An overview of slagging and fouling indicators and their applicability to biomass fuels

Fuel Processing Technology 217 (2021) 106804 Available online 13 March 2021 0378-3820/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Research article An overview of slagging and fouling indicators and their applicability to biomass fuels Jakub Lachman * , Marek Bal´ aˇ s, Martin Lisý, Hana Lis´ a, Pavel Milˇ c´ ak, Patrik Elbl Faculty of Mechanical Engineering, Brno University of Technology, 60190 Brno, Czech Republic ARTICLE INFO Keywords: Biomass combustion Ash fusion temperatures Slagging and fouling prediction ABSTRACT Slagging and fouling are common problems associated with biomass firing. The different nature of the mineral and phase composition of biomass ash makes the vast experience with coal firing insufficient for its translation to biomass fuels, especially when it comes to slagging and fouling behavior. Biomass tends to have lower ash content than coals; however, it is often rich in volatile alkalis. The mineral deposits found on boiler walls and superheater tubes are often comprised of alkali compounds. Numerous studies on ash melting and particle sticking behavior have been conducted. Laboratory observed ash fusion temperatures are commonly used to evaluate the slagging and fouling propensity of fuels. The tests are often time consuming, therefore several predictive indices have been developed to estimate the propensity based on the ash composition alone. Thermodynamic models as well as neural networks have also been applied to this end. However, for practical in the field purposes, the ash fusion tests and predictive indices are preferred because of their convenience. An overview of these indices is presented in this work. A sizeable dataset has been collected in order to statistically evaluate the applicability of the indices and of several AFT prediction formulas. General trends in ash composition on this extensive dataset have also been illustrated. Finally, a more convenient graphical solution is presented for preliminary slagging and fouling predictions. 1. Introduction With the growing demand for renewable energy sources, biomass and derived biofuels have been gaining increasing interest. Biomass combustion is carbon neutral, its utilization reduces the overreliance on fossil fuels and creates a revenue source for manufacturers. On the other hand, biomass combustion presents several technical obstacles. Gratefired biomass boilers are prone to severe slagging and fouling. Ash deposits form both inside the furnace as well as on the adjacent heat exchangers, e.g. superheater tubes. These agglomerates then hinder the effectiveness of the heat exchanger and directly influence high temperature chlorine induced corrosion by creating a reducing atmosphere on the surface of the tubes [1]. This can lead to unscheduled plant shutdowns due to mechanical failures. Both slagging and fouling are complicated processes and have been a subject of several studies, such as by Teiteira [2], Yao [3] and Torusamais´ e [4]. Kleinhans et al. [5] thoroughly covered the physical and chemical transformations in inorganic matter during combustion as well as the different ash deposition mechanisms. Slagging and fouling occur when the ash particles become molten and adhere to the boiler surface. Inorganic vapors may also condense on solid particles at lower temperatures, coating their surface in a sticky molten film. Agglomerates, typical for wood combustion, impact the walls or tubes and rearrange their structure. During this process, they dissipate all kinetic energy, which leads to the particle sticking to the surface rather than rebounding. Solid particles can then become entrapped in the porous deposit after its initial formation through these mechanisms. The melting point of the ash is therefore an essential indicator of the slagging and fouling tendency. Reducing the flue gas temperature at the furnace outlet can significantly affect the fouling tendency in the following superheaters. This, however, has a negative impact on the combustion quality and fuel burnout, especially when firing fuels with high volatile content, such as wood. Slagging can also inhibit oxidation of the fuel and lead to higher CO emissions as noted in [6]. Therefore, fuels with high ash fusion temperatures (AFTs) are preferred. * Corresponding author. E-mail addresses: [email protected] (J. Lachman), [email protected] (M. Bal´ aˇ s), [email protected] (M. Lisý), [email protected] (H. Lis´ a), Pavel. [email protected] (P. Milˇ c´ ak), [email protected] (P. Elbl). Contents lists available at ScienceDirect Fuel Processing Technology journal homepage: www.elsevier.com/locate/fuproc https://doi.org/10.1016/j.fuproc.2021.106804 Received 18 August 2020; Received in revised form 23 December 2020; Accepted 2 March 2021 Fuel Processing Technology 217 (2021) 106804 2 1.1. Ash fusion temperatures Laboratory tests are often carried out to estimate ash fusion temperatures. Several methods are prescribed in different standards and literature, both for sample preparation and AFT evaluation. Ash samples are obtained through controlled thermal decomposition of the fuel in a muffle furnace. The ash is then pressed, based on the standard, into a cone, a pyramid, a cylinder or a cube. The differences in behavior based on the selected shape has not yet been thoroughly studied. The produced specimen is then exposed to a uniform heating rate in either oxidizing or reducing conditions and the AFTs are monitored based on Fig. 1. It is important to note that the AFTs obtained in reducing conditions are generally lower [5] and thus not interchangeable with the AFTs obtained in oxidizing conditions. However, for samples low in Fe content, the difference is negligible [9]. Based on the American standard ash fusion test [7] (as also described in [5]), the characteristic temperatures are referred to as: - initial deformation temperature (IDT), - softening temperature (ST); h =w, - hemispherical temperature (HT); h =½w, - fluid temperature (FT). Where h represents the height of the specimen silhouette and w its width. The European standard ˇ CSN ISO 540 uses a cylindrical ash specimen and the individual AFTs are defined in the same way as in the American standard. While the American standard is commonly used both for coals and biofuels, the application of the European standard method has been restricted to coals only. Standards CEN/TS 15370–1 and CEN/TS 15404 are currently recommended for evaluating the AFTs of biomass and solid alternative fuels respectively. Apart from the differences in sample preparation, the characteristic temperatures have also been changed (see Fig. 1 below). The IDT of the original standard has been replaced by the SST (shrinkage starting temperature), which is defined as the temperature at which the area of the test piece falls below 95% of the original area due to shrinking of the test piece. This is because the shrinkage may be caused by sintering and therefore be the first sign of partial melting. However, shrinkage may also appear due to liberation of carbon dioxide and volatile alkali matter before partial melting even occurs. Moreover, samples enriched in silicates can show swelling instead. Some authors [8,9] have hence questioned the informative value of SST. According to [10], the following temperature ranges can be used when evaluating the slagging potential of fuels: - minor slagging occurs when the ST is above 1390 ◦C, - slight slagging for STs between 1250 ◦C and 1390 ◦C, - severe slagging for STs below 1250 ◦C. Boiler design manuals often suggest the temperature at the furnace outlet to be at least 100 ◦C lower than the ST [11,12] or 50 ◦C lower than the IDT [12] to minimalize fouling in the adjacent convective pass. Kleinhans [5] notes that the AFTs of individual ash particles found in industrial applications may vary as much as 250 ◦C from the experimentally obtained sample. This is due to the different conditions inside industrial boilers and the existence of small ash particles, that are not fully represented by the laboratory obtained specimen when it comes to their composition. Williams [13] also notes that the formation of the laboratory ash sample is much slower than the ash formation inside the boiler, making the comparison between the two questionable. Nonetheless the experimentally determined AFTs are commonly used by boiler designers and plant operators [5,11,12]. Other criteria based on the ash composition or other fuel properties are sometimes used to evaluate the slagging and fouling potential or to expand the precision of the AFT predictions. These will be discussed later. 1.2. Ash composition The ash content of biomass varies substantially between the different plants, but is generally lower than the ash content of coals. Experimental results show that the ash content of biomass generally ranges from 0.5% to 20%. [14] Even though the laboratory obtained ash samples are formed under different conditions than industrial ash, according to [15] the ash formation mechanisms and phase-mineral transformations are similar. Ash comprises a wide variety of inorganic matter. The elements found in ash are reported as the weight percentage of their stable oxides, such as SiO 2 , CaO, K 2 O, P 2 O 5 , Al 2 O 3 , MgO, Fe 2 O 3 , Na 2 O, SO 3 and TiO 2 . Though the elements commonly found in biomass and coal ash are the same, the ash fusion temperatures differ considerably. Moreover, the oxide content within several samples of the same plant can also vary noticeably, as observed by Yu et al. [10]. The melting temperatures of the different inorganic phases and minerals identified in ash [15] span hundreds of degrees Celsius. Furthermore, the eutectics formed by the minerals melt at lower temperatures than the individual minerals [16]. Therefore, the AFTs largely depend on the actual ash composition, the ratios between certain elements and their phase and form. It has been observed that the acidic oxides tend to increase the AFTs while the basic oxides decrease it [17]. The lower AFTs of biofuels compared to coal are attributed to high alkaline content in biomass ash. Llorente [14] notes that the increased alkaline content is particularly noticeable in herbaceous biomass, which is also characterized by higher ash content than woody biomass. Another key difference between biomass and coal composition is the increased presence of Ca, Cl, K, Mg, Mn, Na, P, phosphates, carbonates and organically bound inorganic elements in raw fuel. On the other hand, coal tends to be richer in elements such as Al, Fe, N, S, Si, Ti and certain inorganic constituents, e.g. sulfides, sulfates and silicates [18]. These organically bound ash-forming compounds as well as other included minerals or dissolved salts volatilize during combustion or are entrained in the flue gas stream and form fly ash. The inorganic composition of both raw biomass as well as the produced ash is further Fig. 1. Comparison between the American and the CEN/TS standard. J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 3 described in the works of Vassilev et al. [19,20]. 1.3. Motivation Despite the abundance and variety of plants considered for thermal utilization, only a handful of them see common use, e.g. wood, straw, hay pellets etc. Therefore, experience from industrial firing is limited to a much smaller sample than is available and the laboratory observed AFTs are still the only empirical evidence of the ash behavior of many biomass fuels. A number of regression models and other predictive criteria have been developed over the years, to help evaluate slagging and fouling tendencies based on the ash composition alone. To illustrate general trends in ash composition, analyze the correlation between the ash composition and the AFTs and to evaluate the accuracy of collected predictive criteria, a large statistical sample was needed. To that end, the Phyllis2 database for the physico-chemical composition of biomass and other feedstock, peer-reviewed data from other studies as well as our own data were collected for a total amount of 191 samples. Our data used in this study were analyzed using the European standard methods listed above, while the majority of the collected data were analyzed using the American standard method. A comparison was then made to evaluate the difference between the two standards. Our data comprise both common fuels, used for the standard comparison (wood chips, straw and hay pellets) as well as several less common plants considered for co-firing (sunflower pellets, quinoa waste, flax, etc.). 2. Experimental setup The samples are divided into two main categories: biofuels and alternative fuels, with biofuels being further divided into herbaceous and woody biomass. The herbaceous biomass samples comprise both pelletized samples (straw, hay, and sunflower) as well as untreated samples, such as rye grains, camelina and mustard seeds etc. The sewage sludge was acquired from a municipal sewage treatment plant and the digestate from a biogas plant that processes waste biomass through fermentation. 2.1. Experimental methods and results The ash content was analyzed from two samples per fuel and in accordance with ˇ CSN EN ISO 18122 [21] for all the biofuels and ˇ CSN EN 15403 [22] for the two alternative fuels. The mean values for each plant are listed in Table 1. The ash fusion temperatures were analyzed in accordance with ˇ CSN P CEN/TS 15370–1 [23] and ˇ CSN P CEN/TS 15404 [24] for biofuels and the alternative fuels respectively. The ash samples were prepared in a muffle furnace. They were then pressed into cylinders and exposed to uniform heating rate of 10 ◦C/min in oxidizing conditions. A high definition camera was used to monitor the changes in the samples’ shape and size. The photos were then evaluated using a specialized MATLAB script to investigate changes in the silhouette shape and size. The observed AFTs shown in Table 1 are the mean values of two samples that were analyzed for each fuel. The HT and FT of certain samples could not be evaluated due to atypical or instantaneous deformation of the specimen. The woody biomass samples behave as expected with AFTs falling in the usual range as demonstrated below in Table 3. However, the A1 spruce wood pellet shows uncharacteristically low fusion temperatures. This behavior was observed in both tests and is in sharp contrast with other woody samples. The cause of this behavior was attributed to the poor mechanical resilience of the pellets. For this reason, the A1 spruce pellet was excluded from the subsequent statistical analysis. It serves as an example of how brittle or less resilient samples can result in misleading AFTs. The extremely low AFTs of the herbaceous samples make them unsuitable for proper utilization. Namely quinoa waste, camelina seed and rye grains show fluid temperatures well below the softening temperatures of any wood fuel. The two alternative fuels appear to be more suitable for combustion than all the non-woody fuels. The temperature range between the initial and the final alteration (SST and FT respectively) of the system varies heavily across the samples. Vassilev et al. [9] reports the following temperature ranges: 139–171 ◦C for coal ash and 22–293 ◦C for biomass ash. The samples with the highest AFTs show the lowest temperature ranges. All the nonwoody samples with low AFTs on the other hand, show wide temperature ranges (as much as 500 ◦C and 720 ◦C in the case of flax and mustard seed respectively). The ash fusibility index was proposed in [25] to take the temperature range into account. It is defined as: AFI =4∙IDT +HT 5(1) The index, however, has seen little use so far. Furthermore, the HT of certain samples was not observed due to the reasons mentioned above. The index was therefore omitted in the analysis presented here. The use of additives, such as kaolinite, mullite, bentonite, dolomite, quartz, etc., to increase the AFTs and lower the slagging and fouling tendency has been suggested by some authors [18,26]. According to [27], the addition of marble sludge or its mixture with calcium lignosulfonate eliminated slag formation during barley and husk pellets combustion. This was due to formation of high temperature melting calcium silicates, phosphates and oxides. Similarly, the addition of phosphorous additives produced high melting potassium‑calcium phosphates and inhibited slagging and melting problems [28]. Kaolin and soil show a similar trend in increasing the AFTs, as observed by [29]. The effects of a SiO 2 -based additive, however, have been negative. A more comprehensive summary is available in [30]. Vassilev [18] notes that many solid fuels are rich in refractory minerals while low on fluxing mineral content and could therefore be mixed with the low AFT fuels instead of using additives. Biomass co-firing with coal is a common practice and has been a subject of numerous works (e.g. [31–33] or [34]). However, the knowledge of co-firing different biomass mixtures is, at present, limited and further investigation on co-firing different biomass and waste biomass blends is warranted. It is obvious that the effects of any additive will largely depend on the ash composition and the mineral phases presented in the fuel. The ash composition of our samples analyzed in this work is shown in Table 2 below. Inductively coupled plasma atomic emission spectroscopy (ICP-AES) was used for the ash composition analysis. The four most abundant oxides found in the ash samples are SiO 2 , CaO, K 2 O and P 2 O 5 . The three wood samples with the highest SSTs are Table 1 Ash fusion temperatures (◦C) and ash content obtained from the experiment. Sample SST DT HT FT Ash content A d [%] Woody biomass Wood pellet A1 (spruce) 670 930 – – 0.329 Softwood chips 1310 1320 1370 1390 3.010 Hardwood chips 1480 1500 – – 2.441 Amaranth wood 1380 1450 – – 7.704 Non-woody biomass Hay pellets 870 1080 1160 1230 17.729 Straw pellets 800 840 1000 1040 6.639 Sunflower pellets 890 1140 1210 1270 34.623 Quinoa waste 610 680 750 910 15.161 Flax 970 1170 1350 1470 17.900 Rye grains 620 750 900 970 13.675 Camelina seed 680 950 – – 13.360 White mustard seed 730 970 1170 1450 17.088 Safflower grains 740 840 1010 1060 6.085 Alternative fuels Sewage sludge 980 1160 1260 1270 49.103 Digestate 1180 1240 1260 1270 15.947 J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 4 all rich in CaO (>39%) with the other oxides being below 10%, only amaranth being slightly enriched in K 2 O (14,4%). SiO 2 was found to be the dominant oxide in the medium temperature samples (SST 800–990 ◦C). These include sewage sludge, flax and the three pelletized fuels. The lowest SST ash samples (<800 ◦C) include safflower, mustard seed, camelina seed and rye grains. P 2 O 5 was the most dominant oxide in these samples (>30%), followed by K 2 O (14–28%). Only quinoa waste was low on P 2 O 5 , similar to the high temperature melting samples. Its K 2 O content was, however, the highest across all samples (36%). 2.2. Expanded dataset The Phyllis2 database for the physico-chemical composition of biomass and other feedstock [35] was used to collect more data, so that general trends and correlation could be established with a sizeable statistical sample. Peer-reviewed data from other studies [10,36] have been used to further expand the sample. The 86 samples available in [19,37] have, however, not been used, because the ash composition in those works was already given in rel% and the trends illustrated below are mostly based on wt%. Furthermore, the inclusion or exclusion of certain oxides does not affect the wt% of the other oxides, unlike with rel %, which makes wt% more convenient to use. A total of 177 samples have ultimately been collected. These were further divided into groups based on their origin: - Woody a woody biomass: pine, poplar, spruce, willow, amaranth, wood chips, tree pruning, bark, birch, trunk, forest residue etc. - Grasses and flowers: alfalfa, lucerne, lupin, kenaf, miscanthus, knot grass, reed canary grass, verge grass, flax - Straw: wheat, rice, oat, rye, barley, maize, sweet sorghum, sunflower, corn Table 2 The chemical composition of ash from selected samples given in wt%. Sample K 2 O Na 2 O CaO MgO Fe 2 O 3 Al 2 O 3 SiO 2 P 2 O 5 MnO SO 3 TiO 2 S Cl Woody biomass Wood pellet A1 (spruce) 12.5 0.25 32.1 10.9 0.71 0.51 2.72 1.99 6.43 2.47 0.036 0.99 0.1 Softwood chips 8.71 0.4 39.3 3.03 0.94 1.56 8.62 5.55 0.18 1.86 0.096 0.74 0.1 Hardwood chips 9.81 0.17 42.2 4.34 0.49 0.55 2.8 5.47 0.18 2.17 0.041 0.87 0.1 Amaranth wood 14.4 0.29 39.2 6.74 0.48 0.22 2.38 3.7 0.027 1.87 0.27 0.75 0.17 Non-woody biomass Hay pellets 20.9 0.65 11.6 4.28 1.58 3.91 36.3 5.72 0.26 5.05 0.27 2.02 0.86 Straw pellets 18.8 0.58 8.1 3.2 1.94 4.97 45 5.07 0.19 3.22 0.31 1.29 1.51 Sunflower pellets 14 2.62 9.92 3.51 3.73 5.94 46.8 6.14 0.28 2.61 0.4 1.05 1.16 Quinoa waste 36 0.33 11.4 5.75 0.27 0.11 1.18 4.79 0.067 3.43 0.017 1.37 1.61 Flax 5.48 0.7 2.35 1.45 0.85 3.88 76.9 3.98 0.084 0.86 0.35 0.34 0.17 Rye grains 28.1 0.19 3.13 11.3 0.56 0.45 2.65 45.4 0.18 1.48 0.031 0.59 0 Camelina seed 27.4 0.089 8.12 12.6 0.6 0.12 1.04 34.8 0.089 8.97 0.015 3.59 0.1 White mustard seed 14.6 0.34 12.4 7.72 0.52 2.23 19.6 31.1 0.057 5.93 0.032 2.38 0.1 Safflower grains 23.9 0.28 11.7 14.3 0.56 0.01 0.54 39.3 0.09 2.55 0.009 1.02 0.1 Alternative fuels Sludge 7.33 1.22 11.6 3.07 11.5 7.52 37.3 9.5 0.15 3.69 0.58 1.48 0.43 Digestate 10.6 2.38 22.7 11.8 0.87 0.33 14 17.2 0.34 4.58 0.031 1.83 1.42 Table 3 An overview of the AFTs and chemical composition of the collected samples. Group IDT ST HT FT SiO 2 CaO K 2 O P 2 O 5 Fe 2 O 3 Al 2 O 3 MgO Na 2 O SO 3 Samples ◦C ◦C ◦C ◦C wt% wt% wt% wt% wt% wt% wt% wt% wt% – Wood and woody biomass Mean 1179 1304 1394 1419 19,15 29,68 12,09 5,56 2,49 3,81 5,95 1,40 4,24 44 Minimum 700 730 1140 1160 0,43 9,20 1,55 0,00 0,14 0,25 1,82 0,00 0,85 Maximum 1490 1680 1750 1800 50,20 59,00 35,30 22,80 13,70 14,00 45,88 10,80 19,72 Grasses and flowers Mean 916 1006 1206 1293 43,66 11,10 20,74 4,21 0,45 0,65 3,15 0,49 3,44 53 Minimum 650 680 710 710 2,10 3,00 2,40 1,48 0,08 0,07 0,88 0,07 1,12 Maximum 1585 1750 1800 1800 89,85 37,00 44,20 15,35 2,28 5,80 13,00 3,10 14,60 Straw Mean 913 1040 1191 1254 43,15 8,36 20,55 4,14 0,55 0,87 2,40 0,70 3,29 50 Minimum 600 620 650 680 3,10 1,74 2,10 0,64 0,05 0,09 0,59 0,03 0,55 Maximum 1260 1378 1750 1800 82,13 26,50 43,70 18,50 2,03 3,51 5,25 7,52 10,30 Husk Mean 1131 1222 1174 1206 26,44 12,06 22,49 8,99 1,52 2,71 5,95 4,19 2,65 9 Minimum 740 760 860 1020 0,40 0,20 0,30 0,87 0,10 0,10 0,10 0,00 0,77 Maximum 1400 1450 1400 1400 95,60 26,61 42,00 22,10 4,25 10,42 12,40 33,10 4,24 Organic residue Mean 1105 1173 1219 1253 55,32 6,34 10,76 2,35 5,37 7,84 3,46 0,89 3,12 6 Minimum 870 1000 1065 1070 45,88 2,50 1,67 0,89 0,09 0,16 0,60 0,14 0,40 Maximum 1260 1380 1430 1480 84,20 15,60 24,10 5,96 15,45 20,55 5,50 1,77 8,24 Other Mean 1210 1281 1352 1344 30,57 24,25 6,61 8,23 4,31 11,82 4,22 0,92 3,70 15 Minimum 815 1000 1195 1150 8,10 3,00 0,28 0,19 0,30 0,20 0,87 0,18 0,68 Maximum 1432 1540 1700 1700 59,50 54,50 33,60 30,40 18,00 32,28 9,11 2,00 8,95 J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 5 - Husk: almond shells, cacao, hazelnut shell, olive flesh, cotton and seed husks, rice hulls and husks, sunflower hull and husks - Organic residue: bagasse, sugar cane and sugarcane bagasse, sugarcane fiber, wheat grain, compost - Other: cattle and chicken manure, domestic organic waste, waste from brewing and malt industries, paper, sludge, sewage sludge. The mean values of the AFTs and ash composition are shown in Table 3 along with the minimal and maximal values to better illustrate the range of the values. The majority of the samples have been analyzed using the American standard ash fusion test. All the samples have been analyzed in oxidizing conditions. The woody samples from our experiment follow the general trend in AFTs, being close to the mean values observed across the other 44 samples collected for reference. Their SiO 2 content is, however, well below the average. Higher SiO 2 content is most likely caused by soil inclusions, as elements such as Si, Fe and Al often represent inorganic matter of soil origin [52]. The content of all the other oxides falls close to the mean values. The SST of the woody samples does not deviate noticeably from the IDT of the corresponding reference samples, showing a negligible difference between the temperatures obtained with the two standards. It is important to note that the comparison between the American and the European standard is not based on parallel ash samples and therefore some deviations might occur. Hor´ ak et al. [38] compared the two standards on four different biomass samples in both oxidizing and reducing conditions. The differences between the two methods ranged from negligible to around 100 ◦C maximum. The differences between the results obtained in oxidizing and reducing conditions were mostly negligible. While [38] recommends the use of SST instead of IDT, the consensus on the matter is presently not uniform. The medium temperature fuels found to be rich in SiO 2 follow the general trends of straws fairly accurately, with most values again falling near the mean values in Table 3. The less commonly utilized plants rich in P 2 O 5 and with the lowest AFTs (safflower, mustard seed, camelina seed and rye grains) do not fit into any of the groups. The lowest temperatures found in the 177 samples appear in straws; however, the maximal value of P 2 O 5 is only 18.5%. In fact, none of the samples collected for the expanded dataset show such high amounts of P 2 O 5 content. The inclusion of own data in the expanded dataset is therefore crucial to cover as wide ash composition variety as possible. The general trends in composition illustrated in Fig. 2 are in accordance with the trends observed by Vassilev et al. in [9]. The SiO 2 and K 2 O content decreases with increasing IDT while the CaO, MgO and Al 2 O 3 content increases. The analyzed samples behave similarly to these trends, the only major difference again being the high P 2 O 5 content in the low AFT samples and its sharp decline with increasing AFTs. The correlation between the AFTs and the ash composition (each oxide content) across the combined dataset of 191 samples was then analyzed. The aim was to investigate which oxides have a positive and which have a negative effect on the AFTs. As no significant outliers were found and the variables were shown to have linear relationships, Pearson’s correlation coefficient (defined below) was used. Rxy =∑ n i=1 (xi−x)(yi−y)  ∑ n i=1 (xi−x)2 √  ∑ n i=1 (yi−y)2 √(2) where: n: sample size x i , y i : individual sample points indexed with i x=1 n∑n i=1xi: the sample mean (analogously for y). The results are presented in Table 4. Correlation coefficient values above ±0.35 were highlighted as they were deemed substantial. Negative values indicate a negative linear correlation, i.e. a decreasing trend in the temperatures with increasing content of the particular oxide. When examining the combined dataset, it was also found that the samples with large temperature spans (difference between IDT and FT) show similar behavior. The large gap in temperatures often occurs between ST and HT, occasionally between IDT and ST. The gap between FT and HT is generally below 100 ◦C as the final two stages often occur at similar temperatures. This is also evident when examining the correlation coefficients between the AFTs, as IDT shows strong correlation with ST (R =0,86), but significantly lower correlation with HT (R =0,46) and FT (R =0,36), while the correlation between HT and FT is noticeably stronger (R =0,93). 3. Slagging and fouling prediction The dataset collected by Vassilev in [9] was further examined by Garcia-Maraver et al. in [39] and the correlation between the AFTs and different predictive indices was tested. However, as mentioned above, the ash composition found in those works is given in rel.% and therefore unfit for use in the common slagging and fouling predictors as well as several AFT predictive models. Further investigation on the correlation was therefore warranted and is presented below. 3.1. AFT prediction The routine AFT tests are fairly time consuming and multiple researchers have therefore attempted to find a method for AFT prediction based on the ash composition. Different approaches have been tried to that end. Thermodynamic modelling [40,41] through the minimization of the Gibbs energy of a system is a powerful tool for predicting the chemical behavior of complex systems. However, a more convenient method is often desired for practical in the field purposes. While formulas derived through regression analysis are purely statistics-based 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% IDT < 900 900 – 1100 1100 – 1300 IDT > 1300 SiO₂ CaO K₂O P₂O₅ MgO Al₂O₃ Fe₂O₃ Na₂O SO₃ Fig. 2. The mean ash composition of the reference samples sorted by their IDT. J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 6 and do not fully reflect the physico-chemical and thermodynamic processes of the system, their application is more convenient. Linear regression has seen a frequent use in coal AFTs prediction. The results found in [42,43] or [44] are satisfactory. When applied to biomass fuels, their predictive powers drop significantly. This is likely due to the vastly different composition of biomass ashes as well as the different minerals and phases found in them. Only a handful of studies, such as [10,45] or [46], have derived empirical formulas for AFT prediction from biomass datasets. To evaluate their accuracy, the ash composition from our experiment were used in these regression models and the predictions were then compared to the observed AFTs. While the models appear to be accurate when applied to their original datasets, their predictive powers drop drastically when applied to some of the samples presented in this work. The predicted values as well as the relative errors of the models are shown in Table 5 below. It is apparent that the accuracy of the models varies across the presented samples. The model for woody biomass available in [10] shows satisfying results with all samples with STs above 1000 ◦C. However, the predicted values of the low ST samples are significantly higher than the observed values. This is true for all four selected models and could have severe consequences if used in place of the routine AFT test. The poor predictive power for the low ST samples can be attributed to several factors: The dataset used to derive the models is comparatively small to the sizeable coal datasets. This gives the models limited, narrow range. Moreover certain abundant oxides were also omitted, such as P 2 O 5 in both [10,46] or SiO 2 in [45]. P 2 O 5 was found to be the dominant oxide in the samples, where the accuracy of the models was the lowest. On the other hand, overfitting the regression model with too many regressors compared to the amount of observations, such as in [46], often leads to high correlation with the original dataset, but limited predictive power. Another common issue is the pursuit of high R 2 values through negligent elimination of outliers, curve fitting using polynomial terms, etc. This often leads to situations where the regression coefficients start to represent the noise rather than the actual relationship and the applicability of the model outside the original dataset deteriorates. Linear multivariable regression analyses were carried out on the acquired dataset in order to evaluate the applicability of the statistical approach to predict IDT. Only the four major oxides (SiO 2 , P 2 O 5 , CaO and K 2 O) were used as regressors in the first analysis. The second analysis took into account the content of all the oxides. The results are summarized in Table 6 below. The table shows both the R 2 value as well as the adjusted R 2 value to evaluate the importance of the individual regressors. The standard error of the estimate shows that using multivariable regression to predict the IDT could lead to inaccurate results. It was found that the noticeable outliers with predicted values greater than ±2 σ were often the samples with large temperature spans. These samples hinder the applicability of the linear regression analysis; however, their exclusion would not be justified. Because the statistical approach does not take into account the complex chemical reactions and because the accuracy of the predictions was shown to be insufficient, its use is currently not recommended. 3.2. Predictive indices Apart from the AFT test, a number of indices have been developed to evaluate the slagging and fouling potential of coals. These include: Basic to acidic compounds ratio (B/A) [48]: B A=Fe2O3+CaO +MgO +Na2O+K2O SiO2+Al2O3+TiO2 (3) where each oxide is represented by the mass fraction in the ash (wt%). As mentioned above, the acidic compounds have been shown to increase the overall melting temperature while the basic compounds have the opposite effect. The inclusion of P 2 O 5 is somewhat inconsistent in Table 4 The correlation coefficient values between the AFTs and the ash composition. P 2 O 5 SiO 2 Fe 2 O 3 Al 2 O 3 CaO MgO Na 2 O K 2 O SO 3 Correlation with IDT −0,05 −0,24 0,21 0,41 0,58 0,21 0,10 ¡0,43 −0,02 Correlation with ST −0,06 −0,11 0,16 0,31 0,54 0,17 0,10 ¡0,47 −0,01 Correlation with HT 0,12 ¡0,37 −0,01 0,10 0,47 0,18 −0,03 −0,16 0,09 Correlation with FT 0,11 ¡0,39 −0,07 0,03 0,39 0,14 −0,06 −0,08 0,07 Table 5 Comparison between the observed SSTs and the values predicted by selected models (◦C) with relative errors in %. Regression model: [10] Woody [10] Herbaceous [45] [46] Sample observed predicted rel. Error predicted rel. Error predicted rel. Error predicted rel. Error Hardwood chips 1500 1494 0,4 1354 9,7 1162 22,5 1609 7,3 Amaranth wood 1450 1443 0,5 1342 7,4 1116 23,1 1656 14,2 Softwood chips 1320 1398 5,9 1343 1,7 1173 11,1 1448 9,7 Digestate 1240 1306 5,4 1283 3,4 1230 0,8 1384 11,6 Sludge 1160 1139 1,8 1253 8,0 1254 8,1 1221 5,3 Flax 1170 1106 5,5 964 17,6 1198 2,4 1150 1,7 Sunflower pellets 1140 1132 0,7 1135 0,4 1133 0,6 1199 5,2 Hay pellets 1080 1143 5,8 1123 4,0 1058 2,0 1216 12,6 Straw pellets 840 1127 34,1 1073 27,7 1063 26,6 1189 41,5 Safflower grains 840 1266 50,7 1214 44,5 1247 48,5 2308 174,8 White mustard seed 970 1191 22,8 1195 23,2 1285 32,5 1280 32,0 Camelina seed 950 1217 28,1 1132 19,2 1207 27,1 1736 82,7 Rye grains 750 1177 56,9 960 28,0 1252 66,9 1394 85,9 Quinoa waste 680 1177 73,1 1135 66,9 940 38,3 1691 148,7 Table 6 Linear multivariable regression analysis results with IDT as the dependent variable. Sample size n R 2 Adjusted R 2 Standard error of the estimate σ Number of outliers > ±2 σ Only major oxides 191 0.453 0.441 166.7 13 All oxides 191 0.555 0.530 156.5 9 J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 7 literature: grouped with basic compounds in [39], acidic in [44] or omitted completely in other works [47–49]. SO 3 is usually omitted as well, only grouped with acidic compounds in [44]. The slagging tendency of the ash increases with increasing B/A ratio. The ash is said to have low slagging inclination when B/A <0.5, medium for 0.5 <B/A < 1, high for 1 <B/A <1.75 and severe for B/A above 1.75. Simplified B/A ratio [39]: B As =Fe2O3+CaO +MgO SiO2+Al2O3 (4) The simplified B/A ratio is sometimes used instead to reduce the analysis time needed to identify the different oxides. However, the omission of certain crucial compounds, especially K 2 O, makes it inappropriate for biomass fuels. Low slagging is to be expected when B/As is below 0.75. Bed agglomeration index [50]: BAI =Fe2O3 Na2O+K2O(5) The bed agglomeration index has been developed to evaluate operational problems during fluidized bed combustion. Bed agglomeration is often caused by low AFT ash, therefore higher coefficient values should indicate higher AFTs. According to [50], bed agglomeration occurs when BAI is below 0.15. Fouling index [39]: Fu=(B A)∙(Na2O+K2O)(6) The fouling index is based on the B/A ratio in eq. 3. While the sodium content in biomass ash is generally low, potassium has been shown to be one of its major constituents. When potassium condenses on flyash particles, it increases the surface stickiness and is therefore a major cause of fouling. This was observed by Miles et al. [16], who found K 2 O to be one of the major constituents (along with SO 3 and CaO) of the ash deposits on superheater tubes. Low fouling inclinations are to be expected for Fu <0.6, high for Fu values up to 40 and extremely high for values above 40. Slag viscosity index (silica ratio) [48]: Sr =(SiO2 SiO2+CaO +MgO +Fe2O3)∙100 (7) The grate and furnace wall deposits are almost identical to the original ash composition and retain most silica present in the ash. The slag viscosity index is hence used to evaluate the slagging tendency inside the furnace. High values correspond to high viscosity and therefore low slagging. Low slagging occurs when Sr is above 72, medium for 65 <Sr <72 and high for values below 65. The alkali metals form lowmelting eutectics with silica and their ratio to silica (AtS) defined as (K 2 O +Na 2 O)/SiO 2 is sometimes used to evaluate fouling as well. Najser et al. [51] reports the safe range to be between 0.35 and 0.45 for furnace temperatures up to 1100 ◦C. Alkali index [52]: Al =(Na2O+K2O)∙Ar HHV (8) The alkali index expresses the quantity of alkali oxides in the fuel per unit of fuel energy. The fuel energy is commonly expressed as the higher heating value (HHV) in GJ/kg, while A r represents the ash content in raw fuel (wt%). According to the thresholds presented in [52], fouling is probable above 0.17 kg alkali/GJ and virtually certain to occur above 0.34 kg alkali/GJ. The higher heating value of the selected samples was taken from [53], where a more thorough chemical analysis is also available. The indices were applied to our own data (Table 7) as well as the expanded dataset (Table 8) and compared to the laboratory-observed AFTs. The slagging and fouling propensity based on the AFTs is evaluated by the following thresholds: Slagging and fouling propensity SST/IDT (based on [9]) DT/ST (based on [10]) Low >1300 >1390 Medium 1100–1300 1250–1390 High <1100 <1250 Not only do the indices predict drastically different tendencies than observed through the AFTs, the predictions for each sample vary based on the index that is applied. For example, the prediction based on the B/ A ratio for softwood chips shows extremely high fouling, while the prediction for the same fuel based on the Fouling index shows low fouling, which correlates more with the observed AFT. All the criteria show high slagging and fouling tendencies for commonly used fuels, such as wood chips, while on the other hand showing low tendencies for low AFT fuels, such as straw pellets. Experience from firing these two fuels are better represented by the AFTs. It is therefore recommended to use the laboratory-obtained AFTs as a guideline over the use of slagging and fouling indices. Table 7 Comparison between the observed AFTs and the predictions made by the indices. Sample SST DT B/A B/As BAI Fu Sr Al Hardwood chips 1480 1500 extremely high high high high high medium Amaranth wood 1380 1450 extremely high high high high high high Softwood chips 1310 1320 extremely high high high low high medium Digestate 1180 1240 extremely high high high low high high Sludge 990 1160 medium low low low high high Flax 970 1170 low low high low low high Sunflower pellets 890 1140 medium low low low low high Hay pellets 870 1080 medium low high low medium high Straw pellets 800 840 medium low high low low high Safflower grains 740 840 extremely high high high high high high White mustard seed 730 970 high high high low high high Camelina seed 680 950 extremely high high high high high high Rye grains 620 750 extremely high high high high high high Quinoa waste 610 680 extremely high high high high high high Table 8 Correlation between the AFTs and selected indices across the expanded dataset (191 samples). B/A B/As BAI Fu Sr AtS Dol Mean 5.02 2.72 0.31 1.25 0.55 2.99 0.49 Minimum 0,09 0.03 0.003 0.002 0.01 0.03 0.14 Maximum 92.25 49.47 15.00 23.02 0.98 62.19 0.96 Correlation with IDT 0.06 0.22 0.15 −0.08 −0.35 −001 0.60 Correlation with ST −0.04 0.14 0.11 −0.19 −0.26 −0.11 0.60 J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 8 To further demonstrate the poor applicability of the slagging and fouling indices to biofuels, the expanded dataset was used to establish correlation between the laboratory observed AFTs and the indices. The HHV of many samples in the extended dataset was not available, therefore the Alkali index was omitted. Table 8 below shows the mean, minimum and maximum values of the indices across the expanded dataset. Pearson’s correlation coefficients between the two AFTs and the predictive indices were established to better illustrate the weak correlation. The correlation coefficients indicate insignificant relationships, only silica ratio (Sr) shows weak correlation with the AFTs. Adjusting the thresholds of the indices to better fit biomass fuels would therefore be ineffective, due to the weak correlations. The dolomite ratio (listed as Dol in Table 8 and defined below) is normally used to describe the relative % of dolomitic compounds to the rest of the inorganic matter [52]: Dol =CaO +MgO CaO +MgO +SiO2+K2O+…(9) Interestingly, it shows a relatively strong correlation with the AFTs. This is due to multicollinearity induced by using CaO both in the numerator and the denominator. It was shown earlier in Table 4that CaO has a strong correlation with IDT and using the same oxide in the index twice leads to artificial increase in the correlation value. 3.3. Ternary plot Ternary diagrams can be used as a relatively reliable alternative to the predictive criteria. The data collected in this work were plotted into a ternary diagram (Figure 3). The samples were divided into three groups based on the temperature ranges shown in the diagram. The diagram can be used for preliminary IDT or fouling estimation based on the ash composition alone. The oxides were split into three groups based on the works of Vassilev [9,15] and are presented in rel. %. Chlorine and certain oxides, such as SO 3 and TiO 2 , were omitted and only the most abundant oxides were chosen to reduce the necessary analysis time. The two dolomitic compounds CaO and MgO were grouped together as they are often found in high melting temperature samples and show positive correlation with the AFTs. SiO 2 was grouped with the other two metallic compounds, Al 2 O 3 and Fe 2 O 3 , which are often found in low quantities. Lastly, the two alkali compounds K 2 O and Na 2 O were grouped with P 2 O 5 as they were found to be the most abundant oxides in the low melting temperature samples. The diagram was further divided into three distinct regions based on the prevalent IDT group found within them: Region (1): samples found within this region show high IDTs and should therefore cause little to no fouling problems when burnt at common flame temperatures. The majority of the samples in this region consist of raw woody biomass (spruce, pine, poplar, willow) or in the form of chips, pellets and bark. Region (2): is the transitionary region between the high IDT and low IDT regions. It consists mostly of certain low IDT wood, such as amaranth wood, contaminated wood or waste wood. Other samples found within the region are paper waste, sludge and sewage sludge, rice husks and shells, hemp and kenaf. Region (3): is mostly made out of straw (wheat, rye, barley, oat, rice, sunflower, corn, maize and sweet sorghum), grasses and flowers (miscanthus, flax, alfalfa, reed canary grass), sugarcane bagasse and fiber, sunflower husks and seed and chicken and cattle manure. Plants found within this region show high slagging and fouling propensity and should therefore not be used without additives or co-firing with a less problematic fuel. The definition of the boundaries within these three regions can further be improved as more samples are analyzed and added to the diagram. The low IDT outliers found within region (2) were mostly woody samples that showed large temperature spans between IDT and FT. As mentioned before, this behavior is problematic and difficult to predict through mere statistical analysis. The high IDT samples found within region (2) and (3) were most likely contaminated with soil, which is demonstrated by their large SiO 2 content. Another possible cause is the fact that silica can form both high temperature melting calcium silicates, phosphates and oxides as well as low temperature melting eutectics with sodium and potassium. Therefore, both high and low IDT samples can be enriched in SiO 2 content. 4. Conclusions The investigation has led to several conclusions which are summarized here: 1) The silica and potassium content tends to decrease with increasing AFTs. The calcium content on the other hand increases. CaO was found to be the dominant oxide in most high AFT samples. 2) Certain low IDT samples were found to show large temperature spans between the initial and the final fusion temperatures. The largest gaps were found to occur between ST and HT, demonstrating that partial melting and possible slagging problems can occur at significantly lower temperatures than complete melting of the specimen. The behavior of these samples is difficult to predict through statistical means. Further investigation into this behavior is therefore warranted. 3) The accuracy of the available multivariable regression models for AFT prediction was found to be insufficient when applied to the experimental data and their use is therefore not recommended. 4) The applicability of slagging and fouling predictive indices was shown to be poor when applied to biomass-based fuels. Most of the tested samples, including wood, were attributed high or extremely high fouling tendencies, even though the AFTs and experience from firing wood shows low to moderate fouling instead. Certain medium AFT and low AFT samples on the other hand, were attributed low fouling tendencies by some indices. Moreover, in some cases the indices would contradict each other. The application of these indices to biomass-based fuels is therefore not recommended. 5) The proposed graphical approach to preliminary AFT evaluation could be convenient for in the field purposes if proven to be sufficiently accurate. Fig. 3. Ternary diagram with all the 191 samples plotted and grouped based on their IDT. J. Lachman et al. Fuel Processing Technology 217 (2021) 106804 9 Author statement Jakub Lachman - writing of the original draft and the revised manuscript, formal analysis and statistical investigation. Marek Bal´ aˇ s – supervision and conceptualization. Martin Lisý – funding acquisition and project administration. Hana Lis´ a – laboratory investigation. Pavel Milˇ c´ ak – resources. Patrik Elbl – visualization and editing. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements This work was supported by the Ministry of Education, Youth and Sports of the Czech Republic under OP RDE grant number CZ.02.1.01/ 0.0/0.0/16_019/0000753 “Research centre for low-carbon energy technologies” and the Brno University of Technology, Faculty of Mechanical Engineering in frame of the Specific Research Fund, grant no.: FSI-S-20-6280. References [1] Wenchao Ma, Terrence Wenga, Flemming J. Frandsen, Beibei Yan, Guanyi Chen, The fate of chlorine during Msw incineration: vaporization, transformation, deposition, corrosion and remedies, Online, Prog. Energy Combust. Sci. 76 (2020), https://doi.org/10.1016/j.pecs.2019.100789. [2] P. Teixeira, H. Lopes, I. Gulyurtlu, N. Lapa, P. Abelha, Slagging and fouling during coal and biomass cofiring: chemical equilibrium model applied to Fbc, Online, Energy Fuel 28 (1) (2014) 697–713, https://doi.org/10.1021/ef4018114. [3] Xiwen Yao, Kaili Xu, Fang Yan, Yu Liang, The influence of ashing temperature on ash fouling and slagging characteristics during combustion of biomass fuels, Online, Bioresources 12 (1) (2017) 1593–1610, https://doi.org/10.15376/ biores.12.1.1593-1610. [4] Abel Tortosa-masi´ a, Falk Ahnert, Harmut Spliethoff, Claude Loux, Klaus Hein, Slagging and fouling in biomass co-combustion, Online, Therm. Sci. 9 (3) (2005) 85–98, https://doi.org/10.2298/TSCI0503085T. [5] Ulrich Kleinhans, Christoph Wieland, Flemming J. Frandsen, Hartmut Spliethoff, Ash formation and deposition in coal and biomass fired combustion systems: progress and challenges in the field of ash particle sticking and rebound behavior, Online, Prog. Energy Combust. Sci. 68 (C) (2018) 65–168, https://doi.org/ 10.1016/j.pecs.2018.02.001. [6] Jan Hrdliˇ cka, Pavel Skopec, Tom´ aˇ s Dlouhý, Frantiˇ sek Hrdliˇ cka, Emission factors of gaseous pollutants from small scale combustion of biofuels, Online, Fuel 165 (2016) 68–74, https://doi.org/10.1016/j.fuel.2015.09.087. [7] ASTM D1857 / D1857M-18, Standard Test Method for Fusibility of Coal and Coke Ash, ASTM International, West Conshohocken, PA, 2018. www.astm.org. [8] A. Garcia-Maraver, J.A. P´ erez-Jim´ enez, Biomass pelletization: standards and production, Online, Chem. Eng. Prog. 111 (6) (2015) 50, http://search.proquest.co m/docview/1705958793/. [9] Stanislav V. Vassilev, David Baxter, Christina G. Vassileva, An overview of the behaviour of biomass during combustion: Part II. ash fusion and ash formation mechanisms of biomass types, Online, Fuel 117 (PA) (2014) 152–183, https://doi. org/10.1016/j.fuel.2013.09.024. [10] L.Y. Yu, L.W. Wang, P.S. Li, Study on prediction models of biomass ash softening temperature based on ash composition, Online, J. Energy Inst. 87 (3) (2014) 215–219, https://doi.org/10.1016/j.joei.2014.03.011. [11] V. Ganapathy, Industrial Boilers and Heat Recovery Steam Generators: Design, Applications and Calculations, CRC Press, New York, 2003. [12] Kumar Rayaprolu, Boilers: A Practical Reference, CRC Press, Boca Raton, 2013 (ISBN 978-1-4665-0053-2). [13] A. Williams, 95/00598 understanding slagging and fouling during Pf combustion, Fuel Energy Abstr. 36 (1) (1995), https://doi.org/10.1016/0140-6701(95)958517, 35–35. Online. [14] M.J. Fern´ andez Llorente, J.E. Carrasco García, Comparing methods for predicting the sintering of biomass ash in combustion, Online, Fuel 84 (14) (2005) 1893–1900, https://doi.org/10.1016/j.fuel.2005.04.010. [15] Stanislav V. Vassilev, David Baxter, Christina G. Vassileva, An overview of the behaviour of biomass during combustion: part I. Phase-mineral transformations of organic and inorganic matter, Online, Fuel 112 (2013) 391–449, https://doi.org/ 10.1016/j.fuel.2013.05.043. [16] Thomas R. Miles, Larry L. Baxter, Richard W. Bryers, Bryan M. Jenkins, Laurance L. Oden, Boiler deposits from firing biomass fuels, Online, Biomass Bioenergy 10 (2) (1996) 125–138, https://doi.org/10.1016/0961-9534(95)00067-4. [17] W.J. Song, L.H. Tang, X.D. Zhu, Y.Q. Wu, Z.B. Zhu, S. Koyama, Effect of coal ash composition on ash fusion temperatures, Online, Energy Fuels 24 (1) (2010) 182–189, https://doi.org/10.1021/ef900537m. [18] Stanislav V. Vassilev, Christina G. Vassileva, Vassil S. Vassilev, Advantages and disadvantages of composition and properties of biomass in comparison with coal: an overview, Online, Fuel 158 (2015) 330–350, https://doi.org/10.1016/j. fuel.2015.05.050. [19] Stanislav V. Vassilev, David Baxter, Lars K. Andersen, Christina G. Vassileva, An overview of the chemical composition of biomass, Online, Fuel 89 (5) (2010) 913–933, https://doi.org/10.1016/j.fuel.2009.10.022. [20] Stanislav V. Vassilev, David Baxter, Lars K. Andersen, Christina G. Vassileva, Trevor J. Morgan, An overview of the organic and inorganic phase composition of biomass, Online, Fuel 94 (1) (2012) 1–33, https://doi.org/10.1016/j. fuel.2011.09.030. [21] ˇ CSN EN ISO 18122: Solid Biofuels - Determination of Ash Content, UNMZ, Praha, Czech Republic, 2016. [22] ˇ CSN EN 15403 (838307): Solid Recovered Fuels - Determination of Ash Content, UNMZ, Praha, Czech Republic, 2011. [23] ˇ CSN P CEN/TS 15370–1: Solid Biofuels - Method for the Determination of Ash Melting Behaviour - Part 1: Characteristic Temperatures Method, UNMZ, Praha, Czech Republic, 2007. [24] ˇ CSN P CEN/TS 15404: Solid Recovered Fuels - Methods for the Determination of Ash Melting Behaviour by Using Characteristic Temperatures, UNMZ, Praha, Czech Republic, 2007. [25] A.R. Mclennan, G.W. Bryant, C.W. Bailey, B.R. Stanmore, T.F. Wall, Index for ironbased slagging for pulverized coal firing in oxidizing and reducing conditions, Online, Energy Fuel 14 (2) (2000) 349–354, https://doi.org/10.1021/ef990127d. [26] Arthur F. Stam, Gerrit Brem, Fouling in coal-fired boilers: biomass co-firing, full conversion and use of additives – a thermodynamic approach, Online, Fuel 239 (2019) 1274–1283, https://doi.org/10.1016/j.fuel.2018.11.127. [27] L. Wang, G. Skjevrak, O. Skreiberg, H. Wu, H.K. Nielsen, J.E. Hustad, Investigation on ash slagging characteristics during combustion of biomass pellets and effect of additives, Online, Energy Fuels 32 (4) (2018) 4442–4452, https://doi.org/ 10.1021/acs.energyfuels.7b03173. [28] Qian Wang, Kuihua Han, Jiamin Wang, Jie Gao, Lu. Chunmei, Influence of phosphorous based additives on ash melting characteristics during combustion of biomass briquette fuel, Online, Renew. Energy 113 (2017) 428–437, https://doi. org/10.1016/j.renene.2017.06.018. [29] Yanqing Niu, Zhizhou Wang, Yiming Zhu, Xiaolu Zhang, Houzhang Tan, Shi ’en Hui, Experimental evaluation of additives and k2o–sio2–al2o3 diagrams on high-temperature silicate melt-induced slagging during biomass combustion, Online, Fuel 179 (2016) 52–59, https://doi.org/10.1016/j.fuel.2016.03.077. [30] Liang Wang, Johan E. Hustad, Øyvind Skreiberg, Geir Skjevrak, Morten Grønli, A critical review on additives to reduce ash related operation problems in biomass combustion applications, Online, Energy Procedia 20 (2012) 20–29, https://doi. org/10.1016/j.egypro.2012.03.004. [31] Mohammad S. Roni, Sudipta Chowdhury, Saleh Mamun, Mohammad Marufuzzaman, William Lein, Samuel Johnson, Biomass co-firing technology with policies, challenges, and opportunities: a global review, Online, Renew. Sust. Energ. Rev. 78 (C) (2017) 1089–1101, https://doi.org/10.1016/j. rser.2017.05.023. [32] Ezinwa Agbor, Xiaolei Zhang, Amit Kumar, A review of biomass co-firing in North America, Online, Renew. Sust. Energ. Rev. 40 (2014) 930–943, https://doi.org/ 10.1016/j.rser.2014.07.195. [33] Ming-che Hu, Chun-hung Lin, Chun-an Chou, Shao-yiu Hsu, Tzai-hung Wen, Hu. Ming-che, Analysis of biomass co-firing systems in taiwan power markets using linear complementarity models, Online, Energy Policy 39 (8) (2011), 4594–4594, http://search.proquest.com/docview/888103642/. [34] Yuanyuan Shao, Jinsheng Wang, Fernando Preto, Jesse Zhu, Xu. Chunbao, Ash deposition in biomass combustion or co-firing for power/heat generation, Online, Energies 5 (12) (2012) 5171–5189, https://doi.org/10.3390/en5125171. [35] Phyllis2: Database for the Physico-chemical Composition of (treated) Lignocellulosic Biomass, Microand Macroalgae, Various Feedstocks for Biogas Production and Biochar, Online, https://phyllis.nl/. [36] Shenglei Du, Haiping Yang, Kezhen Qian, Xianhua Wang, Hanping Chen, Fusion and transformation properties of the inorganic components in biomass ash, Online, Fuel 117 (PB) (2014) 1281–1287, https://doi.org/10.1016/j.fuel.2013.07.085. [37] Stanislav V. Vassilev, David Baxter, Lars K. Andersen, Christina G. Vassileva, An overview of the composition and application of biomass ash. Part 1. Phase–mineral and chemical composition and classification, Online, Fuel 105 (2013) 40–76, https://doi.org/10.1016/j.fuel.2012.09.041. [38] J. Hor´ ak, Z. Jankovsk´ a, M. Branc, F. Straka, P. Buryan, P. Kubesa, F. Hopan, K. Krpec, Problematika Stanovení Charakteristických Teplot Tavitelnosti Popela Biomasy, Online, Chem. List. 107 (6) (2013) 502–509. [39] Angela Garcia-maraver, Jesus Mata-sanchez, Manuel Carpio, Jose A. Perezjimenez, Critical review of predictive coefficients for biomass ash deposition tendency, Online, J. Energy Inst. 90 (2) (2017) 214–228, https://doi.org/10.1016/ j.joei.2016.02.002. [40] Evgueni Jak, Sergei Degterov, Peter C. Hayes, Arthur D. Pelton, Thermodynamic modelling of the system Al 2 O 3 SiO 2 CaO FeO Fe 2 O 3 to predict the flux requirements for coal ash slags, Online, Fuel 77 (1) (1998) 77–84, https://doi.org/10.1016/ S0016-2361(97)00137-3. [41] Daniel Lindberg, Rainer Backman, Patrice Chartrand, Mikko Hupa, Towards a comprehensive thermodynamic database for ash-forming elements in biomass and waste combustion — current situation and future developments, Online, Fuel J. Lachman et al.