Bioresource Technology 101 (2010) 9387–9390

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage:

Short Communication

Empirical modelling and dual-performance optimisation of a hydrogen sulphide removal process for biogas treatment
Gabriela Soreanu a,b, Patricia Falletta a,*, Michel Béland a, Kara Edmonson a, Bert Ventresca c, Peter Seto a

Environment Canada, Aquatic Ecosystems Management Research Division, 867 Lakeshore Road, Burlington, Ontario, Canada L7R 4A6 Gh. Asachi Technical University of Iasi, Faculty of Chemical Engineering and Environmental Protection, Iasi 700050, Romania c Environment Canada, Chief Information Officer Branch, 867 Lakeshore Road, Burlington, Ontario, Canada L7R 4A6

a r t i c l e

i n f o

a b s t r a c t
This study was conducted in order to investigate the use of a simple, low-cost technology for the removal of hydrogen sulphide (H2S) from biogas, consisting of an anoxic biotrickling filter. Modelling and optimisation of the process was achieved by studying two independent variables (H2S concentration and biogas flowrate) and two simultaneous performance criteria (H2S removal efficiency (%) and H2S loading rate (g/ (m3 bed day)), which were inversely related. The experiments were carried out on a bioreactor with a 12 L packing volume. H2S concentration and biogas flowrate were varied in the range of 2000–4000 ppmv and 10–70 L/h, respectively. A model sensitivity analysis indicated the influence of the process variables on the bioreactor performance. Process optimisation was undertaken on a H2S removal efficiency basis, while maintaining a target H2S loading rate, depending on the desired quality for the biogas use and the technological requirements. Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved.

Article history: Received 29 February 2008 Received in revised form 13 June 2010 Accepted 24 June 2010 Available online 3 August 2010 Keywords: Hydrogen sulphide Anoxic Biotrickling Modelling Optimisation

1. Introduction Extensive research conducted over the course of the last decade in the field of applied biotechnology for the treatment of gas fluxes recognised non-conventional biological methods as an environmental-friendly and cost-effective alternative to more expensive traditional technologies, depending on the gas source, composition and the purpose of the treatment. Two criteria, the removal efficiency and the loading rate, are often considered in the determination of the process performance of a bioreactor treating gas streams. It should be noted, however, that in this process, these criteria have competing responses, such that the maximisation of one occurs with the minimisation of the other. Neither of these criteria alone completely describes the bioreactor performance (Devinny et al., 1999), rather both criteria need to be examined. A recent study focused on hydrogen sulphide removal from biogas under anoxic conditions (Soreanu et al., 2008a,b) describes the use of a biotrickling filter (BTF) in which simultaneous hydrogen sulphide oxidation in biogas and denitrification in wastewater, by specific denitrifying bacteria, takes place. Under nitrate nonlimiting conditions, the process performance of this dynamic system was found to be strongly dependant on H2S concentration and biogas flowrate. The goal of this study was to mathematically analyse the data obtained in these studies through process

modelling in order to understand the influence of experimental factors in the operation of the BTF and to determine the optimal conditions for a robust performance. 2. Experimental methodology 2.1. Experimental setup A PVC BTF packed with 12 L plastic fibres (Duststop Air Filters Inc., Petrolia, Ontario, Canada) was continuously fed with H2S rich biogas (prepared from digester and synthetic biogas, Soreanu et al., 2008a) at the bottom and with a recycled nitrate-based nutrient solution (prepared from sequencing batch reactor effluent and sodium nitrate, Soreanu et al., 2008a,b) at the top (Fig. 1). The biogas flowrate (Q) and H2S concentration (c0) were varied between 10 and 70 L/h and respectively between 2000 and 4000 ppmv, under nitrate-non-limiting conditions (nitrate excess), at a constant nutrient solution flowrate (500 mL/min), pH of 6.5 and at room temperature (23 ± 2 °C). Nutrient solution was recycled at a rate of 30.64 times/day, corresponding to an overall daily residence time of 640 min in the loop; the retention time of the liquid in one pass through the bioreactor was 20.88 min, versus 9–62 min for the gas (i.e. based on a 87% packing media void space). These experiments are summarised in Tables 1 and 2 and were selected from an extensive experimental program carried out for several months under various operating conditions (Soreanu et al., 2008a,b). The BTF response was expressed as H2S removal

* Corresponding author. Tel.: +1 905 336 4563; fax: +1 905 336 4858. E-mail address: (P. Falletta).

0960-8524/$ - see front matter Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2010.06.113

the absence of H2S in biogas for a short-time (i.3. m3/day. Nitrate was analysed according to SMEWW (2005) using a Technicon TRAACS Autoanalyser. 2001): zi ¼ xi Dzi þ z0i (4). The mathematical modelling. (5) and (6).9388 G.85 35. . g/m3 *.14 28. % N/A = Not applicable. x1. The H2S loading rate (R. % (real units).34 86. Y = H2S removal efficiency. Results and discussion 3. and an interaction model (Eq. microorganisms were able alternatively to use other S-degradation products available in the system (Soreanu et al. the quadratic factor interactions (x1 and 2 x2 were found insignificant (P > 0.. Except x2 1 in Eq. The determined R-sq values were above 98%. (1)) and represents an average of at least two values recorded daily during 2–4 days of steady performance. Rpred.14 75. A 91100-40 model Digi-Sense thermocouple thermometer equipped with T probe was used for the temperature measuring in the middle of the reactor. Cole-Palmer instruments were used for the gas flow rate measurement (PMR1-010296 model flowmeter) and for solution recycling (77200-62 model MasterFlex pump). Experimental design An orthogonal central composite experimental design (CCD) (Tables 1 and 2) was employed for the elaboration of the H2S removal efficiency objective function (Ypred) and the H2S loading rate restriction (Rpred) as a function of the coded independent variables: biogas flowrate (x1) and H2S concentration (x2). (4). State College.76 66. Real versus coded values.035 cents/ 1g of H2S removed (Harben and Tehune.48 84.28 50. Soreanu et al.00 10. R À Sq ¼ 98:4% ð5Þ Rpred ¼ 42:854 þ 32:145x1 þ 14:285x2 þ 10:715x1 x2 .00 14. Sensitivity analysis From the analysis of the regression coefficients in Eqs. (2)) was considered as a restriction for Y. Eq. Eq. one day) did not significantly compromise the biomass and the very fast recovery of the BTF’s removal efficiency was recorded. confirming good models fits for the experimental data. Q = biogas flowrate. R À Sq ¼ 100% ð6Þ where Ypred.. 2. efficiency (Y. The GC was coupled to the inlet and outlet gas sampling ports of the reactor. Table 1 Process variables and their variation interval.05) and excluded from the models. m3. It should be noted that the empirical models are valid only in the tested range of the operating parameters for which they were obtained. 3. % Response Y. Table 1). Fig. These models were validated for the significance level (a) of 2 0. Experimental BTF diagram.00 7. xi = coded values. L/h H2S concentration. In such cases. Process modeling A second-order empirical model (Eq.12 99. * conversion factor at 20 °C: [g H2S/m3] = 1.57 42. ppmv Center level (0)a 40 3000 Variation interval 30 1000 Superior level (+1)a 70 4000 Inferior level (À1)a 10 2000 a coded values.21 99.79 87. The pH of nutrient solution was measured by using a 912600 model Orion pH electrode coupled to a PBX0922110BA model ETRATON DS pH controller. (2008b) (0. Y pred ¼ 86:996 À 18:827x1 À 12:343x2 À 6:316x2 1 À 12:203x1 x2 . 2006). in terms of coded variables. model G280) equipped with a thermal conductivity detector and two analytical columns. as these criteria are inversely related.23 99. 2008b). the cost of raw material for the nutrient solution amounts to approximately 0. 3.4167 Á [ppmv H2S] Á 10À3 .1. V = packing bed volume.85 42.. Independent variable Biogas flowrate.2. (5). R = H2S loading rate. (6)) for the restriction Rpred. x2. The relationship between real and coded variables (Montgomery.e. exp.28 N/A 99. Analytical and measurements methods Biogas composition was analysed using a gas chromatograph (Agilent 3000A. cf = final H2S concentration. g/m3. varied between À1 and 1 (coded values. / Bioresource Technology 101 (2010) 9387–9390 Based on the nitrate consumption determined in Soreanu et al.05. 2. Dzi and z0i represent the values of the variation interval and center level. the graphical representation and the optimisation of the process performance were performed using Minitab 14 (Minitab Inc. No. (5)) was obtained for the objective function Ypred.25 Restriction R. Input variables x1 coded 1 2 3 4 5 6 7 8 9 10 +1 À1 +1 À1 +1 À1 0 0 0 0 x2 coded +1 +1 À1 À1 0 0 +1 À1 0 0 100. It is important to note that as observed in other trials (not shown). it can be observed that both factors (x1 and x2) have a negative effect on Y and a positive effect on R. gH2 S=ðm3 dayÞ ¼ where c0 = initial H2S concentration. Eq.%. PA). A MolSieve 5A Plot column was used for methane analysis and a Plot U column for the analysis of carbon dioxide and hydrogen sulphide. respectively. Table 2 Orthogonal central composite design matrix (CCD). since the increase of x1 and/or Y. where: zi = real values. % ¼ c0 À c f Á 100.%. c0 0 R R.2.32 g N-NO3À/g H2S removed).71 57. 1. as required by the simultaneous optimisation procedure described in the Process Optimisation section. % (normalised units) and x1 and x2 are coded variables. % ¼ 0 max Á 100 R Q Á c0 V ð 1Þ ð 2Þ ð 3Þ R0 .

using Eqs. R = H2S loading rate. P30%).98 a H2S rem.000 À1. The overlay of the Y and R contour plots. For this inlet H2S concentration. for the same H2S loading rate (R).25 times. since high Y and R are observed in opposite corners. Ypred. The obtained equations (7–12.24–1. but the discussion has to take into consideration that Y criteria is optimised with respect to R (i. The Y curvature in Fig. the linear effect of x1 appears to be more pronounced (1. 33. Soreanu et al. corresponding to about 238 g H2S/(m3 day)).e. the process performance was simulated for this scenario using Eqs.3. Alternatively. x1 = biogas flowrate.7 L) operated under equal percent variation (50%) between 25and 75 L/h biogas flowrate and 500–1500 ppmv H2S was affected by the biogas flowrate. 98. the removal performance of a similar but smaller scale BTF (6.00 40. Table 3) indicate that the effect of x2 on Y is 1. (4). 2008).000 45. depending on the quality of biogas supplied or required after treatment. (5) and (6). The evolution of the system response towards the optimum point according to formulation (13) is summarised in Table 4.e.G.78 94.00 55.000 a H2S concentration z1. .98 L/h. x2 occurs with the decrease of Y and respectively the increase of R.81 times more significant than x1. (5) and (6) according to the formulation (Montgomery. x2)}.54 30.98 49. % a H2S loading rate Rpred..300 0. Process optimisation A common method for formulating and solving optimisation problems involving multiple performance criteria was applied in this study... / Bioresource Technology 101 (2010) 9387–9390 Table 3 Ypred and Rpred models for equal x1 and x2 percent variation (33.00 Correspondence between the real and the coded values was achieved according to Eq.99 g H2S /(m3 day)) is considered.52 and 2.e. Y = H2S removal efficiency. x2 = H2S concentration. while maintaining at least 30% H2S loading rate (i. Performance models. Table 4 Dual-performance maximisation. H2S concentration = 2000 ppmv.000 40. This behaviour suggests a limited mass transfer due to the large ratio of substrate to cell density in the biofilm (Soreanu et al. x2) P Rtarget.99 62.533 0. L/h (real) 41.000 35.33%). In practical situations. 2001): Max {Y (x1.33%).99 277. the H2S loading rate (R) decreases with the decrease of x1 and the increase of x2. However. likely due to the larger x1 percent variation of ± 75% (40 ± 30 L/h) versus 33.86 96. 2.00 45.86%) can be achieved when the optimal determined operating conditions are: biogas flowrate = 41. The optimum points corresponding to the formulation (13) can be located by the overlay of Y and R contour plots. the concentration of H2S in biogas fluctuates depending on its source. the response of Y decreases with the decrease of x1 and the increase of x2. respectively) than the effect of x2. the optimum H2S removal efficiency (i.66–53. 2. as it is shown in Fig. Correspondence between the real and the coded values is presented in Table 1. % Rpred. 2 indicates the approach of the maximum.e. In order to compare the significance of the factors under equal percent variation (i. with restriction: R (x1. (6) and CCD.000 À1. Table 1. selected as a technological requirement according to previous investigations (Soreanu et al.766 98. this optimum H2S removal efficiency ensures outlet H2S concentrations below 150 ppmv. it can be observed that for the same Y. (5).00 90. From the intersection of the Y and R contour plots. This formulation allows the maximisation of H2S removal efficiency. Also. in terms of coded variables Model valid interval Biogas flowrate H2S concentration 2000–4000 ppmv Coded: À1 to 1 2000–4000 ppmv Coded: À1 to 1 2000–4000 ppmv Coded: À1 to 1 9389 Ypred ¼ 86:996 À 8:367x1 À 12:343x2 À 1:247x2 1 À 5:423x1 x2 ð7Þ ð8Þ ð9Þ ð10Þ ð11Þ ð12Þ 26. In order Fig. According to Eqs. Constraint Rtarget. These trends suggest that the simultaneous high Y and high R values are favoured at high x1 and low x2. As expected.000 À1. 2009). Table 4) may be selected.33% (3000 ± 1000 ppmv) used in the experiments. % Biogas flowrate x1 coded) 30. 2009). 2009).33 356. Eq. Y and R are antagonistic. (4).000 z2. ppmv (real) 2000 2000 2000 2000 0. which are within the stated maximum limits for the use of biogas for combined heat and power generation (500 ppmv) or for internal combustion gas engines (150 ppmv) (Deublein and Steinhauser. rather than H2S concentration (Soreanu et al.00 35. where Rtarget = 30% (13).e.33 L/h Coded: À1 to 1 35–70 L/h Coded: À1 to 1 16–32 L/h Coded: À1 to 1 Rpred ¼ 42:854 þ 14:286x1 þ 14:285x2 þ 4:762x1 x2 Ypred ¼ 78:054 À 14:052x1 À 17:427x2 À 2:149x2 1 À 7:118x1 x2 Rpred ¼ 56:247 þ 18:751x1 þ 18:749x2 þ 6:250x1 2 Ypred ¼ 95:240 À 3:223x1 À 5:834x2 À 0:449x2 1 À 3:254x1 x2 Rpred ¼ 25:710 þ 8:572x1 þ 8:570x2 þ 2:857x1 x2 3.99 x2 (coded) À1. eff. If the constrained Rtarget of 30% (237. other optimal operating conditions (i. gH2S/(m3 bed day) 237.66 317.066 0.

P. The use of H2S sensors coupled to automatic pipeline valves and gas flow controllers would enable a portion of the treated biogas to be introduced back into the untreated gas stream thus diluting it. The uniqueness of this biotechnology comes from the anoxic nature of the biological process. Weinheim. Acknowledgements The authors wish to thank Natural Resources Canada’s Decentralized Energy Production Initiative. They also wish to thank John Salvatore and Mohamad Al-Jamal (Environment Canada. / Bioresource Technology 101 (2010) 9387–9390 to ensure that the concentration of H2S did not exceed the capabilities of the process conditions. CO. Krukowski. when necessary. Biofiltration for Air Pollution Control. Seto. Society for Mining.. M... Alternatively. K... Béland. Sci. . 1999. Eng. Ventresca. Deublein. CRC/Lewis Publishers. Biogas from Waste and Renewable Resources. Webster. SMEWW (Standard Methods for the examination of water and wastewater). Wiley-VCH Verlag GmbH & Co. Falletta. J. Edmonson. D. Soreanu. 2005. Soreanu et al. M. An Introduction. Lttleton.. Berker. Environ..T.. Trivedi. Technol. Aquatic Ecosystem Management Research Division) for their important contributions to this study. S. S. Edmonson. Design and Analysis of Experiments.. Metallurgy and Exploration. Harben. 7. Laboratory pilot scale study for H2S removal from biogas in an anoxic biotrickling filter. 21st ed.KGaA. 2008. 2008a. Soreanu. the simultaneous maximisation of H2S removal efficiency and H2S loading rate has been described and validated models were proposed. Falletta. Tehune. M. A. Seto. Evaluation of different packing media for anoxic H2S control in biogas. allowing H2S removal from biogas under anaerobic gas phase conditions. J. A.. P. the results of this study indicate that the influence of H2S concentration on removal efficiency was more significant. John Wiley & Sons.J.E..M.. USA. 2006. Environment Canada and the Natural Sciences and Engineering Research Council of Canada for funding and administration of this research.. USA. and uses.. Investigation on the use of nitrified wastewater for the steady-state operation of a biotrickling filter for the removal of hydrogen sulphide in biogas.C. References Devinny. Technol.. Conclusion This study was focused on the mathematical analysis of the performance of a new BTF treating H2S in biogas. Particularly. In: Kogel. Inc. S. Germany. M. B. Nitrogen and nitrates industrial mineral & rocks: commodities. Montgomery. J.. 4.. P.. Deshusses. K. 2008b. under the experimental conditions specified in the paper... Water Sci. 30 (12). P. Béland.9390 G. P. Falletta. D. FL. BTFs in series could be installed for treating biogas flow with elevated H2S levels and for biogas polishing purposes. markets. Béland.. Environ.. 543– 552. 1249–1259. Boca Raton. Soreanu.. 2001.. Seto.T.W. G. Although the key factors in the control of biofilter performance were demonstrated to be the biogas flowrate and H2S concentration. G. G. P. C.. (Eds. P. automatic control of H2S could be performed. fifth ed. 201–207. 57 (2). Steinhauser.. American Public Health Association/American Water Works Association/Water Environment Federation. 2009.C.)..

Sign up to vote on this title
UsefulNot useful