Role of Primary Freeboard on Staged Combustion of Hardwood Pellets in a Fixed Bed Combustor

In staged fixed bed biomass combustion, primary air is supplied beneath the fuel bed with secondary air then provided above in the freeboard region. For fixed bed configurations, the freeboard is further divided into a primary freeboard length (LI), which is upstream of the secondary air and a secondary freeboard length (LII), measured from the secondary air all the way to the exhaust port. Despite extensive research into fixed bed configurations, no work has been successfully completed that resolves the effects of changing LI on fuel conversion, both in the fuel bed and within the freeboard of batch-type biomass combustors. In this study, experiments on a 202 mm diameter and 1500 mm long batch-type combustor have been conducted to determine the effects of changing primary freeboard length over three secondary to total air ratios (Qs/Qt) and two total air flow rates (Qt). The impact of these conditions has been studied on (i) intra-bed fuel conversion, measured through burning rate (kg/m−2 s−1), fuel bed temperature (°C) and ignition front velocity (mm-s−1), as well as (ii) post-bed fuel conversion in the freeboard, expressed through freeboard temperatures and emissions (NOx ppm, CO2%, CO ppm, O2%). The fuel used throughout the above experiments was Australian hardwood pelletised biomass. Results show that changes to primary freeboard length over LI = 200 mm, 300 mm and 550 mm, or LI/D = 1.00, 1.48 and 2.72, respectively, affect both intra-bed and freeboard (post-bed) performance indicators. The highest values of burning rate, ignition front velocity and fuel bed temperature were observed for interim values of LI/D = 1.48 at Qs/Qt = 0.25 and Qt = 0.358 kg/m−2 s−1. Primary freeboard lengths of LI/D = 1.00 and 1.48 were found to have higher freeboard temperatures, NOx and CO2 as well as lower CO and O2 values as compared to LI/D = 2.72 at Qs/Qt = 0.50 and 0.75. Increasing Qs/Qt from 0.25 to 0.50 for LI/D = 1.00 and 1.48 initially increased freeboard temperatures, with an accompanying increase in NOx and CO2 as well as decrease in CO values. However, further increase in Qs/Qt to 0.75 lead to lower freeboard temperatures for all primary freeboard lengths.


Introduction
Renewable energy sources, such as biomass, can play a significant role in providing energy security and reducing environmental impact [1]. Sustainably sourced biomass can be an alternative to fossil fuel usage in many existing combustion processes [2], particularly commercial-scale heat and power systems [3]. However, despite its potential for replacing fossil fuels, direct thermal conversion of biomass via combustion also requires continued insights into reducing or managing the resulting gaseous and particulate emissions as well as ash generation [4]. Whilst chemical and physical properties of biomass can significantly affect its utilisation [5], this paper addresses the role of process design (air staging) on important performance indicators such as temperatures, emissions and fuel consumption rates.
Commercial scale power plants in the range of 100-3000 MW [6] mostly incorporate moving, travelling or vibrating grate furnaces [7]. However, at the lab scale, fixed bed combustors have been mainly used due to the fact that their results can be used to analyse processes, performance and model conversion efficiency in commercial moving grate combustors [8]. Furthermore, laboratory-scale combustors are widely used because of their lower capital and operating costs, in addition to their good test reproducibility [9]. The bed arrangement in fixed bed combustors can be classified into two types. In the cocurrent configuration, the flame front and air stream move upward in the same direction; however, this produces high amounts of tar below the ignition front that may result in blockage of any underlying perforated grates or of the air inflow opening for the combustor. Additionally, the manufacturing cost of a co-current combustor is higher due to its complex design [10]. For these reasons, the present investigation uses the second type of fixed bed combustors, termed counter-current, in which the flame is ignited at the top of the bed and thus propagates downwards against the direction of upward flowing primary air [11].
Further to this, the air in fixed bed combustors can be staged with the total air divided between a primary stream that is mainly supplied through the plenum from beneath the fuel bed and a secondary air stream, supplied in the freeboard region above the fuel bed. In the absence of any secondary air supply, non-staged combustion involves all combustion air being supplied from beneath the fuel bed, as shown in Fig. 1(a) and (b). Accordingly, fuel conversion in counter-current and staged fixed bed configurations occurs over two zones: (i) drying, devolatilisation and combustion in the fuel bed, followed by (ii) post-bed conversion in the freeboard region. When secondary air is used, the freeboard can further be divided into the primary freeboard, comprising the space between the top of the fuel bed to the secondary air injection point, as well as the secondary freeboard, formed by the space downstream of point(s) for injecting secondary air. This paper focusses on the role of the primary freeboard on fuel conversion in both fuel bed and freeboard.
In counter-current fixed bed combustors, fuel ignition at the top of the fuel bed causes a downward propagating reaction front. This results in the release of volatile matter from the fuel bed, leaving behind a solid layer of char that is combusted [12]. The devolatilisation process in the reaction front is sustained by heat from partial oxidation of volatile gases (with the air), whereby it is labelled as an auto-thermal process [13]. Gaseous products rise through the solid layer of hot char, leave the fuel bed and can subsequently undergo further thermal conversion in the freeboard region due to secondary air [14]. In batch type fixed combustors, in which the fuel is not continually replenished through a screen feeder-like mechanism, the downward moving reaction front causes the majority of the volatile matter to be released, subjecting the char, consisting mainly of fixed carbon, to combustion [15]. This process continues until the entire packed bed is consumed where only ash is left behind as a solid by-product. With the above in mind, fuel conversion in fixed bed combustors appears to be mainly affected by three factors: (i) fuel thermochemical composition properties and its morphology [16], (ii) operational conditions such as residence time and air staging and (iii) the occurrence of instabilities such as channelling [17,18]. The current study focusses on the role of operational conditions, specifically air staging, in the counter-current combustion of densified biomass fuel in the form of commercially available pellets. Fig. 1 a Staged fixed bed biomass combustor, b non-staged fixed bed biomass combustor, c biomass combustor with load cell assembly, d various primary freeboard lengths (LI) investigated and e grate and spacer assembly Secondary emission reduction strategies, such as selective non-catalytic reduction (SNCR), are expensive to apply to small-scale domestic boilers running on biomass [19]. In response, studies like this that investigate fixed bed combustor operational conditions can help better understand the factors influencing process efficiency and emissions. Additionally, fuel conversion efficiency, emissions and burning rate are dependent on parameters such as primary and secondary air flow rate, insulation of combustor wall, fluid flow characteristics and air-fuel ratio [17]. Primary air flow rate mainly affects burning rate, ignition rate and the distribution of volatiles on the top of the fuel bed [20]. In fixed bed configurations, burning rate increases with the increase in primary air flow rate until it reaches a peak point [21,22], where further increase in primary air flow rate results in reduced burning rate. This has been attributed to non-uniformity of flow distribution inside the fuel bed and the onset of instabilities such as channelling [23]. Previous studies on staged biomass combustion show that the burning rate improves with a primary to secondary air ratio (Q p / Q s ) of 4:6 to 6:4 [24]. Whereas freeboard temperatures are mainly affected by primary air and Q p /Q s ratio [25]. Contrastingly, emissions appear to be mainly influenced by secondary air and Q p /Q s ratio, with a significant reduction in CO emissions achieved by reducing Q p /Q s ratio to 71% and 83% [25]. CO emissions can also be reduced by 50% in staged combustion of wood pellets as compared to non-staged, by using a uniform secondary air module at a higher position from the bed. However, NO x concentration can be reduced by 37% in a staged combustor, compared to non-staged, by using a non-uniform air distribution module at the higher flow rate with a lower distance from the fuel bed [26]. However, these studies on staged combustion of wood pellets did not present the effect of Q p /Q s and primary freeboard length on fuel bed temperatures.
Despite there being much earlier research on laboratory scale fixed bed biomass combustion, there is a lack of systematic investigation on the impact of air staging under varying primary freeboard lengths (LI) on overall combustion characteristics. The present study focuses on the effect of variation in primary freeboard length on the following: (i) In fuel bed performance indicators, including fuel bed temperature, burning rate and ignition front velocity, and (ii) freeboard performance indicators, such as emissions and freeboard temperatures, both upstream and downstream of the point for introducing secondary air. Outcomes from this study will provide a better understanding for the role of various process parameters and design variables on the operation of biomass combustors.

Methodology
Combustor Figure 1a-e shows that the fixed bed batch type combustor is comprised of three parts, namely a plenum for introducing primary air, a secondary air assembly and a column in which its unoccupied length constitutes both the primary freeboard (LI), upstream of secondary air and secondary freeboard (LII), downstream of secondary air. The plenum has a diameter of 202 mm, with two side ports supplying primary air. A stainless steel grate, placed above the plenum to carry the fuel and its spacer, has circular perforations with a total open area of around 32%. Further to this, flow conditioning mesh screens are placed inside the plenum. The combustor column is made of 310 stainless steel, with an inner diameter (D) of 202 mm, a maximum freeboard length (L) of 1500 mm and a wall thickness of 8.5 mm. Column walls have multiple thermocouple ports, which are positioned for the acquisition of axial temperature data inside the fuel bed and freeboard region. The secondary air assembly contains a circular ring with eight holes of 3 mm each that run along its inner periphery. The downpipe for this assembly enters via the combustor top flange and allows the location of the secondary air and relative distances of the primary freeboard (LI) and secondary freeboard (LII) to be adjusted for a given total length of the freeboard (LI + LII), as shown in Fig. 1d. Additionally, the total length of freeboard can also be changed, independent of fuel bed length, by using a variable height spacer as seen in Fig. 1e that elevates the fuel bed into the freeboard. By increasing the height of spacer legs, the remaining freeboard length can be reduced from its as-fabricated length of 1500 mm. The spacer contains another steel grate with an open area of 32%. In order to prevent the combustor from overheating, no wall insulation has been applied to the present tests.

Air Supply
Air supply to the combustor is provided by a screw type air compressor connected to a flow board with pressure regulation and four calibrated flowmeters (make: influx, model: LV2S15-AI 27) with an accuracy of ± 5% full-scale.

Temperature Measurements
Eleven thermocouples were used to measure freeboard and fuel bed temperature. The thermocouples were N-type, with a 300 mm long Nicrobel (protective) sheath of 3 mm diameter (make: TC Measurement, model: 2I-Nickel-Silicon-Magnesium). The temperature range of these thermocouples was − 270 to 1300 °C, with a resolution of 0.15 °C. Thermocouples were inserted through the combustor wall using stainless steel compression fittings to avoid leakage between the combustor and surroundings. All the thermocouples featured a factory integrated 5 m PVC insulated extension cable for connection to a high-density thermocouple module (make: National Instruments, model: NI 9213) mounted on the data acquisition system (model: cRIO-9074XT). Temperature data was acquired at a frequency of 0.2 Hz via a Labview interface (version: 2014). Similar frequency of temperature data was also used by Rashidian et al. [18]. Fuel bed has a total height of around 150-165 mm. Fuel bed thermocouples were located at x = − 50 mm and − 100 mm beneath the top surface of the fuel bed. Freeboard thermocouples were located according to the position of secondary air from the fuel bed. Ignition front velocity (IFV) was calculated using the temperature data of both fuel bed thermocouples. Whereas the data of fuel bed temperature is of peak fuel bed temperature observed during the experiment and is only from the thermocouple located at x = − 100 mm. However, since the current study was carried out in a batch-type combustor, the continuously depleting fuel bed also changed the distance of thermocouples with respect to fuel bed during the experiment. Further details of the thermocouple location are displayed in Table 1.

Load Cell
The combustor assembly was mounted on the load cell using a hydraulic lift platform, as shown in Fig. 1c. The load cell (make: Wedderburn, model; SUK300) had a measurement range of up to 300 kg and a sensitivity of 100 g. The sensitivity of 100 g corresponds to roughly 4 mm of fuel bed height based on the fuel's bulk density and combustor column geometry. Data from the load cell was logged using a digital indicator (make: Rinstrum, model: RAR320A) via its Rinstrumview software.
Mass loss of fuel in fixed bed combustion generally shows a linear pattern over time [27]. The measured mass loss in the current work was also noted to occur in a linear pattern. Fuel burning rate (Br . ) was obtained by using mass loss during the steady state period in relation to the cross-sectional area of the reactor, as shown in Eq. 1.
where, Δt is the time difference (seconds), Δm is the mass difference (kg) of fuel for selected time range during the steady state and A is the cross sectional area of fuel bed (m 2 ). It should be noted here that the values of Br . found in the literature [17,28,29] were calculated based on fuel bed thermocouple data, represented by ignition front velocity (IFV) and fuel bulk density (ρ fuel ) (Br . = IFV × ρ fuel ) because the fuel mass loss was not measured in related studies. The use of IFV for obtaining Br . does not consider the mass of remaining char in the combustor, leading to an overestimation of fuel mass flux when compared with direct measurement of fuel mass loss [30]. Consequently, in the current study, the burning rate was calculated using the incremental change in mass loss obtained during the steady state period. Additionally, the data from the fuel bed thermocouples was also used to calculate the IFV using Eq. 2.
where, d is distance between two adjacent fuel bed thermocouples (50 mm) and t is the time (seconds) required by the ignition front to transition between two adjacent thermocouples. In the current study, the desired combustion temperature for obtaining IFV was considered as 600 °C, similar to previous study [23].

Emissions
Gaseous emissions were measured using a Testo 350 X/ML gas analyser. A gas sampling hose connected with an industrial gas sampling probe was used to measure emissions of CO 2 , O 2 and NO x with a sampling rate of 0.2 HZ. The measurement probe was placed at the centre of the combustor exhaust pipe to provide an indication of relative changes in emissions over various test conditions. Accordingly, the measurement principles of various sensors in the analyser are based on a galvanic oxygen sensor (O 2 ), non-dispersive infra-red (CO 2 ) and electrochemical (NO x ). The analyser has a resolution of 1 ppm for NO for emission levels up to 3000 ppm, 0.1 ppm for NO 2 ppm for emissions levels up to 500 ppm and 0.01% for CO 2 and O 2 . Values of gaseous emissions were corrected to 10% of O 2 to allow comparison between this study and previous work done at similar experimental conditions [31]. Emission data presented here was measured during the steady state. The data from startup and shutdown has transient effects, as seen in Fig. A.1ab of supplementary materials. Therefore, it was neglected. Furthermore, the steady state was identified using the time series freeboard data as seen in Fig. A.1a-b.

Fuel and Test Conditions
Commercially available cylindrical biomass pellets (make: Maxiheat Hardwood Pellets) were used as fuel, with a diameter of 6.5 mm and a nominal length of 5-40 mm. Pellets were produced using hardwood and typically hardwood contains 20-25% hemicellulose, 39-41% cellulose and 17-24% lignin [32]. The moisture content of the pellets was 6.62%. Moisture content of the pellets was calculated using ASTM E871 standard [33]. Prior to testing, bulk density of fuel was determined as per ASTM E873 standard [34]. Table 2 shows the proximate and ultimate analysis of the fuel. Values for volatile matter, ash content and higher heating value (HHV) in Table 2 were determined using ASTM E872, ASTM D1102 and ASTM D5865-12 standards, respectively [35][36][37]. Fixed carbon was calculated by the difference of volatile matter, moisture and ash content. The standard deviation for parameters listed in Table 2 was calculated from three repeats. These ultimate analyses yield formula of fuel on dry basis and ash free as CH 1.45 O 0.62 and stoichiometric air to fuel ratio ( S ) of 6.24. For each test, the combustor was charged with a batch of 3.9 kg fuel using its side ports (60 mm dia) to form a fuel bed of around 150-165 mm in height. To ignite the fuel through the same side port, 15 mL of methylated spirt (96% ethanol) was injected and lit using a hand-held butane flame troch. To examine the effect of change in primary freeboard length (LI), three values for LI = 200 mm (LI/D = 1.00), 300 mm (LI/D = 1.48) and 550 mm (LI/D = 2.72) were investigated whilst keeping the secondary freeboard (LII) constant at 700 mm. This was the main feature of the approach used in this research as it means variations in primary freeboard are independent of secondary freeboard. These various primary freeboard lengths were achieved by changing spacer heights (100 mm, 350 mm and 450 mm) inside the combustor. Two total air flow rates of 0.358 kg-m −2 -s −1 and 0.192 kg-m −2 -s −1 were used, and the effect of three secondary to total air ratios (Q s /Q t ) of 0.25, 0.50 and 0.75 was also examined at these two total air flow rates. Figure B.1 of supplementary materials shows results from experiments undertaken to find the stoichiometric range of combustion air within the fuel bed by measuring the burning rate using primary air (Q p ) only over the range of Q p = 0.224 to 0.358 kg-m −2 -s −1 . Similar approach using burning rate to quantify the stoichiometric air in batch type combustors have been employed previously but not in conjunction with studying the role of primary freeboard whilst the secondary freeboard length is unchanged [38]. It can be observed from Fig. B.1 that burning rate increases as the primary air flow rate increases until a peak point. Further increase in the primary air flow rate reduces the burning rate. In this study, the maximum burning rate was observed for the air flow rate of 0.358 kg-m −2 -s −1 and therefore, was selected to investigate the effect of air staging. Decrease in the burning rate after a peak point with further increase in primary air flow rate was also observed in the study conducted by Porteiro et al. [38] as seen in Fig. B.1. In this context, any partitioning of Q t = 0.358 kg-m −2 -s −1 into supplying both the primary and secondary air renders combustion inside the bed subject to less than 0.358 kg-m −2 -s −1 , whereby the combustor operates in rich mode. The total air flow rate Q t = 0.358 kg-m −2 -s −1 was then selected as the baseline case, with Q t = 0.192 kg-m −2 -s −1 indicating fuel rich. Moreover, these total flow rates correspond to the excess air coefficient (n) = 0.48 and 0.62, respectively, which was calculated as per Eq. 3 [38].
It is obvious that increasing Q t from 0.192 to 0.358 kg-m −2 -s −1 will increase the burning rate. However, the effect of decrease in Q t on emissions in an air-staged combustor is not reported in the literature. Therefore, a lower Q t of 0.192 kg-m −2 -s −1 was also investigated in addition to Q t = 0.358 kg-m −2 -s −1 . Further details of the tested experimental conditions are listed in Table 1 as well as uncertainty analyses for experimental data type in Tables C.1 and C.2 of supplementary materials.

Results and Discussion
In resolving the sensitivity of fuel conversion relative to the primary freeboard length (LI) over the range of experimental conditions listed in Table 1, this section first considers effects on the packed fuel bed with regard to its burning rate (Br . ), fuel bed temperature (FBT) and ignition front velocity (IFV). This section then considers freeboard progress variables, such as temperatures, over the combustor column and emissions.
(3) n = Qt Sa.Br .   (Fig. 4), also evident in the lower temperatures immediately above the bed in the freeboard at LI/D = 2.72 as shown in Fig. 5. However, the fact that fuel conversion at LI/D = 1.48 appeared optimal across all freeboard stoichiometry (Q s /Q t ) and total flow rates (Q t ) indicates the need for further study into complex flow dynamics that may arise from introducing secondary air in the freeboard, which have only been studied to a limited extent [39].

Role of Primary Freeboard on Fuel Conversion in the Packed Bed
Fuel conversion in the packed bed improved as the Q s /Q t ratio decreased from 0.75 to 0.25 and the Q t increased from 0.192 to 0.358 kg-m −2 -s −1 across all freeboard lengths. The increase in the amount of secondary air (Q s ), at the expense of primary air (Q p ), occurred as Q s /Q t rose from 0.25 to 0.75. This, however, did not lead to improved fuel conversion compared to Q s /Q t = 0. However, the calculation of the burning rate using load cell data is more accurate as the data from IFV might be affected by channelling. Burning calculated using IFV may also be overestimated as the mass of remaining char inside the combustor is not considered in it. Figure 4(a) and (b) show the correlation between burning rate and fuel bed temperatures at different operating conditions. It can be observed that burning rate is directly proportional to fuel bed temperature, with the highest values, across all the primary freeboard lengths, observed at Q s /Q t = 0.25 and Q t = 0.358 kg-m −2 -s −1 . Higher values of burning rate and fuel bed temperature at this condition may be due to greater primary air flow rate through the packed bed (0.269 kg-m −2 -s −1 ), which is closer to the stoichiometric condition of Q p = 0.358 kg-m −2 -s −1 , Q s = 0. Decreasing primary air, by increasing Q s /Q t ratio and decreasing Q t , reduces both burning rate and fuel bed temperature irrespective of primary freeboard length. Whilst this is consistent with earlier observations that burning rate and IFV are significantly affected by primary air (Q p ) flow rate in a non-staged batch type fixed bed combustor [38], it further highlights that the optimisation of air staging (Q s /Q t ) and its location (LI) in freeboard can indeed influence upstream conditions inside the packed bed, which ultimately dictates the burning rate.    Fig. 2 and 3, optimises intra-bed progress variables (Br . , IFV and temperatures). As for the lower freeboard temperatures in the vicinity of the secondary air at Q t = 0.192 kg-m −2 -s −1 and Q s /Q t = 0.75 for LI/D = 2.72, these could be due to the distant location of secondary air from the fuel bed in addition to higher secondary air flow rate. Providing a higher amount of secondary air at this location did not improve the combustion of unburned volatiles emitted from the fuel bed, but rather it caused quenching. Whilst the secondary air caused a distinct (localised) increase in freeboard temperatures near the injection point for both LI/D = 1.00 and 1.48 at Q t = 0.358 kg-m −2 -s −1 for Q s /Q t = 0.25 and 0.50, increasing the primary freeboard to LI/D = 2.72 caused localised temperature jumps at 0.358 kg-m −2 -s −1 to become less apparent and limited to Q s /Q t = 0.25. As for the fuel-rich condition of Q t = 0.192 kg-m −2 -s −1 , there were no such distinct temperature jumps with any Q s /Q t ratio and primary freeboard. The localised temperature increase was due to combustion completeness by oxidising the product of incomplete combustion in the downstream of the secondary air [40]. Further to this, Fig. 5 shows that that freeboard temperatures increased with an increase in Q s /Q t from 0.25 to 0.50 for LI/D = 1.00 and 1.48 at Q t = 0.358 kg-m −2 -s −1 . Further increasing the ratio to 0.75 decreased the freeboard temperatures for these conditions. Additionally, for LI/D = 2.72, the increase in Q s /Q t ratio had a negative effect on freeboard temperatures, as with an increase in Q s /Q t ratio from 0.25 to 0.75, the freeboard temperatures decreased continuously. Figure 6a shows NO x emissions to have comparable values for both LI/D = 1.00 and 1.48, where these are also generally greater than those for LI/D = 2.72 at Q s /Q t = 0.5 and 0.75 for Q t = 0.192 kg-m −2 -s −1 and 0.358 kg-m −2 -s −1 . However, LI/D = 2.72 had higher NO x emissions compared to both LI/D = 1.00 and 1.48 at Q s /Q t = 0.25 and Q t = 0.358 kg-m −2 -s −1 . A similar trend of NO x emissions across all the freeboard lengths can also be observed for CO 2 emissions, as shown in Fig. 6b. Higher values of NO x and CO 2 emissions, for LI/D = 2.72 at Q s /  [18,26]. Furthermore, it can be observed from Fig. 6(a) and (b) that NO x and CO 2 emissions for LI/D = 1.00 and 1.48 increased with increase in Q s /Q t ratio from 0.25 to 0.50 and then decreased with the further increase in the ratio to 0.75. However, this trend was only limited to Q t = 0.358 kg-m −2 -s −1 for CO 2 emissions. This behaviour might be due to more efficient combustion of post bed volatiles in the freeboard region, which can also be observed in Fig. 5, with higher freeboard temperatures for Q s /Q t = 0.50 and Q t = 0.358 kg-m −2 -s −1 . The main reason for NO x formation from fuel-based NO x relates to HCN and NHi (i = 1, 2, 3), released during devolatilisation of fuel in addition to CO, H 2 , C x H y , H 2 O and CO 2 [41]. The highest temperatures in downstream of the secondary air were observed for 0.358 kg-m −2 -s −1 at Q s / Q t = 0.50 (1014 °C) (LI/D = 1.48). This higher temperature (1014 °C) may enhance the formation of NO. The combustion of CO, CH 4 and H 2 introduces O and OH radicals that can then be used to convert HCN (volatile) to NCO, which then reacts with O radical to produce more NO as observed in Eqs. 4-7 [42][43][44]. For the same total air flow rate of 0.358 kg-m −2 -s −1 , at Q s /Q t = 0.75, a higher secondary air flow rate decreased the temperature to 705 °C (Fig. 5) in the downstream of secondary air injection, thereby increasing O 2 concentration during HCN oxidisation that leads to an NO reduction. Furthermore, NO x and CO 2 emission decreased with the reduction in total air flow rate from 0.358 to 0.192 kg-m −2 -s −1 irrespective of freeboard length and Q s / Q t ratio. Interestingly, at Q t = 0.192 kg-m −2 -s −1 and Q s / Q t = 0.75, LI/D = 1.00 had significantly higher CO 2 emissions as compared to LI/D = 1.48 and 2.72. This significant reduction in CO 2 emissions for LI/D = 1.48 and 2.72 can be due to incomplete combustion related to instabilities and warrants further investigations. Figure 6c shows that that O 2 had an opposite trend from CO 2 emissions, where values of O 2 in the flue gas were lower for cases with higher CO 2 emissions (Q t = 0.358 kg-m −2 -s −1 ). As expected, the amount of O 2 in the flue gas kept on increasing with increases in Q s /Q t ratio from 0.25 to 0.75. This behaviour is due to an increasing amount of secondary air supply at the later stage of the combustor. Higher concentration of O 2 in the flue gas mainly indicates the availability of excess air that may contribute to lower freeboard temperatures [25]. O 2 values reported here are generally higher than normal biomass combustion. This can be due the location of gaseous emission measurement probe. As the probe was measuring emissions at the exhaust of the combustor, oxygen concentration from the environment may have effect on the reading. The trends of freeboard temperature for the three primary freeboard heights tested can be correlated to O 2 values in the flue gas. As the ratio of Q s /Q t increases from 0.25 to 0.50 for Q t = 0.358 kg-m −2 -s −1 , despite increasing secondary air supply, O 2 values in the flue gas decreased for LI/D = 1.00 and slightly increased for LI/D = 1.48. Therefore, freeboard temperatures in these cases are also higher for Q s /Q t = 0.50 as compared to Q s / Q t = 0.25. As the ratio further increased to Q s /Q t = 0.75, O 2 values increased significantly for LI/D = 1.00 and 1.48, causing a drop in freeboard temperatures due to convective cooling caused by higher secondary air flow rate [26]. This behaviour can also explain the localised increase in freeboard temperature in the vicinity of secondary air for for the same Q s /Q t ratio and primary freeboard length that resulted in lower freeboard temperatures. This observation shows that supplying higher secondary air flow will not be sufficient to complete the combustion of unburned hydrocarbons in the freeboard region if primary air flow rate is low. Although the secondary air supply is important for combustion in the second stage, this process is also equally dependent on the supply of primary air. Interestingly, O 2 values for LI/D = 1.00 at Q s /Q t = 0.75 and Q t = 0.192 kg-m −2 -s −1 were significantly lower than LI/D = 1.48 and 2.72 cases, which were subjected to instabilities and are further discussed in the next section.

Emissions
Carbon monoxide emissions from biomass combustion are influenced by excess air ratio, primary air flow rate, secondary air flow rate, mixing of air and flue gas in the secondary combustion zone, lower concentration of oxygen in the freeboard region and lower residence time, as well as fuel bed and freeboard temperatures. Additionally, wall effects and the presence of char in fixed bed reactors may also increase the formation of CO in small laboratory scale combustors as compared to other large-scale biomass combustors [26]. CO emissions for the investigated experimental conditions are presented in Fig. 6d. Previous studies have presented that CO emissions in laboratory-scale combustors are higher than the standard norm [18,45]. The cases with * in Fig. 6d have CO emissions above 125,000 ppm and were above the measurement range of gas analyser. These cases were for Q s /Q t of 0.25 at 0.358 kg-m −2 -s −1 and 0.192 kg-m −2 -s −1 and for Q s /Q t = 0.50 at 0.192 kg-m −2 -s −1 . In this study, CO emissions are very similar for all the freeboard lengths at a Q s /Q t of 0.50. However, as the Q s /Q t ratio increases to 0.75, smaller primary freeboard length, especially LI/D = 1.00, had lower CO emissions as compared to a high primary freeboard length (LI/D = 2.72). For Q s / Q t = 0.75 and Q t = 0.358 kg-m −2 -s −1 , CO emissions for LI/D = 1.00 were around 70.84% less than LI/D = 2.72. This value increased further to 90.34% for the same Q s /Q t = 0.75 but for a lower total air flow rate of 0.192 kg-m −2 -s −1 . Furthermore, this indicates that locating secondary air closer to fuel bed for low primary air flow in combination with higher secondary air flow rate can reduce CO emissions significantly. CO emissions reduced significantly to around 35,914 ppm for Q s /Q t = 0.50 and further to 534 ppm for Q s / Q t = 0.75 at 0.358 kg-m −2 -s −1 for LI/D = 1.00. A similar trend for CO emission reduction with increasing Q s /Q t ratio has also been observed in previous studies [25,46]. A higher CO concentration in some of the investigated cases can be related to the staged operation of the combustor. As in staged combustion, the primary combustion zone (fuel bed) of the combustor operates in fuel-rich conditions, whereby a large amount of unburnt species (including CO) form in the primary stage. The burnout of these unburnt species should take place in the freeboard. If sufficient air is provided in the freeboard, the CO level will be very low; otherwise, the CO level will be high. Accordingly, the higher deviation for CO emission in the staged combustion is a result of variations in secondary air excess ratio, where more fuel-rich conditions in the secondary stage can combine with imperfect mixing of flue gas and secondary air, causing higher fluctuations and concentration of CO. Figure 6d shows that for the same primary freeboard height of LI = 1.00, a lower value of Q t = 0.192 kg-m −2 -s −1 at Q s /Q t = 0.75, CO emissions are lower as compared to Q t = 0.358 kg-m −2 -s −1 at Q s /Q t = 0.25. This indicates that CO emissions, unlike burning rate, are more influenced by secondary air flow rate as compared to primary air flow rate. Additionally, CO and other gaseous emission concentration can also be influenced by residence time. Due to change in primary freeboard length in this study, residence time also changes. However, residence time is also affected by other operational parameters such as primary and secondary air flow rates.

Instabilities
Whilst investigating the experimental condition listed in Table 1, it was observed that some conditions were instable. These conditions are highlighted in green in Fig. 5. Instabilities during these conditions occurred at the start of the steady state period (temperatures inside the packed bed and freeboard had stabilised). When initiated, these instabilities also caused a sudden drop in freeboard temperatures in addition to being both audible and visible (supplementary material E: Video file). It would therefore appear that in some conditions, introduction of higher secondary air flow rate, as evidenced by greater values of Q s /Q t , results in lower temperatures due to convective cooling, thereby promoting incomplete combustion. These conditions also have low freeboard temperature, high CO emissions and low CO 2 emissions compared to stable conditions at similar Q s /Q t and total air flow rate. Instable conditions for LI/D = 1.48 and 2.72 at Q s /Q t = 0.75 and 0.192 kg-m −2 -s −1 produced CO emissions of roughly 104,208 ppm and 125,300 ppm, respectively. These emissions were approximately 761% and 935% higher than stable conditions for the same Q s /Q t ratio and total flow rate but with different primary freeboard length of LI/D = 1.00. Additionally, freeboard temperature in the downstream of secondary air for these conditions was around 312 °C, and 422 °C lower than stable conditions. Furthermore, freeboard temperature in the vicinity of secondary air for the instable conditions at a higher flow rate of 0.358 kg-m −2 -s −1 was 335 °C less than the stable condition of LI/D = 1.00. These instabilities were preliminarily related to the secondary air flow rate and distance between secondary air injection point and fuel bed top, i.e. primary freeboard. Further information about instable conditions can be found elsewhere [47].

Conclusions
Pelletised biomass was converted in a laboratory scale batchtype fixed bed combustor with the effects of primary freeboard length (LI) were investigated. The following conclusions can be drawn: The burning rate, ignition front velocity and fuel bed temperature were affected by the primary freeboard length. However, these indicators appear to have higher values for low and an interim primary freeboard lengths of LI/D = 1.00 and 1.48, with higher primary freeboard length of LI/D = 2.72 leading to lower values for these indicators, especially at Q s /Q t = 0.50 and 0.75. Similar peak freeboard temperatures were observed in the vicinity of the secondary air for LI/D = 1.00 and 1.48 at Q s /Q t = 0.50. Similarly, comparable NO x , CO 2 , emissions and O 2 values in the flue gas were also observed at these conditions. For LI/D = 2.72, at Q t = 0.358 kg-m −2 -s −1 , lower values of freeboard temperatures, NO x and CO 2 as well as higher CO emissions were observed over the range of Q s /Q t = 0.50 and 0.75, except for Q s /Q t = 0.25 in which no appreciable variation occurred compared to LI/D = 1.00 and 1.48. Moreover, CO emissions were mainly affected by Q s /Q t ratio in comparison to primary freeboard length.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions

Conflict of Interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.