Microbial kinetic for In-Storage-Psychrophilic Anaerobic Digestion (ISPAD)

Download Microbial kinetic for In-Storage-Psychrophilic Anaerobic Digestion (ISPAD)

Post on 22-Feb-2017

217 views

Category:

Documents

4 download

Embed Size (px)

TRANSCRIPT

<ul><li><p>lable at ScienceDirect</p><p>Journal of Environmental Management 146 (2014) 59e68Contents lists avaiJournal of Environmental Management</p><p>journal homepage: www.elsevier .com/locate/ jenvmanMicrobial kinetic for In-Storage-Psychrophilic Anaerobic Digestion(ISPAD)</p><p>Mahsa Madani-Hosseini, Catherine N. Mulligan, Suzelle Barrington*</p><p>Department of Building, Civil and Environmental Engineering, Concordia University, 1455 de Maisonneuve, Montreal H3G 1M8, Canadaa r t i c l e i n f o</p><p>Article history:Received 9 May 2014Received in revised form21 July 2014Accepted 22 July 2014Available online</p><p>Keywords:Psycrophilic anaerobic digestionSwine manureKinetic coefficients* Corresponding author. Tel.: 1 1 450 773 6155x6E-mail address: suzellebarrington@sympatico.ca (S</p><p>http://dx.doi.org/10.1016/j.jenvman.2014.07.0280301-4797/ 2014 Elsevier Ltd. All rights reserved.a b s t r a c t</p><p>In-Storage-Psychrophilic-Anaerobic-Digestion (ISPAD) is a wastewater storage tank converted into ananaerobic digestion (AD) system by means of an airtight floating geo-membrane. For process optimi-zation, ISPAD requires modelling with well-established microbial kinetics coefficients. The present ob-jectives were to: obtain kinetics coefficients for the modelling of ISPAD; compare the prediction of theconventional and decomposition fitting approach, an innovative fitting technique used in other fields ofscience, and; obtain equations to predict the maximum growth rate (mmax) of microbial communities as afunction of temperature. The method consisted in conducting specific Substrate Activity Tests (SAT) usingISPAD inoculum to monitor the rate of degradation of specific substrates at 8, 18 and 35 C. Microbialkinetics coefficients were obtained by fitting the Monod equations to SAT. The statistical procedure ofLeast Square Error analysis was used to minimize the Sum of Squared Errors (SSE) between the measuredISPAD experimental data and the Monod equation values. Comparing both fitting methods, thedecomposition approach gave higher correlation coefficient (R) for most kinetics values, as compared tothe conventional approach. Tested to predict mmax with temperature, the Square Root equation betterpredicted temperature dependency of both acidogens and propionate degrading acetogens, while theArrhenius equation better predicted that of methanogens and butyrate degrading acetogens. Increasingtemperature from 18 to 35 C did not affect butyrate degrading acetogens, likely because of theirdominance, as demonstrated by microbial population estimation. The estimated ISPAD kinetics co-efficients suggest a robust psychrophilic and mesophilic coexisting microbial community demonstratingacclimation to ambient temperature.</p><p> 2014 Elsevier Ltd. All rights reserved.1. Introduction</p><p>In-Storage-Psychrophilic-Anaerobic-Digestion (ISPAD) is theconversion of a standard exterior storage tank into an anaerobicdigester by means of an airtight geo-membrane cover collectingbiogas (King et al., 2011b). With ISPAD, the tank is gradually filledover an extended period of at least 100 days, thus compensating forits psychrophilic temperature fluctuating with ambient climaticconditions. For ISPAD, microbial acclimation, protein degradation(King et al., 2011a, 2011b) and biogas generation were investigated(Giard et al., 2013; Nohra et al., 2003). Nevertheless, process opti-mization is still required through modelling with well-establishedmicrobial kinetic coefficient values, such as maximum microbialgrowth rate (mmax) and fluctuation with temperature, microbialyield (Y), and substrate half saturation constant (Ks). In AD systems,microbial kinetics coefficients are useful modelling parameters to03; fax: 1 450 773 3133.. Barrington).evaluate the behaviour of microbial population, the rate of sub-strate degradation and the biogas production (Jimenez et al., 2006).Based on the determination of kinetics coefficients, Nwabanne et al.(2009) concluded that the rate of digestion of a municipal solidwaste digester could be corrected through inoculation.</p><p>A variety of methods have been used in kinetic parameteridentification, because the AD process is characterized by its highcomplexity and non-linearity. This causes variability in reportedkinetic parameter values even when the same operational andenvironmental conditions have been evaluated (Donoso-Bravoet al., 2011). The mode of operation (e.g., batch vs. continuous)and the environmental and operational conditions (e.g., pH, tem-perature, organic load) are other factors resulting in kineticparameter variability (Pavlostathis and Giraldo-Gomez, 1991). TheISPAD system operates at a temperature varying with that ofambient and under a steadily decreasing loading rate betweenbatch-feeding, as compared to constant parameters in conventionalAD systems. Therefore, ISPAD systems operate under very specificmicrobial kinetic values which need to be determined.</p><p>Delta:1_given nameDelta:1_surnameDelta:1_given namemailto:suzellebarrington@sympatico.cahttp://crossmark.crossref.org/dialog/?doi=10.1016/j.jenvman.2014.07.028&amp;domain=pdfwww.sciencedirect.com/science/journal/03014797http://www.elsevier.com/locate/jenvmanhttp://dx.doi.org/10.1016/j.jenvman.2014.07.028http://dx.doi.org/10.1016/j.jenvman.2014.07.028http://dx.doi.org/10.1016/j.jenvman.2014.07.028</p></li><li><p>Nomenclature, subscript and expressions</p><p>SubscriptM Acetoclastic methanogensAB Butyrate degrading acetogenicAP Propionate degrading acetogenicA Acidogensglu Glucosepr Propionatebut Butyrateac Acetate</p><p>SymbolAD Anaerobic DigestionISPAD In-Storage-Psychrophilic Anaerobic DigestionVFA Volatile Fatty AcidGRG General Reduced Gradientmmax Maximum microbial growth rateKs Half saturation constantY Yield coefficientKd Decay rate constantm Microbial growth rateX Microbial population densityBMP Biochemical Methane Production</p><p>SAT specific Substrate Activity TestSSE Sum of Square ErrorTObsi Observed dataTPredi Model predicted outputn Number of dataK Reaction rateR* Universal gas constantEa Activation energyb Regression coefficientT TemperatureTmin Apparent minimum temperatureMAE Mean Absolute ErrorMSE Mean Squared ErrorFB Fractional BiasRSME Root Mean ErrorNMSE Normalized Mean Squared ErrorMAPE Mean Absolute Percentage ErrorR2 Coefficient of determinationR Correlation coefficientCOD Chemical Oxygen DemandTS Total SolidVS Volatile SolidVSS Volatile Suspended SolidVDS Volatile Dissolved SolidFS Fixed Solid</p><p>M. Madani-Hosseini et al. / Journal of Environmental Management 146 (2014) 59e6860The objectives of this study were to: 1) statistically fit laboratoryspecific substrate activity test (SAT) results to that of the Monodequation applied in series to the various AD microbial groups togenerate ISPAD kinetics coefficients and microbial densities; 2)compare the prediction accuracy of the conventional fittingmethod to that of the decomposition approach, and; 3) comparethe prediction accuracy of both the Arrhenius and Square Rootequations to describe the relationship between temperature andmmax. The laboratory data was produced using SAT conducted withISPAD inoculum acclimated to swine manure and individual sub-strates namely, glucose, propionate, butyrate and acetate, at 8, 18and 35 C. The curve fitting process was limited to known kineticparameter ranges and yielded microbial population densities, X, forall 4 main degradation groups.</p><p>2. Kinetic coefficient determination</p><p>A variety of laboratory AD methods are used to obtain data forthe computation of microbial kinetics coefficients through statis-tical fitting. Three different feeding methods are used, namelybatch (Donoso-Bravo et al., 2009; Flotats et al., 2003), continuous(Batstone et al., 2009; Bernard et al., 2001) and fed-batch (Redzwanand Banks, 2004; Rodrigues et al., 2003). Two batch tests arecommonly used in the determination of AD kinetics coefficientsbecause of their simplicity and short experimental duration:biochemical methane production (BMP) and specific substrate ac-tivity test (SAT). The BMP assay is a long-term batch incubation ofcultures with periodic biogas and reactor content sampling toprovide a temporal profile of substrate consumption and methaneproduction from a known initial concentration of active biomassand substrate (Chynoweth et al., 1993; Shelton and Tiedje, 1984).However, the SAT is a shorter assay measuring the rate of con-sumption of individual substrates used by one of the main ADmicrobial groups. Such an assay provides more specific data for thefitting of individual AD process. Donoso-Bravo et al. (2009) usedstarch, glucose and acetic acid as the main substrates in SAT toobtain kinetics coefficients for hydrolysis, acidogenesis and meth-anogenesis, respectively. Using SAT, Flotats et al. (2003) also esti-mated kinetics coefficients for the AD of valerate.</p><p>The experimental data is used to estimate kinetics coefficientsby changing their value till the model predicts a response corre-sponding to that obtained experimentally. The selection of anappropriate set of modelling equations is important. In kineticscoefficient estimation, the simple Monod equation is accurate andsimple when a limited number of parameters are unknown. For AD,the kinetic values of mmax, Ks, Y, and X are interdependent, whichmakes their estimation difficult, challenging and tedious. Used inother scientific fields, the decomposition approach is applied in thiswork as an innovative method of estimating interdependent pa-rameters through data fitting (Bahn et al., 1996; Harjunkoski andGrossmann, 2001). This approach requires the decomposition ofthe designed algorithms into sub-problems and the solving of in-dividual parameters by data fitting, in order of importance. Theadvantages of this approach are to: 1) reduce the possibilities ofdropping into local minima in the process of optimization; 2)reduce the computation time; and 3) increase the fitting accuracycompared to other approaches (Jiang and Cheng, 2005).</p><p>Using the decomposition approach, each sub-problem can besolved through an objective or cost function, recognized as themost common fitting tool. The most useful cost function is the Sumof Square Error (SSE) (Batstone et al., 2009; Donoso-Bravo et al.,2010; Lopez and Borzacconi, 2010; Noykova and Gyllenberg, 2000)which assumes that the standard deviation of the measurementerrors is constant:</p><p>SSE Xn</p><p>i1</p><p>tobsi tpredi</p><p>2(1)</p><p>where the observed data, the model-predicted outputs, and thenumber of data are shown as tobsi , t</p><p>predi , and n, respectively.</p><p>To optimize the cost function, several algorithms were devel-oped, namely grouped under the Exact and Heuristic methods. An</p></li><li><p>M. Madani-Hosseini et al. / Journal of Environmental Management 146 (2014) 59e68 61Exact method is a mathematical procedure generating a sequenceof solutions improving the order of a class of problems. Thismethoduses convexity assumptions for the cost function to obtain theoptimized values. The main drawback is the risk of getting trappedin local minima, specifically when the cost function is non-linear.The common approach to solve this trap is to start the search us-ing several randomly selected initial parameters, which is called amulti-start strategy. Solving the problem more quickly, the Heu-ristic methods are multi-start by nature, but are highly sensitive tothe initial parameters. Also, the Heuristic methods are better atfinding an approximate solution when classic methods fail. This isachieved by trading optimality, completeness, accuracy, and/orprecision for speed.</p><p>Both the Exact and Heuristic methods were used in AD kineticdetermination. Examples for the Exact methods are: LevenbergeMarquardt (Garca-Ochoa et al., 1999; Lokshina et al., 2001),Sequential Quadratic Programming (Aceves-Lara et al., 2005; Sales-Cruz and Gani, 2004); Multiple Shooting (Lopez and Borzacconi,2010; Mller et al., 2002), and Direct Search also called SimplexAlgorithm (Haag et al., 2003; Mosche and Jordening, 1999;Simeonov, 1999). Applications of the Heuristic methods are:Simulated Annealing (Haag et al., 2003), Genetic Algorithms(Wichern et al., 2009), and Particle Swarm Optimization (Wolfet al., 2008).</p><p>The Heuristic methods require a certain level of knowledge andexperience, making them difficult and expensive to use, besidestheir sensitivity to initial values and stopping criteria. Conse-quently, Heuristic methods may be more inefficient than the Exactmethods with local optima issues.</p><p>Because of limitations offered by Heuristic methods, this studyused an Exact method, namely the multi-start strategy, to obtainISPAD kinetics coefficients. Since the risk of getting trapped in localminima is sensitive to the initialization parameters, the kineticscoefficient obtained for the Keshtkar et al. (2001) model were used.The Keshtkar et al. (2001) model was designed for manure sub-strates, includes inhibition factors for volatile fatty acid (VFA), pHand free NH3, and describes a cyclic batch reactor, similar to ISPAD.This model considers five steps: 1) the hydrolysis of particulatesubstrate by extracellular enzymes; 2) the consumption of solublesubstrates by acid forming bacteria; 3) the consumption of VFA; 4)the formation of acetate by propionate and butyrate degradingacetogens, and finally; 5) the consumption of acetate to generatemethane by methanogens.</p><p>3. Temperature functions</p><p>In AD systems, two equations are generally used to predicttemperature dependence, namely the Arrhenius and the SquareRoot equations. According to the Arrhenius equation (Equation (2)),the reaction rate roughly doubles for a temperature increase of10 C:</p><p>K AeEaR*T (2)</p><p>where K is the reaction rate, A is a constant, R* is the universal gasconstant (0.008314 kJ/mol K), T is temperature (K) and Ea is theactivation energy (kJ/mol).</p><p>The Square Root equation is simpler and purely descriptive ofthe evolution of microbial growth rate with temperature(Ratkowsky et al., 1982). The Square Root equation describes a lessthan optimal temperature adaptation of microbial growth in pureculture:</p><p>ffiffiffiffiK</p><p>p b</p><p>T Tmin</p><p>(3)where K is the reaction rate (or growth rate in the case of microbes)at temperature T (K), Tmin is the apparent minimum temperaturefor growth (K), and b is the regression coefficient.</p><p>In most AD models, the Arrhenius equation was applied to thegrowth rate, m (Siegrist et al., 2002; Sinechal et al., 1979; Srisertpolet al., 2010), the maximum growth rate, mmax (Angelidaki andAhring, 1993; Hashimoto, 1983), the saturation constant,Ks (Dague et al., 1998; Siegrist et al., 2002), the hydrolysis rate, K, thedeath rate, Kd (Donoso-Bravo et al., 2009; McKinney, 1963; Siegristet al., 2002; Veeken and Hamelers, 1999), the inhibition constants,Ki (Siegrist et al., 2002), and the yield coefficient from substrate tobiomass, Y (McKinney, 1963). However in some cases, the SquareRoot equation provided a better temperature prediction than theArrhenius.With granular sludge adapted to 10 C over 235 days, thetemperature dependence for the methanogenic conversion of ac-etate was well described by the Ar...</p></li></ul>