*5.1. Filtration Process*

All filtration tests were carried out at filtration temperature equal to 450 ± 5 ◦C and a constant stream of sorbents. Three of the tested sorbents gave the desired filtration/regeneration patterns. Thus, the sorbents fulfilled their basic requirement of keeping the filtration process stable. Only dolomite (T4) led to a complete failure of the filter. With dolomite, the filter cake resisted pulse cleaning and, thus, a constant rise in dP was measured.

Figure 3 presented below depicts the results of filtration data for all four sorbents. For tests 1–3, the time from the start until the first pulse cleaning (formation of filter cake layer), ranged from 1 h 45 min to 2 h 40 min. These values correspond well with the data previously collected. Moreover, the time between subsequent pulses was consistent and repeatable (from 40 min to 1 h 15 min) as can be seen during the analysis of the registered pressure drop vs. time patterns. The repetitive regeneration cycles indicate a stable operation of the filter. For T1, another interesting characteristic feature can also be noted. After the fourth hour of operation, the filter cake collected adopted a self-cleaning (without pulses) characteristic. This phenomenon was noted before, when the filter cake adopted a loose structure and linear velocities of syngas were su fficiently small. Furthermore, a well-selected particle size of sorbent and ratio of sorbent to char in the filter cake positively influenced the observed self-cleaning property. This phenomenon is rarely registered for filtration of gasification gases; however, it is quite conventional for the HT filtration of flue gases.

During T4, the severity of pulse jet cleaning was increased gradually in the search for any signs of improvement. An increase in pulse times from the conventional 200 ms to 500 ms and in the pressure of the pulses from 6 bar to 8 bar did not give any signs of cake recovery. Conversely, the severity of pulses led to some mechanical degradation of the filter elements, and pit holes at the dirty side of the filters were noted during the later inspection.

**Figure 3.** Change in pressure drop on the high-temperature (HT) filter during filtration/adsorption tests. Pressure drop (dP) presented in a standardized form that takes into account the actual flow rate of syngas.

Concerning previous experience with CaO-based sorbents, dolomite should not have brought such a quick and decisive filter failure as noted for T4. At 60%, it is composed of CaO and flows well, similarly to chalk; thus, it is surprising to discover this problem. In the past, operation of the filter was tested for CaO precoating and co-filtration runs where the ratio of char to chalk was higher than 1:0.2 and stable filtration patterns were recorded. Noteworthy, the optimum point for operation of the filter with a pressure drop of 1 kPa was found to be the ratio of 1:1–1:2.

From preliminary testing, halloysite was shown to be a very promising material that enables both good gas cleaning and filter regeneration properties. For this reason, it was chosen for further experiments where limits of its filtration-enhancing properties were determined; thus, two additional tests were done (T2.1 and T2.2) where the mass stream of halloysite was kept constant and gasification conditions were varied through the use of fuel with different moisture content. The changes in reactor operating conditions induced changes in the quality of produced syngas, which subsequently led to pronounced variations observed in the filtration process itself.

Analysis of the three halloysite experiments was primarily done with attention to temperatures of syngas at the reactor exit, at the filter inlet, and inside the dirty side of the filter. The data comparing the halloysite filtration tests are presented in Figures 4 and 5.

As is often common for pilot installations, here, the operation of the reactor and the gas cleaning unit was also stabilized by trace heating. For the GazEla installation, the trace heating is mounted starting from the reactor outlet, throughout and past the HT filter. Due to the thermal mass of the filter, it has the most stable characteristic of temperature changes. The heating is primarily used for start-ups, as well as to keep the operation of the cleaning unit safe when an intermittent, unstable operation of the reactor happens. In normal operation of the pilot GazEla reactor, 550 ◦C to 650 ◦C syngas is produced at its outlet. The raw syngas temperature depends on reactor power and quality of the gasified fuel. As a standard, the filter body is kept at a minimum temperature of 430 ◦C.

**Figure 4.** Change in pressure drop on the HT filter during filtration/adsorption tests: comparison of three experiments with halloysite.

**Figure 5.** Changes in the main process parameters regarding the halloysite-assisted filtration of syngas.

For T2.1, the biomass was artificially wet to the point of 29 wt.%, while, for T2.2, the feedstock was dried to the point of 8.9 wt.%. The GazEla reactor was tested previously on biomasses of moisture content up to 40%; however, at this point, operation of the fixed bed changed dramatically, as did the tar content and characteristics of produced syngas. For the three experiments, the stream of fuel fed into the reactor (on a dry basis) was kept stable to simulate the conditions where the reactor power demand was kept constant and the quality of available fuel changed. This operational difference led to the variable stream of wastewaters condensed from syngas, as well as the stream of syngas calculated from the mass balance of the reactor. In a fixed bed, the instantaneous stream of fuel fed into the reactor cannot be measured directly with the precision known and needed for stable operation of FB gasifiers. Here, the changes in fuel stream were determined based on time-average mass balances and they were shown to be in good agreemen<sup>t</sup> with calculated mass and elemental balances.

In line with the above-mentioned findings, during T2.1, lower minimum and average temperatures of raw gas were noted at the reactor outlet. This stems from the higher demand for heat needed for evaporation of water contained in the feedstock. Thus, it may be counterintuitive that, during the same test, the temperature inside the filter body was on average 14–29 ◦C higher in comparison to the two other experiments. Such a characteristic was noted before for the GazEla reactor, where some residual oxygen was present in raw syngas. During the extended residence time in the HT filter body, and due to the high temperatures of the filters, the oxygen was consumed in exothermal reactions. Such phenomena are predominant for fixed-bed gasification of feedstocks with very high moisture content, and they are also often connected with the production of very high amounts of heavy tars. The reason for this is that the temperature of the pyrolysis zone in a fixed-bed reactor drops and shifts toward the production of higher-molecular-weight tars (lower thermal breakdown of the tars). The heavy tars are di fficult to measure analytically as they are very reactive and easily convert to polyaromatic hydrocarbons (PAH). The PAHs in tar can then undergo spontaneous polymerization, as a result of the high availability of C–O–H bonds, e.g. in phenols, or they can polymerize/carbonize and form solid-state char. These changes in the chemistry of tars are also one of the main reasons why HT filters often lose their porosity and eventually increase their baseline dP. To support this notion, the authors hold experience gathered throughout the past five years of operation of the filter, whereby a situation never occurred where the filtration of post-combustion gases (flue gases) led to a complete failure of the filter elements (pressure drop exceeding 6kPa), despite the grea<sup>t</sup> range of tested operational parameters. In flue gases, the stable filtration behavior is adopted much more quickly, mostly during the first 1–2 h, and it is possible to cycle between operational baseline set points without problems. The filter also operates on flue gases at much higher temperatures without instability. Supporting this thesis seems also to be the experiences from entrained flow, membrane reactors, where completely tar-free syngas is produced, e.g., utilized in integrated gasification combined cycle power plants. There also, the syngas ceramic filters are commonly operated close to their upper temperature limits of 900–1000 ◦C with acceptably long maintenance cycles (>5000 h).

In Figure 5, it can be seen that, in T2.1, the filtration behavior adopted two separate patterns. In the first part of the tests, the pressure drop increased steadily, up to the set point of baseline dP (<2.5 h). At this time, regeneration pulses occurred every ca. 4–5 min, which was never observed for continuous stable runs, indicating that the filter cake kept sticking to the surface of the filters. To keep the unit in operation, the pressure drop set point was not changed until ca. the fourth hour of the test. From previous experience, a stable filtration pattern can be recovered for some cases where the thickness of the filter cake layer is deliberately, substantially increased. A thicker layer of filter cake tends to drop off the elements in larger flakes; however, this can also lead to patchy cleaning and irreversibly higher operational baseline of the filter. In this case, the method was successful, and, after the fifth hour, the time intervals between pulses started to increase. This can be observed by the dropping slopes of the dP curve between subsequent regeneration cycles. In stable filtration runs noted for both T2 and T2.2, the duration between regeneration pulses reached the conventional 1 h to 1 h 15 min.

Neither the temperature at the filter inlet nor the temperature in the filter body seemed to have any correlation with the moment when the filtration started to run stably. The moment when the filtration stabilized for T2.1 was in good relation to the time when the temperature profile of the pyrolysis zone of the reactor reached 500 ◦C. No noticeable change in the composition of syngas with respect to permanent gases (O2, CO2, CO, CH4, H2), the stream of fed fuel, or stream of recovered wastewater was registered. Thus, the three performed tests indicate the predominant role of tars in keeping the syngas filtration stable. Noteworthy for T2.1 was also the much higher actual stream of syngas filtered. The higher amount of water in raw syngas led to an increase in the filtration velocity, Uf (linear velocity of gas passing through a filter element, m/s). For HT applications, a commonly accepted value of maximum allowable Uf is 3 m/s. However, from experience, the filtration of syngases from fixed-bed reactors at Uf > 1.5 m/s cannot be kept with a pressure drop lower than 2kPa. Hence, to keep the dP low, the operational optimum was found for this unit to keep the Uf in the range from 1 m/s to 2 m/s. During T2.1, the Uf reached the maximum of 1.9 m/s.

### *5.2. Water, Tar, and Solid Particle Content*

The water content of syngas at the outlet from the gasifier is directly connected to fuel composition, its moisture content, and the amount of steam used as a gasifying agent. Secondary reactions connected to hydrogen present in fuel further lead to the formation of steam, which can take part in tertiary reforming reactions. For the GazEla reactor, a fuel of moisture content 25 wt.%. was found to be optimal. The moisture content of the fuel and the amount of steam added for temperature control of the reactor's bed together have a high influence on the amount of tar produced. For the pilot-scale GazEla reactor, the content of organics in raw syngas can reach up to 50 g/Nm<sup>3</sup> for waste fuels such as sewage sludge. For the majority of conventional feedstocks, the value does not exceed 25 g/Nm3. Table 4 presents concentrations of the basic contaminants in raw syngas (at reactor outlet) and after passing through the HT filter. For all tested points, one prevalent observation can be reported. Both water and tar contents were slightly reduced in the HT filter, even though the filter was operated at a moderate temperature of 450 ◦C and the filters were constructed from theoretically chemically non-active Al/Si material composites. For chalk, halloysite, kaolinite, and dolomite, the values of tar reduction were equal to 18.2%, 10.3%, 10.4%, and 16.9%, respectively. Thus, the CaO-based sorbent may give the best tar reforming characteristics while AlO–SiO materials take part in tar reforming to a lesser extent (halloysite, kaolinite). MgO is also known to participate in tar reforming in FB reactors; however, here, dolomite performed slightly worse than clean chalk. From current experience, it is hard to determine to what extent this e ffect is connected to the action of the sorbents themselves, because tar reforming on HT filters operated without sorbents (char alone) was also noted previously. It is noteworthy to say that most ceramic filters are built of mullite or other aluminosilicates and, thus, in their chemical composition, show similarities to halloysite and kaolinite. Thus, it was proposed that the reduction of tar content in the filter vessel may originate from a combination of basic thermal decomposition of high-molecular-weight tars or their polymerization and phase change to soot as a result of long residence time, which may be catalyzed by the presence of Ca, Mg, Al, and Si.


**Table 4.** Concentration of water, tar, and particle matter present in raw syngas (1) and after its filtration/sorption (2) in the HT filter.

In the analysis of particulate matter content of syngas, there is always grea<sup>t</sup> uncertainty connected with the determination of solids entrained from fixed-bed reactors. This subject is connected with technical limitations in the size of sampling probes, as well as the size distribution of particles entrained from a fixed bed of the reactor. For this reason, the particulate matter (PM) presented in Table 4 was only determined through the gas sampling of syngas downstream of the ceramic filter. The amount of PM in syngas upstream of HT filter was calculated through a mass balance of C and ash present in recovered filter cake. For T1, T2, and T4, no PM was measured downstream of the filter. Only for T3 was a small amount of PM registered at the outlet from the filter, indicating a breakthrough of dust which can happen if the layer of filter cake after pulse cleaning is too thin. Previously, another source

of PM in syngas filtered on ceramic elements was also found. In HT dry gas cleaning systems, where the installation is subjected to large quantities of tars and does not operate continuously (the pilot system is a research installation), small amounts of polymerized carbon deposits can form at cold spots of the ceramic filter and pipelines where the temperature of syngas is not su fficiently high to keep condensation of high-molecular-mass tars from occurring. At such places, tars tend to condense and polymerize, thus creating deposits of very brittle carbon structures. However, this problem was never noted at installations where tars originate from FB reactors, because these tars are of much di fferent composition (mostly light tars), and the concentrations of gravimetric tars rarely exceed 15 g/Nm3.

Proximate and ultimate analysis of chars recovered from the ceramic filter unit was done to assess the ratio of char entrained from the reactor in relation to the stream of fed sorbent (Table 5). For T1 and T4, the ratio of sorbent to char in gas was similar as can be seen from the proximate and ultimate analysis of the filter cakes. Similarly, T2 and T3 had similar char-to-sorbent ratios. Filter cakes from T1 and T4 were much richer in C and showed much smaller amounts of ash. Regarding the filter cake obtained from the reference test, the sample was a mixture of biomass char and ash with high C, A, and Cl contents. It can also be seen that the ratio of char to the sorbent for T1–T4 was equal to 1:0.9; 1:4.2; 1:5.0, and 1:1.25, respectively.

**Table 5.** Proximate and ultimate analysis of filter cakes recovered from HT filtration/sorption of syngas (as received).


Subsequently, qualitative and quantitative analyses of AAEMs present in filter cake samples were performed, and Table 6 collates the results. Values presented in parentheses in each cell correspond to a given content of the ash species in the fresh sorbent. The last column of the table is fitted with data from the analysis of clean char from the gasification process of wood chips where no co-filtering agen<sup>t</sup> or sorbent was used. Based on this comparison, the analysis of changes in ash chemistry was possible, giving grounds for raising conclusions as to which AAEMs are volatilized during gasification in the GazEla reactor and can be captured from syngas with the use of the filtration/sorption method.

It is visible that, for all test points, dilution e ffects of the sorbents with ashes from biomass took place. Thus, with the use of the mixing law, it was checked whether there were any high deviations from the mixing proportions of ashes obtained from the sorbents and biomass chars. Such findings can indicate the capture of species volatilized in a reactor, as well as volatilization from the filter cake, thus indicating lower e fficiency for its removal.

To some extent, chars from di fferent tests interact with each other and are extracted together from the filter vessel even after prior cleaning of the elements and their precoating with another sorbent. Thus, for T1 and T4, it can be seen that the sorbents used were rich in Ca and Ca–Mg, respectively. Sorbents applied in T2 and T3 had similar chemical compositions (rich in Al and Si); however, halloysite is also naturally rich in Fe. It can be seen here that the sample from T2 (halloysite) was partially polluted with chalk from T1, whereas samples from T4 (dolomite) contained some kaolinite from T3. It is interesting that halloysite, which contains high quantities of Fe, did not contaminate the sample from T3. This finding may be partially supported by the observed good filtration-enhancing properties and flowability of halloysite. Noteworthy also, even though the amount of ashes in biomass was small and, thus, the dominance of compounds originated from sorbents was very high (ratio of ashes

1:10–1:50) for all samples, the effect of ash enrichment in constituents characteristic for biomass ashes such as K and Na was high.


**Table 6.** Alkali and alkali earth metal content in chars forming filter cake on the surface of the ceramic filter (unit: wt.%).

For a more visual comparison of the results, Figure 6 was prepared where samples were compared through mass balance analysis, and the results are presented as ratios of species calculated from the mass balance related to the concentrations measured in samples. Thus, values above 1 indicate here that the content of the specie in the filter cake was lower than expected from the mixing law and may indicate lower removal efficiency of the element. On the other hand, values below 1 indicate that the content of an element in the filter cake was higher than predicted and, thus, may indicate its preferential removal from syngas. Results fitting into a range of ±25% are generally accepted to be in agreemen<sup>t</sup> with the mixing law. The streams of char and sorbents were all in good agreement, and the ratios of values calculated ranged from 0.92–0.97.

**Figure 6.** Ratio of the major elements in filter cake calculated from the mixing law of sorbents and char vs. values measured analytically in recovered filter cake.

The best correlation with the mixing law was found for dolomite, where only concentrations of P and Na in filter cake were well below half of the expected value, which may indicate lower retention in the filtration/sorption system. At the other end of the scale was chalk, which was found to be in good agreemen<sup>t</sup> only for S, while lower than expected contents of K (2.17), Na (5.17), and P (8.19) were noted (P and Na values are o ff the scale presented in Figure 6).

For halloysite, K was the only measured element which might indicate preferential adsorption (0.53) from syngas. For all sorbents, the scatter of the results indicates that it is very di fficult to close mass balances of elements in fixed-bed gasification systems. Primarily for halloysite, S concentration in the filter cake was more than 20 times lower than the mass balance indicates (o ff the scale in Figure 6). On the other hand, for dolomite, the Na content in filter cake was equal to less than 4.5 times the prediction. For future work, it is, thus, advised to try applying other analytical procedures in the search for these elements in syngas; however, as already mentioned, ion chromatography of gas sampled through absorption in NaOH does not provide as clear results as hoped for.

The di fficulty in closing mass balances of elements in gasification systems can be seen here. A sampling of char from syngas in the real installation is only possible for a test designed specifically to be done without the use of sorbents. Even though the biomass source was kept constant during the research, its exact composition, as well as its gasification conditions, varied. Taking into account that components of biomass ash volatilize in a manner related to their ash composition and process parameters, the char bed does not remain constant nor does its ability to readsorb AAEMs from raw syngas.
