A study of black liquor and pyrolysis oil co-gasification in pilot scale

During the experimental run, the gasifier was continuously operated on BL and three PO/BL blends at eight different settings. One of the BL settings was repeated in triplicate. Fuel feed data for all the unique operating points (OPs) is presented in Table 1. In total, more than 130 tons of Kraft BL and ~5 tons of PO were gasified over 5 days. In the present study, “standard operating conditions” denote a thermal load of 2.75 MW and a CH₄ content of 1.2 mol% in syngas, which was obtained by adjusting the oxygen flow to the reactor. This can be compared to a typical operating load of ~3.1 MW_th on BL. A slightly reduced thermal load was used to maximize operational stability and the potential for quantifying the effects of the investigated parameters.

The effect of the blending ratio was investigated by gasifying BL and blends with PO/BL ratios of 10/90, 15/85, and 20/80 on a mass basis at standard operating conditions. The effect of the reactor temperature was studied by operating the gasifier on a 15% PO blend at three different temperature settings denoted by syngas CH₄ fractions of 0.5%, 1.2%, and 2.1% and labeled “hot,” “standard,” and “cold,” respectively. Although the day-tank capacity placed an upper limit on the duration of co-gasification experiments, all the OPs were in place for at least 6 h, which is three times the hydraulic residence time of GL. Due to the partial dissociation of organic acids found in PO, a temperature rise of 5–10 °C was observed in the mixer. However, the temperature of the fuel feed to the reactor was maintained between 140 and 145 °C by regulating the temperature of the BL to the mixer. A detailed breakdown of BL and PO compositions is presented in Table 2. In comparison with BL, PO contains significantly more carbon, marginally more oxygen, and virtually no ash. Meanwhile, a substantial fraction of the oxygen and carbon in BL is present as inorganic carbonates of sodium and potassium. Accordingly, PO has a considerably higher heating value than BL. Hence, at constant thermal load, the total fuel feed rate decreased with increasing PO fraction. The 10%, 15%, and 20% PO blends had PO/BL ratios of 17/83, 25/75, and 32/68, respectively, on an energy basis.

A number of calibrations were performed to improve flow measurement accuracy, and they are discussed in more detail in section A1 of the Supplementary Information. Media samples from the study were analyzed at a number of laboratories, which are collated in Tables 2 and 3. Note that the general procedure used for the calculation of the energy balances was detailed recently [7]. The standard conditions used as the reference points for the calculation of BL standard enthalpy of combustion were defined as T _ref = 25 °C and P _ref = 1 bar, with K₂CO₃ (s), Na₂CO₃ (s), CO₂ (g), H₂O (l), and Na₂SO₄ (s) as the stoichiometric combustion products. The total titrable alkali (TTA) was calculated by summing the sulfide, carbonate, and bicarbonate concentrations.

Similarly to coal, the difference in heating values between biomasses with low ash content has been correlated to differences in O/C and H/C ratios [17]. Although both of these ratios offer a useful means of fuel classification, they do not explicitly take into account the effect of variations in inorganic content on fuel properties such as heating value. Thus, for alkali-rich fuels like BL, a parameter such as alkali/C ratio, which decreases with increasing PO fraction, provides a more representative means of characterization. In the present work, the blends are classified in terms of their molar alkali-to-carbon ratio for quantitative comparisons.

The reactor temperature has been shown to be a nearly linear function of λ in entrained-flow gasification at steady-state conditions [18]. By definition, λ does not explicitly take into account variations in fuel oxygen content [19]. This can conceal the role played by the oxygen native to a fuel which, in most instances, is also available to the gasification reactions taking place in the reactor. In this study, the relative oxygen content (ROC) is used as a fuel-independent measure of the total oxygen available in the reactor. ROC, which has been used previously in studies on biomass gasification [20], is defined as follows [19]:

In Eq. 1, O_{2_additional} stands for gaseous oxygen added to the gasifier, O_{2_fuel} represents the fuel oxygen content, and O_{2_stoich} denotes the gaseous oxygen required for stoichiometric combustion. The λ values of all OPs are plotted as a function of ROC in Fig. 2. ROC and λ have a unique relationship for each distinct fuel composition, which is determined by both the oxygen content of a fuel and the amount of added gaseous oxygen. This is exemplified by the positions of the lines representing pure BL and the 15% PO blend relative to each other. However, for a given blend ratio, each point along the λ-ROC line represents variations in the amount of added gaseous oxygen alone, which in turn are caused by changes in desired operating temperatures. A comparison of the 10% PO OP with one of the pure BL OPs shows that fuels with different compositions can give rather similar λ values even when their gasification is likely to produce syngas with significantly different compositions. However, the differences between fuels are more clearly distinguishable in the ROC values, which capture variations in fuel oxygen content explicitly. Hence, compared with the λ value, ROC is believed to be better suited for directly comparing and assessing the effect of fuel composition on certain gasification performance parameters of varied biomass-based feedstocks.

Although Fig. 2 is based on a constant PO and BL solids composition (Table 2), it does incorporate the effect of variations in BL solids fraction, which are chiefly responsible for any deviations from the straight line. The solids fraction in BL from the SKKP mill has been shown to vary cyclically [7], and it went down by nearly 2% over the course of the present experimental run. On Fig. 2, the standard OP lies much closer to the cold OP than to the hot OP for the 15% PO blend. Since additional oxygen requirement is determined by controlling syngas methane content, the amount needed to reach the set value of 1.2% was less than expected. One potential cause may be an uncontrolled variation in BL solids composition, which led to a small decrease in BL carbon content, and thus the oxygen demand.

In view of the likely applications, BLG energy efficiency was quantified using three different parameters on a LHV basis. CGE_power takes into account the heating value of all syngas components. CGE_fuel only considers CO and H₂, the two most important components for chemical synthesis, while CGE_{fuel + S-free} is calculated on a sulfur-free basis to capture the importance of sulfur recovery to the pulp mill [7]. Similarly, the carbon conversion efficiency used in the present study was recently defined as follows [7]:

In Eq. 2, m _{C, fuel} is the mass flow rate of fuel carbon, η _carbon is the carbon conversion efficiency, and m _{C, TOC} is the mass flow rate of the dissolved organic carbon in GL (GL TOC). For pure BL OPs, m _{C, fuel} was equal to the mass flow rate of carbon in BL (m _{C, BL}) while for the blends, it was determined by adding together m _{C, BL} and the carbon in PO (m _{C, PO}). The m _{C, BL} was calculated by multiplying the averaged weight fraction of carbon in BL solids obtained from the ultimate analysis of two separate samples by the solids fraction and the mass flow rate of BL for a given OP. Meanwhile, the m _{C, PO} was obtained by multiplying the weight fraction of carbon in wet PO solids by the mass flow rate of PO. Similarly, m _{C, TOC} for each OP was calculated by multiplying the GL mass flow rate by the mass fraction of the dissolved organic carbon (TOC).

Note that the definition in Eq. 2 does not account for potential accumulation of char in the system. However, it is worth recalling that the quenching and cooling of syngas take place in a wet environment. The design of the cooling system ensures that both condensed tars from syngas and unconverted carbon fragments from the reactor end up and leave the gasifier in the liquid GL. Hence, char build-up is considered unlikely to occur under steady-state operation, as supported by the experience gained from continuous plant operation over long periods [1].

While the CO₂ content in syngas decreased with increasing PO fraction, the CO and H₂ contents increased almost linearly, as shown on Fig. 3a. Since the increase was less pronounced for H₂, the H₂/CO ratio decreased from 1.36 for BL to 1.23 for the 20% PO blend. Seeing as the fuel inorganic content decreased with increasing PO fraction, the increases in H₂ and CO yields can be attributed to the resultant reduction in reactor thermal ballast, which meant that fewer oxidation reactions were required to reach the gasification temperature. Accordingly, the amount of carbon bound to sodium and potassium in smelt also decreased with increasing PO fraction. The effect was so pronounced that, in spite of 5%-points lower fuel carbon throughput compared with BL, the 20% PO blend actually resulted in a higher combined yield of CO and CO₂ on a mass basis. Overall, the use of a 20% PO blend instead of pure BL yielded a 17% reduction in the additional oxygen demand per mole of H₂ + CO produced.

Figure 4a shows that the standard temperature setting, which is denoted by a syngas CH₄ fraction of 1.2 mol%, yielded the highest CO and H₂ flow rates for both BL and the 15% PO blend. The increase in reactor temperature from the cold to the standard setting led to a 7% increase in H₂ flow rate. However, a further increase in temperature promoted the production of CO₂ at the expense of CO, while the H₂ flow rate also exhibited a marginal fall.

The ROC value of a given fuel can be calculated easily provided composition data is available. Hence, in addition to use as a means of differentiation between fuels, under certain circumstances it can also be used as a tool for predicting yields of major syngas species as a function of fuel oxygen content. On Fig. 4a, the flow rates of H₂, CO, and CO₂ for the six points representing four different fuel compositions show a partial correlation with ROC to varying degrees. The yields of all three components are determined by the water-gas shift reaction controlled by thermodynamic equilibrium. Moreover, as was alluded to in Section 2.2, due to the effect of changes in, e.g., the alkali/C ratio, parameters such as heating value and by extension gasification temperature and additional oxygen vary non-linearly with the O/C and the H/C ratios. This is partly responsible for the inability to qualitatively identify variations in fuel composition from changes in ROC alone. At the same time, the absence of a correlation with the rate-controlled CH₄, and to a lesser extent the only partially equilibrium-controlled H₂S, can be seen clearly in Fig. 4b. In the case of CH₄, the changes in yield are solely a function of temperature, which is regulated by the amount of added oxygen, and thus described adequately by λ.

The temperatures recorded by thermocouples (TCs) at three different positions along the reactor length are plotted in Fig. 5 for both the PO/BL blends and pure BL as a function of their CH₄ content at the cold, standard, and hot temperature settings. In comparison with BL, the 15% PO blend yielded slightly higher temperatures at all three positions for the same CH₄ content. The results in Fig. 5 also confirm that the good correlation between the measured temperatures and syngas CH₄ for BL [8] holds true for the PO/BL blends as well. For the 15% PO blend, the correlation coefficients between the upper, middle, and lower temperatures and CH₄ were −0.99, −1.00, and −0.74. The thermocouple in the lower part of the reactor returned erratic readings both during and after the run that is circled on Fig. 5. The measured value is likely an underestimate and the erratic behavior may have been caused by a temporary build-up of smelt. In general, the correlations are particularly good for the upper and middle arrays which are located closer to the flame than the lone bottom TC, which is more likely to provide readings representative of the syngas temperature at the reactor exit. Note that the present results are in agreement with observations that were made during the gasification of sulfite thick liquor in an earlier study. [8]

The temperature readings from the TCs mounted in the reactor wall are susceptible to conductive interference from the wall and radiation from the very hot flame in the upper part of the reactor. However, experience has shown that the recorded readings can consistently reflect and mirror changes in operating conditions over a single experimental run. The temperature rises recorded in the present study are believed to be inconsequential in terms of their impact on the life of the refractory lining.

Syngas from the blends was very clean as, irrespective of fuel composition, C₆H₆ was the only higher hydrocarbon present in amounts greater than 20 ppm. Figure 6 shows that the flow rate of C₆H₆ increases with increasing CH₄ for both BL and the 15% PO blend. Interestingly, the blend yielded noticeably less C₆H₆ than BL at nearly the same CH₄ flow rate and the difference grew larger with decreasing temperature. The phenomenon is currently under investigation.

The fraction of feedstock sulfur released as H₂S, which is also referred to as the sulfur split, varied between 31% and 35% for the PO/BL blends and 28–31% for BL. Figure 3b shows that the decrease in fuel sulfur content with increasing PO fraction resulted in a corresponding decrease in the amount of sulfur release as H₂S. As shown on Fig. 4b, the H₂S flow rate decreased by 6–7% between the hot and the cold temperature settings for both BL and the 15% PO blend. The extent of the sulfur released in gaseous form has been shown to reach a maximum at ~500–600 °C [21], and higher temperatures are thought to favor sulfur recapture [22]. The sulfur that is either recaptured by volatilized sodium or retained in smelt reacts with steam and CO₂ according to the reaction in Eq. 3.

A thermodynamic equilibrium study of BLG found around two thirds of the feedstock sulfur in the gas phase as H₂S at temperatures above 800 °C and pressures above 20 bar [6]. However, the sulfur split for both the blends and BL shows that this was clearly not the case in the present study. Consequently, the reaction in Eq. 3 is only in partial thermodynamic equilibrium, and the final division of fuel sulfur between smelt and syngas at the exit of the reactor is also subject to kinetic limitations.

According to Le Chatelier’s Principle, an increase in reactor temperature would be expected to lead to an increase in Na₂S formation at the expense of H₂S, as the reaction in Eq. 3 shifts to the left. A partial shift did occur, as the fraction of gas-phase sulfur decreased with temperature by up to 4%-points for both BL and the 15% PO blend. On the other hand, the split was found to increase with PO fraction, by up to 6%-points for the 20% PO blend. In this case, the partial equilibrium is affected not only by the temperature, which increases only slightly with PO fraction, but also by the significant variations in the concentrations of the gaseous components.

Similar to previous studies [4, 7], a trivial amount of sulfur was also released in the form of COS at all OPs. The release was in the range 83–96 ppm, and it was not possible to identify any correlation with fuel composition due to the hydrolysis of COS in the quench tube [4]. The molar fraction of N₂ in syngas did not exceed 1.6% for any of the OPs. As the temperature of the gasification products falls after entering the quench tube, the water-gas shift reaction produces more CO₂ and H₂. The extent of this shift, which is influenced by the orientation and flow rate of the cooling sprays [4], could not be quantified, but the spray flow rates were set at 600 kg/h to ensure constant cooling rates.

Table 4 shows that aside from HCO₃ ⁻, which is dependent upon the magnitude of CO₂ absorption, the flow rate of the other elements and ions decreased in proportion, albeit not wholly, with the fuel inorganic fraction. Consequently, with the exception of the slightly anomalistic 15% PO OP, the GL became more dilute with an increasing PO fraction. In principle, the elemental concentrations in GL are a product of fuel inorganic content and the amount of water added to the GL dissolver through the three water streams shown on Fig. A.1 in the Supplementary Information. Although these streams provide a means of regulating GL concentration, due to the design and present setup of the cooling system, the range of possible concentrations is constrained by limits on minimum flows. Hence, in the interest of maximizing operational stability, the water flows into the GL dissolver were not reduced in proportion to the expected reduction in smelt flow. In theory, the system can be optimized to better regulate inflows and increase the TTA by rebuilding the lower part of the gasifier. It has previously been shown that some of the syngas CO₂ is absorbed in the GL [4], which results in the destruction of OH⁻ and the formation of HCO₃ ⁻ ions. The concentrations of these ions, which were present in all GL samples, did not appear to vary systematically with fuel composition.

In Kraft pulping, any sulfur present in a form other than Na₂S or K₂S does not play an active role in the cooking process, which means that sulfide reduction efficiency needs to be maximized. Sulfur reduction efficiency values of 100% within measurement uncertainty limits had previously been reported in an earlier study on the gasification of spent sulfite liquor [8]. Interestingly and somewhat unexpectedly, as seen in Table 4, the HS⁻ concentrations found in the GL were consistently lower than total S concentrations for both BL and the blends. At the same time, the correlation coefficient between the HS⁻ and the total S concentrations was 0.90, which supported the relative accuracy of the measurements. The methods used for the determination of HS⁻ and total S concentrations in GL have specified relative measurement uncertainties of 15% and 10%, which is of the same order of magnitude as the differences between the values (10–30% relative to HS⁻). Moreover, GL from the BLG pilot plant is extremely sensitive to oxidation during sample preparation, which makes the quantification of reduced sulfur forms, such as sulfide ions, very difficult. Consequently, a complete characterization of the GL sulfur species could not be carried out. Given the high reduction efficiencies seen previously, the differences between total S and HS⁻ concentrations may be due to systematic analysis-related errors. Nonetheless, in light of the above results and given the significance of sulfur recovery, the possibility of the presence of non-reduced forms of sulfur in GL needs to be considered in future work.

The changes in the total organic carbon, total inorganic carbon, and sludge content with increasing PO fraction are shown in Fig. 7. GL sludge fractions and TOC concentrations in the GL filtrate from both the blends and BL were comparable and did not vary systematically. In contrast, as expected, the TIC content decreased linearly since it is directly related to fuel inorganic content. Carbon found in formate ions constituted 23–41% of the TOC in the GL filtrate from the PO/BL blends and 24–49% from BL. The rest of the GL TOC is believed to be mainly tars condensed during quenching and gas cooling as discussed in a recent publication [7]. However, in the absence of a method for detailed TOC characterization, it is not possible to go much beyond reasoned speculation. In the aforementioned study, the carbon present as formate had made up 10–28% of GL TOC and measured formate concentrations had correlated reasonably well with the partial pressure of the CO in syngas [7]. However, such a correlation was not found here for reasons that are not presently clear.

Figure 8 shows that, irrespective of fuel composition, neither GL TIC nor GL sludge showed a meaningful change with temperature. While GL TOC did vary for both BL and the 15% PO blend, the changes did not appear to be systematic. In both Figs. 7 and 8, the error bars on the TOC concentrations represent an estimated measurement uncertainty of 15%. It can be seen that although the uncertainties overlap partially, the range of the variations exceeds measurement uncertainty, which was also found to be the case in an earlier study [7]. Coupled to the fact that TOC variations did not appear to be a function of the most probable process parameters such as fuel composition, temperature, or load, this points to the presence of currently unknown, non-controllable factors which influence GL TOC.

The gasification of the PO/BL blends yielded carbon conversion efficiencies of 98.8–99.5%. These values increase to 99.0–99.7% if the carbon present as formate in GL is considered “converted” for the purpose, given its hypothesized origin in syngas CO. The completeness of conversion was also supported by the absence of any detectable unconverted char in GL. Char gasification is the rate-limiting step controlling carbon conversion, and it is greatly facilitated by the intrinsically high reactivity of alkali-rich BL char [9]. A plot of the carbon conversion efficiencies for all the OPs as a function of their ROC values (not shown) found there to be no correlation between the two variables. It may thus be concluded that the changes in η _carbon were independent of changes in blend ratio, fuel composition, and temperature. As a corollary, the decrease in alkali loading with increasing PO fraction did not have a discernible effect on carbon conversion in the range studied.

In energetic terms, the value of increasing the PO fraction at a constant thermal load can be clearly seen from Fig. 9a. As discussed in Section 3.1, the CO and H₂ contents in syngas increased markedly with the increasing PO fraction at the expense of CO₂. As a result, both CGE_power and CGE_fuel also increased. In comparison with BL, the 20% PO blend yielded a 7%-point higher CGE_fuel, a 7%-point higher CGE_power, and a 5%-point higher CGE_{fuel + S-free}. CGE_{fuel + S-free} is the ratio of syngas LHV_fuel, which is presented in Table B.2 of the Supplementary Information, and fuel LHV_S-free, which is influenced by the fuel S fraction. The decrease in syngas H₂S content with the increasing PO fraction meant that while it accounted for, on average, approximately 5% of LHV_power in the syngas from the BL, the same figure dropped to only 3.4% in the syngas from the 20% PO blend.

Figure 9b shows that due to the presence of CO and H₂ yield maxima, the potential for fuel production peaked at the standard temperature setting. Syngas contained significantly less CO and H₂ at the lower temperature setting, while at the higher temperature setting, some of the CO was oxidized to CO₂, thereby resulting in a slightly lower CGE_fuel. Meanwhile, changes in CGE_power are closely linked to the CH₄ content of syngas for a given fuel composition. At the cold temperature setting, CH₄ made up approximately 9% of syngas LHV. Due to the conversion of some of this CH₄ into CO and CO₂, the share dropped to 5–6% at the standard temperature setting. Consequently, CGE_power at the standard temperature might be expected to be slightly lower, but as can be seen from Fig. 9b, the values of CGE_power at the high and standard settings are very similar, which is very likely due to experimental error. At the hot temperature setting, much of this CH₄ was converted to CO₂, thereby leading to a decrease in the syngas heating value. It is likely that the standard temperature setting represents a close-to-optimal point for biofuel production within the operating envelope of the gasifier.