformation and Stickiness of Minerals during Pulverized Coal Combustion ”, Proc. Eng. Found. Conf. on ‘Mineral Matter and Ash Deposition from Coal ’, R. W. Bryers and K. S. Vorres (Eds.), Eng. Found. Press, 1990.

M. Stenholm . P. A. Jensen and P. Hald, “The Fuel and Firing Characteristics of Biomass - Combustion Trials ”, Final Report, EFP Project No. 1323/93- 0015, 1996 (In Danish).

R. T. Streeter . E. K. Diehl, H. H. Schobert: “Measurement and Prediction of Low-Rank Coal Slag Viscosity ”, In: H. H. Schobert (Ed.) ACS Symposium Series, 264, 195, 1984.

G. Urbain . F. Cambier, M. Deletter, M. R. Anseau, Trans. J. Br. Ceram. Soc., 80, 139, 1981.

20 S. Vargas . F. J. Frandsen and K. Dam-Johansen: “ELSAM-ldemitsu Kosan Cooperative Research Project: Performance of Viscosity Models for High- Temperature Coal Ashes”, CHEC Report No. 9719, 1997.

T. F. Wall . S. P. Bhattacharya, D. K. Zhang, R. P. Gupta and X. He: “The Properties and Thermal Effects of Ash Depositions in Coal Fired Furnaces: A Review ”; In: The Impact of Ash Depositions on Coal Fired Power Plants, J. Williamson and F. Wigley (Eds.), Proc. Eng. Found. Conf., Solihull, UK, June 20-25, 1993; Taylor & Francis Ltd., London, UK, 1994.

P. M. Walsh. A. N. Sayre, D. 0. Loehden, L. S. Monroe, J. M. Beer, A. F. Sarofim: “Deposition of Bituminous Coal Ash on an Isolated Heat Exchanger Tube: Effects of Coal Properties on Deposit Growth”, Prog. Energy Combust. Sci., 16, 327, 1990.

J. D. Watt and F. Feredav . J. Inst. Fuel, 41, 99, 1968.

R. A. Wessel and J. Righi: “Generalized Correlations for Inertial Impaction of Particles on a Circular Cylinder ”; Aerosol Sci. and Techn.; 9; 29; 1988.

J. Xie : “Erosion of Heat Exchanger Tubes in the Convective Section of an Industrial Boiler by By Products of Coal Combustion", Department of Materials Science and Engineering, Penn State University, Ph.D. Thesis, 1995.

M. M. Zdravkovich : “Flow Around Circular Cylinders ”, Oxford University Press, 1997.

21 Figures Gas Flow Direction

Captured particle

Impacting Eddies particles

Rebounding particle

Figure 1: Impaction vs. rebounding of residual fly ash particles on a tubular cylinder surface.

60 I 50 £ 40 f 30 a 20 £ 10 “■ 0 0 % 10 % 20 % Straw Straw Straw Feedstock Mixture

Figure 2: Calculated flux of sticking fly ash particles on a single 38 mm outer diameter cylindrical tube in a position equal to the platen superheater, in experiment no. 3 (20 % straw on an energy base), no. 4 (10 % straw) and no. 5 (0 % straw) of the Demoprogramme at the Midtkraft-Studstrup Power Station, Unit No. 1. For further details on the demoprogramme: See Andersen et al. (1997). 60

PLSH SSH RH UPSH Boiler Profile

Figure 3: Calculated flux of sticking fly ash particles in different sections of the Midtkraft-Studstrup Power Station, Unit 1, boiler, during experiment no. 3 (20 % straw on an energy base) of the MKS Demoprogramme. PLSH - platen superheater, SSH - secondary superheater, RH - reheater and UPSH - upper primary superheater. For further details on the boiler configuration and operational conditions: See Andersen et al. (1997).

100 T

80 * •

60 . . ——Theory - - - STA-Curve

Temperature (C)

Figure 4: Determination of melt formation in a KCI-K2S04 binary mixture. The theoretical curve is based on the KCI-K2S04 binary phase diagram, while the experimental determination is based on Simultaneous Thermal Analysis (STA) of the binary mixture. Source: Hansen (1997). 873 °C, 510 °C

Figure 5: The Slagelse CHP straw-fired boiler. Source: Frandsen et al. (1997).

835 °C, 510 °C

650 °C, 510 °C

Figure 6: The Haslev CHP straw-fired boiler. Source: Frandsen et al. (1997). a500 K/cm a700 K/cm

Measured Flux (g/m2/h)

Figure 7: Calculated thermoforetic mass fluxes vs. measured total superheater deposition fluxes from the Haslev and Slagelse wheat-fired experiments. The thermoforetic velocity was determined for a 0.1 pm KCI aerosol in a 500 K/cm and a 700 K/cm thermal gradient, using the Brock (1962) equation. Tables Mechanism: Diffusion: Thermoforesis: Inertial Impaction: Time: 46 days 12 days 37 min.

Table 1: The time needed to build up a 2 mm uniform layer of deposit on the surface of a cylindrical tube by condensation, thermoforesis and inertial impaction. Source: Sarofim and Helble (1994).

Equation: a: b: c: d: Original 0.125 1.25 -0.014 0.00508 Modified - a 0.1425 1.28 0.00215 0.00587 Modified - b 0.1238 1.34 -0.034 0.0289

Table 2: Coefficients for use in eqn. (3-4). The original coefficients fitted by Israel and Rosner (1983) and the modified coefficients fitted by Baxter and Hardesty (1992): with (a) and without (b) thermoforesis.

Boiler section: PLSH: SSH: RH: UPSH: Flue gas temperature (°C) 1257 1112 915 563 Flue gas velocity (m/s) 17 6 6 6

Table 3: Temperature and flue gas velocity as a function of position in the boiler of the Midtkraft-Studstrup Power Station, Unit 1. For further details and a sketch of the boiler profile: see Andersen et al. (1997). Acronyms used: PLSH - PLaten Superheater SSH - Secondary Superheater RH - ReHeater UPSH - Upper Primary Superheater

Variable: Value (comment): ** 5.6 x 10"5 cal/cm/s/K (air at 298.15 K) Xp 2.3 x 10"2 cal/cm/s/K (KCI) Ig 0.119 pm (air at 1000 K) vQ 1.232 x 10"4 m2/s (air at 1000 K)

Table 4: Data used to estimate the thermoforetic velocity of a 0.1 pm KCI spherical particle in a temperature gradient outside a tubular cylinder. Theoretical relation used: Thermofor. velocity, uT (cm/s): Reference: Eqn. No. 100 500 1000 K/cm K/cm K/cm Epstein (1929) (15) 0.003 0.013 0.026 Brock (1962) (18) 0.081 0.403 0.807 Derjaguin and Yalamov (1965) (20) 0.35 1.77 3.54 Friedlander (1977) (21) 0.085 0.424 0.848 Rosner (1986) (22) 0.116 0.578 1.155 Flagan and Seinfeld (1988) (14) 0.077 0.385 0.770

Table 5: Summary of estimated thermoforetic velocities as a function of temperature gradient.

Experiment No.: Straw type: %K [db] ^sa (mg/Nm 3) Haslev -1 Wheat 0.71 454 Haslev - 2 Wheat 1.29 961 Haslev - 3 Barley 2.05 1500 Haslev - 4 Wheat 1.15 843 Slagelse -1 Wheat 0.41 75 Slagelse - 2 Wheat 0.97 388 Slagelse - 3 Barley 1.83 1011 Slagelse - 4 Barley 2.60 1661 Slagelse - 5 Wheat 1.86 601 Slagelse - 6 Wheat 2.07 667 Slagelse - 7 Wheat 0.69 194 Slagelse - 8 Rape 1.69 2094

Table 6: Straw compositional data (%K on a dry fuel base [db]) and aerosol mass loadings, Csa , (mg/Nm 3 (dry flue gas)) from full-scale experiments at Haslev and Slagelse CHPs. Source: %K-data is taken from Jensen et al. (1997). Csa -data is taken from Christensen et al. (1997).