PD Arty he ®Wicas Wb al bd hike PPh ar rer SS Per cea pe cores Per Ba et Prete 4 ) . ans . _ \- eos x C3 Letty TST PS 0 } Ls La " ry Ch ry tat ry eo ry td 2 ks are Pe TL . ne ar eaten lke chanical ary HERR OItC: ESTI jen Mi aa Dic on ) | Mopen Non ba a Ce iy ere ile Crh Cre ee eri Pra Nears Ih Ti i ra ee ASTEWATE i Soares ar bs mall Prh Bie report , ahd Pea) naitie ©° ere D Pett ai sa _ en Pe é Pahl . me or i Ro, SAS aTS Pert hiis ba Vise . A eae Seats r ry a Aah) rr hha = a rechatcal a rep CO hha hae - Pe vs rae PPT Lae > pechnical 7 - ee ae PPP Lae ; PPT at Ks ed Pera ° y, PP aa 7 res a Phas pacwareet ta Mts Cn Cry a ee PF hid hdd PTT 7 a PPC ols cd BG = ITY Par Leda Frc a. ee re CO Pee ina Pi A. heeContents Acknowledgements vii List of Task Group members ie Preface xi Summary 1 1-_Introductioa to the ADM 3 1.1 The importance of anaerobie digestion and a generie process model 3 1.2 Conversion processes in anaerobic digestion 4 2_Nomenclature, state variables and expressions 6 2.1 Units: 6 2.2. Nomenelature and description of parameters and variables 7 23. Dynamic state variables 8 3. Biochemical processes 9 43.1 Stmcture of hiochemical reactions in the ADML 9 3.2. Rate equation matrix. 12 3.3. Disintegration and hydrolysis B 3.3.1 Kinetics of disintegration and hydrolysis 4 3.4__ Mixed product acidogenesis, it 3.4.1_Acidogenesis from monosaccharides 15 3.4.2 Acidogenesis from amino acids 0 3.5. Syntrophic hydrogen-proxiucing acidogenesis and hydrogen-utilsing methanogenesis 18 3.5.1 Form of electron carrier 20 3.5.2. Biological groups and components in the ADM 2 3.5.3. Hydrogen inhibition functions for aeetogenesis 21 ‘Acetclastic methanogenesis 23 36wi Contents 3.7 Inhibition and toxicity 3.7.1 Modelling of inhibition 3.8 Influence of temperature 3.8.1 Modelling of temperature effect on disintegration and hydrolysis, 3.8.2_ Effect of temperature on thermodynamic yields and reaction pathways 31 4_ Physivo-chemical processes 3 4.1 Liquid-iquid processes 3 41 Modelling of acid-base reactions 35 4.2. Liquid-gas transfer 37 42.1 Liquid-gas transfer equations 37 4.3. Variation of physico-chemical parameters with temperature 38 5__Model implementation in a single stage CSTR 40 '5.1_Liguid phase equations 4 5.2_Gas phase equations a2 53. Specific example: inorganic carbon 4B S31 DAE system 4B 53.2 DE system 4 6 Suggested biochemical parameter values, sensitivity and estimation 45 61 Hydrolysis parameters 46 6.1.1 Parameters associated with propionate 48 61.2 Paramet ociated with acetate 48 7 Conclusion 49 References 31 Appendix A: A review ef parameters 56 Appendix B: Supplementary matrix information. 63 Appendix C: Integration with the ASM 68 Appendix. D: Estimating stoichiometric coeffictemts for fermemation 72 Index 751 Introduction 1.1 IMPORTANCE OF ANAEROBIC DIGESTION AND A GENERIC PROCESS MODEL Anaerobic conversions are among the oldest biological process technologies utilised by ‘mankind, initially mainly for food and beverage production. They have been applied and developed over many centuries, although the most dramatic advances have been achieved in the last few decades with the introduetion of various forms of high-rate treatment processes, particularly for industrial wastewater. High organic loading rates and low sludge production are among the many advantages anaerobic processes exhibit over other biological unit operations. But the one feature emerging as a major driver for the increased application of anaerabic processes is the energy production. Not only does this technology have a positive net energy production but the biogas produced ean also replace fossil fuel sources and therefore has # direct positive effect fon the greenhouse gas reduction. This will most certainly ensure the ongoing, and likely drastically increased, popularity of anaerobie digestion processes for waste treatment in the future But why is there a need for a generic model? Several benefits are expected from the production ofthis first generalised model of anacrobie digestion: ‘= increased model application for full-scale plant design, operation and optimization: ‘© further development work on process optimization and control, aimed at direct implementation in full-scale plants; ‘+ common basis for further model development and validation studies to make fouteomes more comparable and compatible assisting technology transfer from research to industry. Many of the above points relate to practical, industrial applications, Indeed, this is one of the areas whete most henefits from the application of a generie process made! ean be gained, While many different anaerobie models have been devised over the years (and indeed for, the basis of the ADMI), their use by engineers, process technology providers and operators has been very limited. Two of the limiting factors have likely been the wide variety of ‘models available and often their very specific nature. STR Draft 18/09/01 0The model presented in this report tries to be as widely applicable as possible for anaerobic processes ~ and therefore will naturally not be as accurate as some specific models developed for certain applications. And it also has been limited to the main relevant processes occurring in order to make it more simple and applicable, This again meant that ‘many known and sometimes relevant aspects have not been included in thi First version of the model. Some of these are highlighted in this report in the form of inserts, which briefly discuss the nature of the excluded processes and may also provide some conceptual ‘modelling approach, ‘We hope that this model will help to achieve a widespread utilisation of the large body of knowledge in anaerobie processes available from research studies and operational experience. This ultimately will support the increased application of the anaerobie {technology as one of the most sustainable waste treatment options in the Future and a viable alternative to other energy generation processes. 1.2 CONVERSION PROCESSES IN ANAEROBIC DIGESTION. Conversion processes in anaerobic digestion can be divided into two main types (Figure 1.1) (a) Biochemical reactions: These are normally catalysed by intra or extracellular enzymes and act on the pool of available biological substrate, Disintegration of composites (such as dead biomass) to particulate constituents and the subsequent enzymatic hydrolysis of these to their soluble monomers are extracellular. Digestion of soluble materials are mediated by organisms intracellularly, Biomass growth and decay isa result of this process. (b) Physicochemical reactions: These are not biologically mediated and encompass ion associationfdissociation, and gas-liguid transfer. Precipitation would be a further physicochemical process, however this is not included in the model Distinguishing between available degradable (substrate) and total input COD important, as a considerable fraction of the input COD is anaerobically not biodegradable Gossett and Belser, 1982). "The ultimate biodegradability factor (D) is one of the most ‘important characteristics of the influent COD, as it fundamentally influences all steps and COD flux, and an influent with COD D = I, or totally degradable substrate is generally rare. In general, we use the term ‘substrate’ to indicate degradable COD, while the inert fraction (1-D) is represented by the soluble (S,) and particulate (X)) inert Biochemical equations are the core of any model and it is possible to represent an anaerobic system using only these, However, o deseribe the effect on biochemical reactions fof the physicochemical state (such as pH and gas concentrations), physicochemical ‘conversions must be included as well, ‘The COD flow chat as used in this model is shown in Figure 1.2. This shows the COD. flow through intermediates for a hypothetical composite particulate material that is 10% inerts and the remainder split equally between carbohydrates, proteins, and fats. The COD flux would change considerably for different primary components, of for different organic acids from monosaccharides and amino acid. STR Draft 18/09/01 PBoO omen > Figure 11: Conversion processes in anaerobic digestion as used for the model. Biochemical reactions are implemented as ireversible, while physicochemical reactions are implemented a reversible. Abbreviations inckide MS, monosaccharides, AA: amino acids, LCFA: long chain fatty acids, LFA’ CFA base equivalent, HVa: valerio acid, Va‘ valerate, HBus butyric ‘acid, Bu’: butyrate, HPr: propionic acd, Pr: propionate, HAc: acetic acid, Ac: acetate. SomporiteParicaate Material (100% Disintegration ws Tne 104] Catena 09] Protas 35 Fa 09 Aydasie “ LGEA a] Acidogeess Acetogensis Acetic «a5] we Figure 1.2: COD flux for a particulate composite comprised of 10% inerts, and 20% each proteins, carbohydrates and fats (in terms of COD). Propionic acid (10%), butyric acd (12%) and valerie acid (7%) are grouped in the diagram for simplicity. Abbreviations as for Figure 1.1 STR Draft 18/09/01 B2 Nomenclature, State Variables and Expressions ‘The IWA Anaerobic Digestion Model (ADM) introduces generic nomenclature, units and definitions. This section summarises all elements and serves as a reference for terms used throughout the report, For biomass stoichiometry, a formula of C;H;O3N is used as in the ASM series (Henze et al, 1986). 2.4 UNITS ‘The units 1 be used were the subject of extensive discussion and a community survey. COD {&3 COD my"=z COD‘) was ehosen asthe chemical component base unit because ofits we in chemistry texts (and thus agreement with pH, pK, etc), as a wastewater characterisation measure in concentrated sircams, its use in upstream and gas utilisation industries, the implicit balancing of carbon oxidation state and to enable partial compatibility withthe [WA Activated Sludge Models (Henze eta, 1986). Molar bass (kg-mole:m? = M) is used for components with no COD such as inorganic earbon (CO, and HCO," and inorganic nitrogen (NIL and NI) ‘The use of ke COD" is notin agreement with the ASM models and general practice in aerobic catment, where g COD: (mg COD") is commonly used, However, implementing in mg COD!" is relatively simple, as it requires only changes in Ks values, and modification of pK, and K, values, and we encourage the use of g COD-an" (img) if required (.. as an add-in to aerobic models). Use ofthe model in g CODim (mg COD") and integration with the ASM models is specially adaressedin Appendix B. Table 2.1: Units Measure Tis ‘Concentration Kg COD AT Goncentration (nor-COD) kgrmole Gm? Concentration (nitogen nor-COD) kgemole Nan Prosaure bar Temperature K Distance mn, Volume. ~ Energy she) Time é STR Draft 18/09/01 io2.2 NOMENCLATURE AND DESCRIPTION OF PARAMETERS AND VARIABLES There are four main types of pas equilibrium constants, kinetic parameters, and dynamic or algebraic variables, and calculated indicators ofthese. meters and variables; stoichiometric coefficients, Table 2.2: Stoichiometic coefficients ‘Symbol Deseripton Units c ‘arbon antent of component Kgrmole C kg COD N nitrogen content of component i kg-mole Nkg COD wi rate coefficients for component ion process nominally kg COD-m* Isaciewnse yield (catabolism oniy) of product on substrate kg CODKg COD" Table 2.3: Equilibrium constants Desorption Tas (g88 law constant (equal to Ky") ‘bar M ‘acid-base equilbria constant M (kg-molem?) Henry's law coefficient ‘bar logrlKal gas law constant (8.214e-2) bart As free energy ‘Simole” Table 2.4: Kinetlo Constants and Calculated Rates Symbol Deserpion This Ia aid Rase kinetic consiant q (Ma orkg CODm' a) hs frstorder decay rate « Ie mee Inhibition function (686 K) Koco first order constant (normaly fr hyerolysis) “ Ka (g20-lqud transtercoetfciont a Kr te 50% inhibitory concentration kgcODm? Ko peor Monod maximum specticupake rate (im) kg COD_Skg COD,X"-s" Keane haf saturation constant kg COD Sm* B generalised rate of process j Vosenn Yield of Biomass on Substrate kg COD_Xxg co s" nse Mona maximum specific grouth rate ¢ Table 2.5: Dynamic and Algebraic Variables (and calculated indicators of variables) Symbol tts pH Pras bar Pe total gas pressure bar s) soluble component kg CODm*? tox ‘extended retention of solids 4 T temperature K, v volume m™ x particulate component kg CODm* STR Draft 18/09/01 15]2.3 DYNAMIC STATE VARIABLES For biomass stoichiometry, formula of CsH,O:N is used as in the ASM series (Henze et al. 1986) Table 2.8: Dynamic state variable characteristics (DAE system) complex Varies varies carbohydrates cosa 0 protains® varies vaties lipids 00220 "0 particulate inert varies varies varies varios soluble inerts varies varies varies varies monosaccharides 78018200313 ‘amino acids” varios varies varie vatiee ‘otal LOFA® ‘258 795 O0ZI7T 0 total valerate 40220800280 total butyrate 88 ye 00250 oo total Propionate 74 12 00288 =o total Acotate 60 6 = 00313 2 16 0 ° 16 Ca Inorganic carbon M48 ° NA 1 Inorganic nirogen = M17 ° a o biomass 113184 0.0272 0.00844 cations Mo varies 0 0 ° anions uM o o ° ‘Siang 32 i See Taies Stand32——— 2, Unless otherwise stated, kg COD-m® 3, See Appendix C 4. Based on palmitic triglyceride and palmitate STR Draft 18/09/01 (6)3 Biochemical Processes 3.1 OVERALL MODEL CONCEPT Most recent anaerobic digestion models include intermediate produets and the skgroup agreed on a structured model because of a number of seientifie and application advantages. The philosophy of process and component inclusion was to maximise applicability while ‘maintaining a reasonably simple structure. Reasons for including specific processes are explained under sub-headings, The model includes the three overall biological (cellular) steps (acidogenesis fermentation}, acetogenesis (anaerobic oxidation of organic acids) and :methanogenesis) as well as an extracellular (partly non-biological) disintegration step and an extracellular hydrolysis step (Figure 3.1). Three of the processes (hydrolysis, acidogenesis, and acetogenesis) have a number of parallel reactions. Complex particulate waste is assumed a homogeneous mass, which disintegrates to carbohydrate, protein and lipid particulate substrate. This was mainly included to facilitate ‘modelling of activated sludge digestion, as a disintegration step is thought to precede more complex hydrolytic steps (Pavlostahis and Gossett, 1988), but is also generally used when the primary substrate can be represented with lumped rate and biodegradability constants (eg. primary sludge, others in Appendix A). The complex particulate pool is also used as a pre Iysis repository of dead biomass. ‘Therefore the disintegration step is intended to include an array of steps such as lysis, non-enzymatic decay, phase separation, and physical breakdown (e.g, shearing). Al extracellular steps were assumed first onder, whieh is an empirical Funetion reflecting the cumulative effect of a multi-step process (Eastman and Ferguson, 1981). Cellular kineties are described by three expressions (uptake, growth, decay, Tables 3.1 and 3.2). The ey rate equation is substrate uptake, which is based on substrate level Monod type kinetics. We chose substrate uptake related kinetics (rather than growth related kinetics) to decouple rowth from uptake, More reasons for this are given in the inhibition seetion (3.7). ‘The basic kinetics used here could also be termed Michaelis-Menten, but this is not a term generally used fo) alysis, and as for Specce (1996), we use the teem Monod:type. Biomass growth is implicit in substrate uptake, as itis Hnked to growth rate. First order biomass decay (to complex organics) was assumed and is an independent set of expressions. 3.2 RATE EQUATION MATRICES he rate equations are shown in Tables 3.1 and 3.2. ‘These are for non-physicochemical equations and do not include transfer tothe gas phase. All acid-base pairs, including organic STR Draft 18/09/01 macids, are repy Sie = Seor+Scos and 'Si25ac+Siae): Mote information about modelling of the physicochemical equations, and the ‘option of splitting these p fen in Section 4, Where fitted or caleulated yields are used, they are referred t0 8 fui, atsuses COD balancing is implicit inthe rate equations. In many cases, inorganic carbon may be the earbon source or product and snted as the sum of the acid/oase pair (e fas dynamic state variables i gi for catabolism or a ta hydrogen; j=5.6,10,11,12), and in these eases, we recommend expressing the inorganie carbon rate coefficient asa carbon balan aholism (i.e. uptake of sugars, amino a Ying=— PMs en For example, Vio the inorganic carbon coefficient for amino acid fermentation is Veaet-Cact Yan) Sonan #U-Yon) Sina Coo IVS pas'Cpo HY Pose Cae FYuCo 82) where C, isthe carbon content (ky mole Ckg COD") of component i, and Chow is the zeneral carbon content of biomass (0.0272 mole C-z COD"). In other processes, (aptake of LCFA, butyrate valerate, hydrolysis, disintegration, deea: j=1-4.7.89,13-19), we decided not to inclide thi term. “There may be an error in the carbon balances ofthese processes because of the different carbon contents of substrate, product and biomass, and if balancing of carbon is critical, expression 3.1 should be used forthe coefficient Vi, forall processes t Figure 3.1: The anaerobic madel_as implemented including biological processes (1) ‘Acidogenasis from sugars, (2) Acidogenesis from amino acids, (8) Acetogenesis trom LOFA, (4) Acotagenesis from propionate, (5) Acetogenesis from butyrate and valerate, (8) Acoliclastic methanogenesis, and (7) Hydragenotraphic methanogenesis. STR Draft 18/09/01 1s]“Tele Vict (and ae enn rable comport a2) om ee ef PL EE BE eminent ‘ baogecn + Jeo Thysopn Crt 2 Fraoyas Petre i Shpssyse supe Ty Ty FUpuiectougn | 10 fa] Np OA fan ace — |= ESP | Me Lnmmcianmonccs | [+ | [aac melt ra terroinn|tte faltartee) | aBEt| SR Uae tate a vvonst ovae3t| caves Hohe bet aye 1 coxaan | orn ame nt Papa + laxgam gas Ne nant ee 1 yo, Eeh= mame {2 et yam 1 axel “nom Dba oie ya Dey a ay Pe ye = ‘poeaystxe ew lel ale il tle i 3 7 i i 3 |3 eT Lee a“ale 12 Getpen t otrta yan cnprensn pr pause component 24-18) comport > 1 Bw we we | we v s mo Phe i aren ee Tou Ipsos Comeraiee ni ss ealupae zt 3 hpi af Pots ET 5 Ute at s % 6 Unset are tee Yo 7 tect. ve 8 Une! Vara Ye © tect aerate Ye {0vpake ot Prepon Ye ote ot onto Ye Seon mts reonat Noa arse tren team Eco cen LEP gms krcoba trcoaa esac reno gen d aici STR Da 8501 or3.3 DISINTEGRATION AND HYDROLYSIS, Disintegration and hydrolysis are extracellular biological and non-biologieal processes mediating the breakdown and solubilisation of complex organic material to soluble substrates. The substrates are complex composite organics and particulate carbohydrates, proteins and fats. ‘The last three substrates are also products from disintegration of the first substrate. Other products of disintegration are inert particulate and inert soluble material ‘The products from (enzymatic) degradation of carbohydrates, proteins and fats are monosaccharides, amino acids and long chain fatty aeids, respectively [A mainly non-biological disintegration step was ineluded as the first process 10 allow diversity of application, and to allow for lysis of biological sludge and complex organie ‘material, The three parallel enzymatic steps were included to account for the difference in hydrolysis rates ofthe three well defined particulate substrates ‘The disintegration step was also included to represent the pool of composite organie ‘material. This is especially important for activated and primary sludge digestion, where the disintegration step represents lysis of whole cells and separation of composites. Vavilin and co-workers (eg.. Vavilin etal, 1999), Pavlostatis and Gossett (1988) and O'Rourke (1968) have used this approach. Inclusion of a composite organic material also allows an elegant ‘method for recycling of dead anaerobic biomass that is more conceptually and numerically correct than recycling decayed biomass to the immediate substrate (e.g., Costello et al, 1991), ‘The term hydrolysis is used here for the degradation of a defined particulate or macromolecular substrate to its soluble monomers, ‘The most significant particulate substrates identified were carbohydrates, proteins and fats, and for these substrates, the polymerisation process matches the formal chemical definition of hydrolysis. In each case, the process is catalysed by enzymes, which are likely to be produced by the organism directly benefiting from the soluble products. Hydeolysis ean be represented by one of two conceptual models: (a) The organisms secrete enzymes to the bulk liquid where they adsorb onto a particle fr react with a soluble substrate, (b) The organisms attach to a particle, produce enzymes in the vicinity of the particle and benefit from soluble products released by the enzymatic reaction ‘The taskgroup agreed that in anaerobic mixed culture systems the dominant mechanisms Found was (b) as shown by Vavilin etal. (1996) and Sanders et af. (2000). Therefore, the forganisms growing on the particle surface, rather than the enzyme produced, should be rogarded as the effective catalyst. Kinetics of Disintegration and Hydrolysis All iterature models utilising a disintegration term (as opposed to a hydrolysis term) have used first order kinetics. This is reasonable, as first order kinetics have been supported by observations, and because the diversity of disintegration processes cannot support a different, more fundamental approach ‘The complete enzymatic hydrolysis step is a complex. multi-step process. for carbohydrates, proteins and fas, which may include multiple enzyme production, diffusion, adsorption, reaction and enzyme deactivation steps. However, the most commonly used Kinetic relationship © describe hydrolysis processes Is frst order and “is an empiri expression that reflects the cumulative effet ofall the microscopic processes occurring (Eastman and Ferguson, 1981). Surface related hydrolysis kineties have been based on STR Draft 18/09/01 nuenzyme production or adsorption (Batstone et a, 2000; Jain er a, 1991), or surface related biomass growth (Vavilin et al, 1996). Walker and Wilson (1991), Negti et al. (1993), and Sanders et al. (2000) have used models to demonstrate the importance of surface based ineties more empirically However, Vavilin et al. (1996) compared a number of hydrolysis Kinetics including a fundamental two-phase model. A first order mode! was only slightly poorer than the more complex two-phase model. A model with Contois kineties (which use a single parameter to represent saturation of both substrate and biomass) was as good at fitting the data as the two- phase model. Valentini er af. (1997) quantitatively assessed the influence of biomass concentration in a first order model, with an exponent of between (and | affecting the ‘biomass concentration, finding that the exponent had a best fit between 0.4 and 0.6 (bateh tests). An exponent of 0 (ie., biomass independent, first order substrate based) was almost fs effective as the optimal exponent with a standard deviation of 35% compared to an optimum of 22%. Batstone (2000) also showed that a fist order model could fit biogas production as well as a complex two-phase model (which included enzyme adsorption). ‘Therefore the taskgroup recommends that first order Kineties be used by default, Contois kinetics could be used in systems where biomass concentrations are low enough to be rate Timiting (e.g, batch digestion; sce Vavilin er a, 1996 for form of equation) 3.4 MIXED PRODUCT FERMENTATION/ACIDOGENESIS Acidogenesis (fermentation) includes the degradation of soluble sugars and amino acids toa number of simpler products. The degradation of LCFA is an oxidation reaction with an extemal electron acceptor and is therefore included in the next section. Unlike anaerobie oxidation, acidogenesis can occur without an external electron acceptor. This, and better energy gains allow the reactions to occur at high hydrogen or formate concentrations. Yields are also generally higher for monosaccharides of amino acid acidogenesis than for the other biological processes. Acidogenesis from Monosaccharides ‘The taskgroup decided to use glucose (hexoses) as the model monomer. Fructose is energetically and stoichiometrically equivalent for modelling purposes, and pentoses will have similar stoichiometric reactions compared to hexoses, with one less CO; or carboxylic ‘acid unit in the products. ‘The most important products and their stoichiomettie reaction Irom, alucose with approximate ATP yields (except reaction i) are ranked in order of importance in Table 3. These acids can also be produced in combination for mixed acid products Table 3.4: Products fram glucose degradation. Produce ion ‘ATP Condiiane Mole Giu () Acetate GeHeOn2HOP20H.COOHACONAH: 4 fowls (i) Propionate OsHsOst2H 20H:CH.COOHI2HO “low —notobooned 2 (iy -Avaate. 8CsH206 ‘ any Propionale 4CH:CH-COOHY2CH:COOH:20042H:0 He (ii) Buyrate— CHuOrPCHCH:CH:COOH2COsr2H: 3 lowe 1 (iv) Lactate (HaOs> 20HsCHOHOOOH 2 any He (Ethane GHi20, 20H OH,OH4260) 2 low pH * Noles: 1. While thermodynamically possible at high He, may be limited by energetics of substate- level phosphorylation (Schink, 2001). STR Draft 18/09/01 ites}2. Not yet observed in cultured environmental samples. Coupling with substrate level ‘Oxidation is more common ae in action 4, Energy yield taken from yeast pathway. Bacterial pathway may have 0 ATP/mole ethanol (Madigan etal, 2000) ALTERNATIVE PRODUCTS FROM FERMENTATION OF GLUCOSE Glucose can form a number of alternative fermentation products apart from organic acids (Madigan et al,, 2000), the most important of which in anaerobic digesters are lactate and ethanol. Lactate is a key intermediate, and dynamic work has indicated that most or all of the monosaccharide substrate may degrade via lactate (Skiadas ct al, 2000, Romli et al, 1995) However, lactate is subsequently degraded very quickly and is therefore seen primarily during transient overload conditions in acidification reactors (e.g., Figure). As demonstrated in the figure, during the concentration overload, the lactate increases from being insignificant to the highest organic acid (in terms of COD). Ethanol is produced as an alternative to acetate at low pH (pH5.0; Ren etal., 1997) Lactate has the same stoichiometry as glucose, and therefore, biological reaction stoichiometry is not affected by its omission from the ADMI. However, lactic acid has a relatively low pKa (3.08), which has a strong effect on pH values. In particular, the ADM will under-predict transient decreases in pH (averpredict pH). This effect is increased for hydraulic increases as compared to concentration increases. The lack of ethanol as an intermediate will cause poor prediction of intermediate organic acids, and pH at low pH levels in acidification reactors. Methanogenic reactors, and low-loaded systems will be largely unaffected by the commision of either lactate or ethanol, as lactate and ethanol aze relatively easily degraded to mixed organic acids and acetate, respectively. The relatively low concentrations of these intermediates in most anaerobic digesters was the main reason for their omissions. In general, it woud he desirable to include them in bi ily loaded acidogenic glucose-fed systems, sith transient concentration and hydraulic conditions (lactate), or when operated at low pH, of deliberately to promote ethanol production, for example to enhance downstream digestion (ethanol, Ren ct al, 1997) Lactate hss been implemented as an intermediate by Costello et al. (1991), Romi et al (1995), and Skiadas etal. (2000). ‘The simplest method is similar to the last, and assumes that all glucose degrades via lactate, whieh is subsequently degraded to mixed organic acids by either glucose degrading bacteria or by a dedicated group. References to models including degradation of glucose o ethanol have not been found, and a pH regulation function is probably necessary to describe the dependence of product yields on pH. q me Figure for Above Insert: Organic acids during a 2x concentration step in a glucose-fed agideeenicetaaatssditom Romi etal, 1995) 113)Reaction (ii), whieh is the uncoupled reaction of glucose to propionate has appeared in several models (Costello et al, 191; Romli er a., 1995; Skiadas eta, 2000). These models are consistent with observations, as the parameters associated with hydrogen regulation do ‘not allow production of propionate only. However the taskgroup recommends that reaction (i) should be used in preference of reaction (ii) for the following reasons: (1). No organism producing propionate only has been cultured, All organisms producing. propionate or succinate (the key intermediate prior to propionate) also produce Acetate with COs as byproduct (Madigan etal, 2000; Gottschalk, 1986) 2) Sourcing electrons from (ie., oxidising) formate or elemental hydrogen is thermodynamically unfavourable except ata high H, partial pressure and is therefore Inconsistent with the release of formate or hydrogen by organisms fermenting slucose to butyrate or acetate, ‘The tskgroup decided to include acetate, propionate and butyrate i the model as they are important end-prodets from glucose acidogenesis, are degraded differently downstream, and ate measured simultaneously by GC analysis. Lactate is very important as an intermediate and modelling work has shown that all glucose may degrade via lactate (Romli e al. 1995; Skiadas eal, 2000), However, lactate is soichiometrically the same as glucose (though not physicochemically) and the kinetics of lactale fermentation to VAS are relatively fast (Romi era, 1995). Therefore the taskgroup has decided not to include lactate inthis model and the Timitations eaused by this assumption ae addressed in an insext. Ethanol is an important end-product at low pH (Ren et a, 1992) but is produced at lower concentrations a the pH lovels normal in digesters (5-8). Its therefore also omitted from the current mode! ‘As many organisms are capable of producing several products, a single group of organisms with the same uptake rate constant should he used. Regulation functions to deseribe the various fractions of produets from glucose under different Hs and pll levels have been described by Mosey (1983) and futher developed hy Costello er a. (1991) and Romi ea. (1995). However, these re described using rections i) and (iv), e0uld not be tscd consistently with a varity of experimental data scts, and require the inclusion of lactate. Therefore, no hydeogen regulation function is wed in the ADM and stoichiometie Yields sw Sse Fran Shan) fe Sel © constants consistent with the equations in Table 3.4 Gee Appendix C). "Fixed stoichiometric yields were used by Skiadas er al, (2010) and ngelidaki etal. (1999) Acidogenesis from Amino Acids ‘There are 20 common amino acids (Appendix C). Products from protein hydrolysis are ‘dependent on protein primary structure (FAO, UN, 1970), There are two main pathways for amino acid fermentation: (1) Stickland oxidation-reduction paired fermentation. (2) Oxidation of a single amino acid with hydrogen ions oF carbon-dioxide as external electron acceptor Stickland reactions occur more rapidly than uncoupled degradation (Barker, 1981) and in ‘normal mixed:-protein systems, there is normally only a 10% shortfall in electron acceptor STR Draft 18/09/01 ayproteins (Nagase and Matsuo, 1982). There ate a number of characteristics of Stickland Termentation oF amino acids (Figute 3.2) (2) Different amino acids ean act as donors, acceptors, or both (Appendix C). (2) Electron donor loses one earbon atom to CO; and forms carboxylic acid with one cearbon shorter than original amino acid (ic, alanine; C3 - acetate; C2) (8) Fleetron acceptor retains carbon atoms t6 form carbox length asthe original amino acid (ie. glycine, C2 acetate, C2) (4) Only histidine cannot be degraded via Stickland reactions, (5) Typically around 10% of total amino acids are degraded by uncoupled oxidation because of a shortfall in electron acceptors and this results in hydrogen oF formate acid with same chain production Onidation Reduction Aline (Domo rye ces “xo v NABH Pyrat NH 5 ia con SD < a cox) wit AceglCon é vy Coa Ace Phpate App yo > ATP \ Actc Ad Ate Ac Amines ADC&R—> Aerie + ATP +00, +10 +a 2Gpeie > 2A 42504 Figure 3.2: Coupled SticKland digestion of Alanine and Glycine (From Madigan ofa, 2000). This makes it relatively easy to predict products of reactions, which are largely C2, C €5. C6 iso and normal organic acids with some aromatics, COs, Hy, NH, and reduced sulphur. Aromatic amino acids (Phe, Tyr, Trp) produce aromatic carboxylic acids. Since these amino acids are in relatively low numbers, only small amounts of the aromatic acids are generated. Ramsay (1997) built a spreadsheet of yields from amino acids to estimate the yields from the amino acid content of a protein substrate (Appendix C). Non-stickland ‘oxidation of amino acids may occur with low hydrogen or formate concentrations, or under thermophilic conditions, when oxidative reactions become more thermodynamically {avourable (Ref from Stams), and oxidation reactions generally yield greater propionate and less acetate and butyrate (in direct contrast to fermentation of glucose). However, the use of f Stickland based spreadsheet is @ reasonable initial estimate of product yields. Because Stickland reactions are generally not inhibited by hydrogen, hydrogen regulation or inhibition functions have been excluded, STR Draft 18/09/01 ls)3.5 SYNTROPHIC OBLIGATE HYDROGEN OR FORMATE PRODUCING ACETOGENESIS AND HYDROGEN OR FORMATE UTILISING METHANOGENESIS Degradation of higher organic acids to acetate is an oxidation step, with no internal electron acceptor. Therefore the organisms oxidising the organic seid (normally Bacteria) are required to utilise an external clectron acceptor such as hydragen ions or earbon-dioxide to produce hydrogen gas or formate respectively. These electron acceptors must be maintained at alow concentration for the oxidation reaction to be thermodynamical possible, (Figures 3.3 and 3.4) and hydrogen and formate are consumed by methanogenie organisms (normally Archaea) ‘The thermodynamics of the two reactions are only possible ina narrow range of hydrogen or formate concentrations. This is illustrated in Figure 3.3, which shows the thermodynamic yield (AG) for methanogenesis and three anaerobic oxidation reactions. ‘The shaded region indicates where methanogenesis and propionate oxidation are simultaneously possible Figure 3.4 shows the operating space in another way, by marking lines of zero AG! for varying acetate and hydrogen concentrations. ‘The shaded region shows the space in which the five reactions are theoretical simultaneously possible, Also shown is the measured threshold for methanogenesis (Cord-Ruwisch, 1988). This would more than halve the available operating space and if the other reactions had similar limitations, a very narrow region of acetate and hydrogen levels would he availabe. Because of the small operating space, and the solubility limitation of hydrogen, syntrophie (e.g. hydrogen producing and consuming) organisms are often closely located and quite istinctive (Harmsen er af., 1996). Because of the size of the syntrophic regions within the anaerobic 0% ortia, diffusion may be rate-limiting and feral substrate and. produet ‘concentrations ean be diferent to bulk concentrations. ‘Table 3.5: Thermodynamics of reactions for fatty acid oxidising orgariems, “subevaie ~——~—SCSC~S~S~S*«*SR aon SSC (ki/g COD) _(ks’g COD) Hi, HOO ‘Het 00,9 CHar BHO 212 0.19 Propionate _ CHsCH:COOH#2H:0-> CH:COOH+3H:+CO; 0.88 0.13, Butyrate CH:CHaCHeCOOH + 2H:0 > 2CHsCOOH+2H» 0.80 018 Palmitate __CHs(CHs] COOH + 14H:0 3 BCHsCOOH+14He 0.55 0.16 ‘AG Calculaled for pH of 7, pH, of Te-Sbar, pGH, of 0.7 bar, OTM HCO; and TmMl organic ‘acid concentrations STR Draft 18/09/01 6)SULFATE REDUCTION AND SULFIDE INHIBITION When oxidised sulfur compounds are present in anaerobic digesters, they will generally be reduced to S*, This ix hecause the oxidised sulfur is reduced in thermodynamic and Kinetic preference to hydrogen ions (10H) oF CO» (to formate). Organisms reducing sulfur ean ‘busin the eleetons dieety by oxidising orzanic acid, or by Oxidising the Hs produced by ctogens. Additionally, organic acids are used as a carbon soutes, and as a result organisms reducing sulfur compete with majority of groups in anaerobic digestion at ‘various levels of influent oxidised sulfur including (a) Hydrogenotrphie organisms for hydrogen (at low levels of infuent SO.) (b) Acetogenie and aveticlasic organisms for electrons and carbon (at medium levels of influent S09). Further complicating the effect on anaerobic systems, the reduced product, sulfide, is inhibitory at 0.003-0.006 M total $, of which the fully associated Form (HS) is the inhibitory agent, at levels of 0.002-0.003 M HLS (Speece, 1996). Tn terms of most affected ed, but also all other groups including sulfate reducing organisms, except perhaps acidogenic organisms. Reduced sulfide has similar acid-base system to the inorganic carbon system, with S®, HS, and HbS as components. Hs is also gas phase component, witha relatively high solubility (0.1 M-bar"), Solubility and acidity constants are stongly affected by temperature (Specce, 1996), and the relationships are well described by the van't Hoff equation. All anaerobic processes simulated by the ADM 1, both biological and physicochemical ‘except perhaps disintegration hydrolysis are affected y either competition for substrate inhibition by HLS, or the acid-base system and gus-ligid transfer of the sulfide product. Because ofits complexity, the sulfate reduction system was not included inthe first version of the ADMI. The ADMI is therefore incapable of modelling systems with low-medium mounts of sulfide (20.002 M influent SO.) The simplest method of modifying the model to incorporate sulfate reduction (at <0.02 M influent $0) is wo include an extra group of organisms degrading oxidised sulfur to reduced sulfides, with electrons and hydrogen sourced from hydrogen, and carbon for growth sourced from acetate. The acid-base pair TS7HLS should also be included, with wansfer wo the gas phase of HsS. AL higher influent oxidised sulfur concentrations, more complicated models, with different sulfate reducing stoups. describing competition for organic acids must be included (Kalyuzhnyi and Fedorovic, 1998). to least affected aceticlasti, hydrogenotrophic and acetogenie organisms are inhil STR Draft 18/09/01 17Jog HCOOH (9 7 27 a7 040 020 AG" at pH 7.0, 25¢ (kyg COD sunstratey ono 2 3 ‘ 5 6 -tog pil, (bar) Fig lure 9.3: AG’ for the reactions shown in Table 3.5 at diffrent hyerogen partial pressures (bottom ° axis) and formate concentrations (top "x axis). Apart from hycrogenvformate, concentrations are 0.1M HCO,’, and tml organic acids at pH 7. The shaded region shows the theoretical operating region for syntrophic acetogenesis of propionate, Valerate is thermodynamically similar to butyrate. AG values taken from Madigan (2000) and AG” calculated from AG’=4G"-RT In (CIIDI" ) in the reaction: a A +b Bere C +d D. rer STR Draft 18/09/01 lus)Lines of constant 4 G'=0 Palmitate > Hact1 4H: HBu— HacezH [vas cm 5 6 4Hi+CO; > OH, ‘4He+CO2 > CHe ens) 7 70 60 50 40 3.0 20 “Jog acetate (M) Figure 3.4: Lines of zero AG for tive reactions with similar assumptions to those in Figure 3.3, caption (except acetic acid concentration), Shaded portion shows regions where all reactions are possible. The boxed hydrogen utllsing reaction and corresponding line is based on ‘measurements as deseribed by Cord:-Ruwlech (1988). Form of electron carrier ‘The electron carrier (acceptor product) can be either hydrogen (Irom hydrogen ions) or formate (from carbon-dlioxide). There are three major differences between the two forms (H, + CO, ¢ HCOOH): (0). Hydrogen has a higher diffusivity, 2) Format (3) Formic aed is stronger acid than earbon-dioxide is more soluble. ‘Therefore, when interspecies distances are short, hydrogen transfer will be faster and when distances are long, the greater solubility of formate allows a greater concentration gradient and therefore better transfer. Additionally, formic avid has a different influence on the physicochemical system duc to the lower pK, compared to CO;, Apart from this, model ‘implementation is largely unaffected as stoichiometry and thermodynamics are virwally identical and hydrogen/formate may be in enzyme assisted equilibrium (Thiele and Zeikus, 1988). Also acetogens may waste electrons as either hydrogen or formate and hydrogen utilising methanogens can accept either (Boone ef al, 1993). ‘The taskgroup has therefore {decided to implement the electron carrier as hydrogen only and to not include formate STR Draft 18/09/01 119}NITRATE REDUCTION Dissimilatory nitrate reduction, a form of anaerobic respiration, isthe reduetion of NO," to nitrogen oxides ~ such as NO,, NO, NO ~ and N,, Because the production of gaseous nitrogen compounds leads t0 a decrease in the nitrogen concentration in the liquid phase, this process is also called deniirification. A wide variety of facultative prokaryotes perform dissimilatory nitrate reduction, most of them belonging to the Proteobacteria (Madigan et al, 2000). A number of obligate anaerobic. facultative anaerobic, and microaerophilic bacteria perform dissimilatory trate reduction to NH,*, mainly in ccarbon-rich, electron-acceptor-poor environments (Tiedje, 1988). Organotrophic denitifiers can use a wide range of natural organic substrates as carbon and electron sources, and have also been found to degrade several anthropogenic compounds (e.2., phenols and benzoates). In addition, several denitvifying bacteria can grow by fermentation. Although of lesser importance for treatment systems, lithotrophic dnitifers ~ which can use Hy, S°, and HS as electron donors ~ and phototrophie dnitifiers exist. In anoxie systems, anaerobic ammonia oxidation (Anammox), in which NO; is redueed to Ny using NH," as the electron donor is mediated by certain autotrophic ‘The accurrence of this process in methanogenic systems is not well microorganisn documented. In the context of an organic matter fermenting, overall methanogenic system, nitrate reduction will have the following effects (D) Channelling of electron equivalents (eeq) away from methanogenesis, which results in an overall decrease in methane production, The reduction of one mole ‘of NOs’ to No requites 5 ecg, whereas reduction to NH," requires 8 seq NO; + 6H" + Se 05N: + 31,0 NO, + 10H? + 8e"—> NH" + 31,0 Decrease in the methane content of the biogas as a result ofthe production of Nz ‘and additional COs (resulting from electron donor oxidation and denitrification), as well as alkalinity and/or NH," production, (3) Competition with other microbial groups for the same substrate(s). For example, denitifiers would compete with methanogens for both acetate and Hs, (A) Inhibition of methanogenesis by nitrogen oxides such as NOs, NO; and NO. @ Based on ths brief discussion, in an overall methanogenic system, nitrate reduction can have significant impact on both the carbon and electron flow, mierobial competition and inhibition, and gas composition. Such interactions were deemed 10 be too complex for inclusion in the first, generic ADM. Inclusion of nitrate reduction in the model will require an additional microbial population (denitriiers) with the corresponding kinetic parameters for substrate uptake (both organic compounds and NO;), functions for Partitioning of the total degradable substrate between fermentative/methanogenic and \denitifying populations, inhibition functions, ete. Although several of the steps could be regulated based on thermodynamic considerations (e.g., free energy), experimental data, kinetic data in particular, will be necessary for a more rational representation of simultancous nitrate reduction and methanogenesis. Pathways and Components ‘The main pathway for anaerobic fatty acid degradation above propionate (C3) is Be ‘oxidation, This isa eyelic process where one acetate group is removed per eycle fora yield of 13 ATP per cycle (Finnerty, 1988). ‘The final product of faty acids with even carbon atoms is aeetate only. When the fatty acid has an odd number of carbon atoms (e.2. valerate, Cy), one mole of propionate is produced per mole of substrate. Most naturally STR Draft 18/09/01 [20]‘curving LCFA have chains of even length (Gunstone, 1996). ‘The taskgroup considered three main fatty acid substrates (above C,) of importance; butyrate, valerate and LCFA. Butyrate and valerate are thought to he degraded by the same organism (included as such in the ADM1) while long chain fatty acids have a dedicated organism in the model because of the transport difficulties and different physicochemical characteristics of these much larger molecules. Thece acetogenic bacterial groups are therefore proposed, ane for propionate, one {or (butyratet valerate) and one for LCFA (>C3) A single group of organisms (methanogens) utilise the hydrogen. During. hhydrogenotrophie methanogenesis, a CO; molecule is activated and sequentially hydrogenated to methyl-CoM, which is then reduced to methane (Thauer ef af., 1993). Homoacetogens are also an important sink for hydrogen, especially at lower temperatures. However, these have not been included in the current model, but are discussed in a special Hydrogen inhibition functions for acetogenesis, Free energies for both acetogenesis and hydrogenotrophie methanogenesis are very low and both organisms may use proton and cation motive forces for partial yields as opposed to substrate level phosphorylation. The taskaroup discussed that a decreased yield could be used at decreased free energy levels rather than a standard inhibition funetion. Another funetion considered was the thermodynamic inhibition model of Hoh and Cord-Ruwisch (1996), in which the equilibrinm constant was used directly in a model, to prevent reaction at thermodynamically unfavourable conditions. However, to reduce model complexity, and increase flexibility (eg. for biofilm systems), the standard inhibition form was preferred for hydrogen regulation in the AMI. Ligvid phase hydrogen concentration was used for hydrogen inhibition. Because co-cultures may locally regulate hydrogen, bulk iguid and gas ‘measurements of hydrogen or formate concentrations may not necessarily dirty reflect the concentrations in the syntrophie consortia, but may be used to indicate responses to disturbances (€... Cord-Ruwisch et al, 1997). Initial testing of the model found that inhibitory hydrogen concentrations were 1-10 kg COD-m” liquid or 7-10° bar gas for propionate and 35:10" kg COD-m” liquid or 210" bar gas for butyrate and valerate (ce. inhibited 50% at these levels, with gas-liquid equilibrium assumed), which agrees partly with the thermodynamic levels. Other studies of biofilm systems (Costello e a, 1994; Romi et 4l,, 1995) have found hydrogen inhibition constants an order of magnitude above the inhibitory level, Other conditions sueh as substrate concent PH, cation levels, and weak acids could also antagonise thermodynamic inhibition by {noteasing maintenance. ions, acetate concentration, 3.6 ACETICLASTIC METHANOGENESIS In the major methanogenic step, acetate is cleaved to form methane and carbon-dioxide (Equation 3.1. CH,COOH + CH, + C0; AG'=31 KIM! (0.25 ATP) Bn Only a limited set of organisms utilise acetate; Methanoxaeta genus (3 species, homotrophic) and Merhanosarcina genus (S species, can also utilise Hz and COs) (Madigan et aly 2000), Methanosarcina dominates above 107M acetate while Methonosaeta dominates below this acetate level (Zinder, 1993). Methanosacia are pH sensitive and may be inhibited below a pH of 7 (Schmidt and Ahring, 1996; Boone etal. 1993) STR Draft 18/09/01 puThe difference in these organisms is the frst step, which is acetate activation to acetyl- CoA. Because Merhanosarcina operates at higher acetate levels, acetate can be activated in a two step process that requires one mole ATP/mole acetate, Methanosaeta uses two moles of ATP to assist activation of one mole acetate (at low concentrations; Ferry. 1993), ‘Therefore, Methanosarcina probably has an overall yield of approximately 1 ATP/mole needs a longer sludge retention time, but ean operate at very low acetate concentrations. more than Methanosaeta, thus allowing greater growth rate while Methanosaeta WEAK ACID AND BASE INHIBITION Free acid and base inhibition is the disruption of cell homeostasis by change in pH, caused by passive eansport of the free acid or base across the cell membrane and subsequent dissociation (Henderson, 1971). Because the relative amount of free acids or bases (compared to the ionic counterpart) is strongly pH dependent, the inhibition is also pH dependent, and the empirical pH inhibition Tunetions may include the cumulative effect of free acid or base inhibition (Figure). Free acid or base pH inhibition is particularly important for organisms utilising substrate to product reactions with a low energy yield, or utilise proton motive forees, such as propionate and butyrate/valerate ‘oxidising organisms, and hydrogen and acetate uilising methanogenic organisms. The following compounds are important as free acid or base inhibitory compounds (all pK, values at 25°C) (1) Free organic acids (HA, HPr, HBu, HVa): Main methanogenie precursors with Ka values from 4.7-49. Mainly include in models as aceti aid inhibition (2) Free ammonia (NH); Main free base in anaerobic digesters with a pK, value of 9.3, Inhibition function inchided in the ADMI for acetate utiliser (3) Hydrogen sulfide (1,8): While itis known thatthe free form of HS is largely inhibitory as compared to HS" or S* (Specce, 1996), the pKa is 6.95 would indicate the free acid butfers rather than disupts homeostasis. In this ease, the mechanism may be different. ‘Therefore, the free acids cause inhibition at lower pH values (HVFA's, HS) and free bases cause inhit Wat higher pH values (NH). The organisms most affected by free acid and base inhibition are (in order of affect) acetoclastie methanogens, hydrogenotrophic methanogens, and organic oxidising organisms, though the last two are in a syntrophie consortia, and decrease in activity of hydrogenotrophic methanogens Will cause apparent decrease in activity of organic oxidising organisms, due 1 accumulation of hydrogen and formate. In the ADMI, the effects of free organic acid inhibition are largely implicitly included in the empirical pH funetion, while the fiee ammonia inhibition is either implicitly included in the upper and lower empirical pH inhibition, or explicitly included in the free ammonia inhibition function. H;S inhibition is not included as sulphate reduction is not included. AS the major forms were implicitly included in other inhibition forms, an explicit form of free acid inhibition was not included. However, as the inhibition depends on the acid concentration as well as the pH, it is reasonable to include free organic acid inhibition when the concentration of free organic acids and pH FHuctuate. Also, because the inhibition may occur by disruption of homeostasis, rather than decrease of activity or increase in eell death, the most appropriate function may be inhibition vin decreased yield (Table 3.6, B2, Beeftink, 1990), rather than as decreased uptake rate STR Draft 18/09/01The presence of the two different organisms in anaerobic digesters is normally mutually exclusive with Meshanosaeta often found in high rate (biofilm) systems (Harmsen et al.. 1906; Sekiguchi et al.. 1999) and Methanosarcina found in solide digesters (Mladenowska ‘and Abring, 2000). Because of the exclusive nature of the system, the taskgroup recommends that a single group of aceticlastic methanogens be used with different kinetic and inhibitory parameters depending on application and experimental observations. 3.7 INHIBITION AND TOXICITY ‘peeve (1996) has «wo valuable definitions within the area of general restriction of biological processes: ‘Toxicity: an adverse effect (not necessarily lethal) on bacterial metabo Inhibition: an impairment of bacterial function" ‘The word bacterial should be expanded to “biological”, which includes other organisms ‘The taskgroup made the further Biocidal inhibition: Reactive toxicity, normally reversible, eg. LFA, detergents, aldehydes, vitro-compounds, cyanide, azides. antbioties and electrophiles: defined. by Speece (1996) a8 toxicity. Biostatc inhibition: Nonresctive toxicity, normally reversible, eg., product inhibition, ‘weak aeid/base (including VFA, NH, and HS) inhibition, pH inhibition, cation inhibition, and anything else that disrupts homeostasis; Loosely defined by Speece (1996) as inhibition. Forms of inhibition could be further separated into those that affect specific targets (eg. detergents on eell membranes) and those that affect overall cell kinetics and function (c.z., pH inkibition) Biostatic inhibition is of the most importance to anaerobic treatment (except perhaps. LCFA inhibition, which is biocidal), and is largely a result of the low yields available 10 anaerobic organisms. Most organisms that have an ATP yicld less than 1 molefmole substrate or eyele utilise cation or proton motive forees for anabolism rather than substrate level phosphorylation (Schink, 1997). This is true of methanogenic archaea (Ferry, 1993) and VFA oxidising organisms (Kleerebezem and Stams, 2000}. Weak acids and hases in Tree (non-ionic) form can pass through the cell membrane and dissociate, which disrupts proton motive force and homeostasis (Henderson, 1971). At ion and pH levels away from the optimum micro organisms must expend energy to maintain homeostasis rather than for anabolism. Therefore, while product uptake may change very litle, the yield decreases. ‘This _was recognised by Pirt (1965) who proposed a growth-independent maintenance coefficient. Increased energy use for maintenance limits the available energy for growth and consequently, the biomass yield. Flexibility t include the different kinetic forms was one of the reasons the taskgroup decided on an uptake related kinetic equation rather than 3 growth related kinctic equation as used in aerobic bioprocess models (e.g, ASMI; Henze et al, 1986), STR Draft 18/09/01 palLCFA INHIBITION Lipids constitute one of the main groups of organic matter and are found in both domestic ‘wastewater, organic household waste agricultural waste and industrial waste, Indeed, special industrial wastes such as abattoir waste, and wastes from oil mills have high content of lipids. Triacylglycerols are the most abundant family of lipids and the major components of depot or storage lipid in plant and animal eels Triacylglycerols represent lipids in the AD-model. During anaerobic digestion, lipids are first hydrolysed to glycerol and long-chain fatty acids (LCFA). This step is catalysed by extracellular enzymes called lipases. Hydrolysis of lipids proceeds rapidly compared subsequent steps (Hanaki et al, 1981); (Angelidaki and Ahring, 1992). The resulting LCFA are further degraded (o acetate and hydrogen via a number of sleps involving activation by acyl-CoA synthetase in cytoplasmatic membrane. The saturated LCFA-aeyl-CoA then directly centers the eyelic reaetion known as B-oxidation, while the unsaturated LCFA-aeyl-CoA are first hydrogenated before entering the 8-oxidation reaction. Acetate and hydrogen are the sole products from LCFA containing even number of carbons, while also propionate is formed from uneven numbered LCPA. G-oxidation of LCFA has been shown to occur both at mesophilic and thermophilic conditions (Weng and Jeris, 1976); (Rinzema et al, 1994); (Angelidaki and Abring, 1995). [Long-chain fatty acids can be inhibitory at low concentrations (Henderson, 1973); (Hanaki et al. 1981): (Roy et al.. 1985): (Rinzema et al., 1989), (Koster and Cramer, 1986) ‘Angelidaki and Abring, 1992). In LCFA Booxidising organisms, the LCFA are detoxified by activation with acyl-CoA to LCFA.CoA. Several mechanisms of LCFA inhibition have been proposed: ‘© they inhibit growth by competitive inhibition of the synthesis of LCFA essential to the structure of new bacteria ‘© they uncouple the electron transport chain from the proteins involved in ATP regeneration or transport of essential nutrients into the coll (Sheu and Freese, 1972). * They adhere to the bacterial cell wall and impede the passage of essential nutrients (Henderson, 1973). 1 has heen proposed that itis the associated form of LCFA that is inhibitory, and the inhibitory function via adsorption on the cell surface. Therefore, factors such as cell surface area to LCFA concentration ratio, and pH may have an influence (Hiwu et al, 1996). In zeneral, heavy inhibition is ireversible (i.e. toxic, Hwu, A&A, 1992), as recovery cannot be effected by decrease in influent LCFA concentrations (Angelidaki and Ahting, 1992); Htwu et al, 1996), While the most heavily inhibited organisms are probably acetoclastic methanogens all organisms are inhibited to a varying degree (Hwu, 1997), Rinzema/Ratledge, (Angelidaki and Ahring, 1992). While LCPA may complicate the process by inhibition, adaptation may also oceur, and a ‘well developed process will readily degrade feeds with a high content of lipids. This is because efficient LCFA degradation (in an adapted culture) will be able to remove LCFA as {ast as they are released from hydrolysis of lipids. However, in order to avoid high transient concentrations of LCFA, gradual acclimatisation is required. ‘Therefore, LCFA inhibition can have a significant impact on process operation when fed lipid-rich waste, and the first ADM will not deseribe reactor behaviour under transient high LCA concentrations; especially if toxie overload occurs. LCFA inhibition has not been jneluded in the first ADM, because of the potential complexity of the inhibition, and it’s lower frequency in processes (a8 opposed to the common included inhibition functions. Models that have successfully modelled LCFA inhibition include (Angelidaki et al., 1999), on manure and oil ‘STR Draft 18/09/07 palModelling of Inhibition Several forms of inhibition were considered apart from the more complex maintenance and decay kineties proposed by Beeftink (1990), based on the work by Pist (1982). While fundamentally sound, this set of kinetic rate equations were considered too complex, and disparate from the Monod kinetics most commonly used. Inhibition kinetics considered are shown in Table 3.6 and are: (a) reversible forms as proposed by Lehinger (1975), of which ‘non-competitive inhibition only was used (extensively) (b) two simplifications of Beettink (1990) that affect yield and decay respectively, which may be Valuable, but were not used in the model, (c) two empirical forms used for pH inhibition (Angelidaki er al., 1993; Ramsay, 1997), (4) competitive uptake, (which is not inhibition, but is included here for completeness), and (¢) secondary substrate Monod kinetics, which are necessary to describe decrease in growth when nitrogen is limited (Olsson, 1976; also not inhibition but included for completeness). More extensive reviews of inhibition and uplake/growth kinetics are presented by Pavlostathis and Giraldo-Gomez (1991), and Dochain (1986). When possible, because the inhibition forms in anaerobie digestion ate Varied and extensive, the forms are expressed as in equation 3.2, o allow for easy substitution of addition of inhibition foems. 62) ‘Where the first part of the equation is uninhibited Monod-type uptake, and I y=f(Si)..) fre the inhibition functions (Table 3.1, 3.2), where S; is the inhibitory compound for that inhibition function and process. When this is not possible, because the inhibition function is. integral in the uptake equation, the full uptake equation is shown in Table 36. PHI inhibition is a combination of disruption of homeostasis and increased weak acids concentration at low pH, or weak bases inhibition and transport limitations at high pH, and affects all organisms to some degree, Both the pH functions in Table 3.6 are useful for uptake equations, as the first form ean he sed in systems that are strongly buffered by ammonia, or other bases (above pH 8), and the second is more flexible when only low pH inhibition is likely to occur, for example, in carbohydrate systems. In the case of hydrolysis, inhibition at low or high pH is probably caused by partial enaturation of enzymes. Boon (1991) demonstrated the effect of batch digestion on primary sludge and showed an optimal hydrolysis at pit 6.8, but little significant change between pH of 6.5 and 7.5. pH inhibition of hydrolysis was not ineluded, but if required, functions are given in Veeken eral. (2000) and Sanders (2001). ‘The other important inhibition function is free ammonia (NH,) inhibition of aceticlastie methanogens and function (a) was used to describe the inhibition by ammonia. Many taskgroup members expressed a desire for the use of more fundamental inhibition funetions such as those in (b) and as used by Beeftink (1990) in the future but currently the pool of Knowledge is too limited to allow this. Hydrogen inhibition of organic acid oxidising organisms is discussed in section 3.5. Although important, the biocidal effect of LCFA was ‘ot included at present (but is addressed in the accompanying insert). Non-compettive inhibition was used in general, because it the most commonly used form in the literature, ‘which makes previously published inhibitory constants directly applicable, although some papers have found other inhibition Forms more ideal for (for example) hydrogen inhibition (Hoh and Cord-Ruwisch, 1996), organie acid inhibition (Mische and Jordening, 1999), and ammonia inhibition (Siegrist and Batstone, 2001). 'Noa-inhibition terms that ace teated as for inhibition terms include ammonia imitation as 4 secondary substrate for all organisms (necessary to describe halt in growth at very low ammonia, and to prevent negative ammonia concentrations), and competitive uptake of STR Draft 18/09/01 psibutyrate and valerate by C,, oxidising organisms. Competitive uptake is explicitly stated in ‘Tables 3.1 and 3.2 for clarity. Table 3.6: Inhibition forme. ‘Description uation ————SSCSCSC~ seo teageeetve froverentitos 772 compete — {(b) Reduction in yield ‘Not used sdoath rate pH inhibition when © moeraper and both high and low pH 5-12" lower inhibition 1 tower inhi inhibiton occur Ieexp -f =D PH inhibition when Empirical lower only ai Neca. f Oniylow pH inhibition 5-12” =1) Butyrate and valerate gg (4) Competitive uptake campettion for Gy (e) Secondary substrate ee All uptake 52 Table notes: 1, Processes where inhibition term used 2. Only one pH inhibition term used, and form 1 (with upper and lower inhibition) should not be used with free ammonia inhibition. For the First pH function, ply, and pHi. are upper and lower limits where the group of organisms is 50% inhibited respectively. For example, acetate utilising methanogens with a pHy, of 7.5 and a pHy, of 65 have an optimum at pH 7, For the second function, pH, nd pH, are points at which the organisms are not inhibited, and at which inhibition is full respectively. Acetate utilising methanogens with a pHy. of 7 and a pl, of 6 will be completely inhibited below pH 6 and not inhibited above plH 7. ‘Nomenclature: K, = inhibition constant, p,= rate for process j § = substrate for process j S, inhibitor concentration, X= biomass for process j STR Draft 18/09/01 6)3.8 INFLUENCE OF TEMPERATURE ‘Temperature ean affect biochemical reactions in five main ways: (1) Increase in reaction rates with increasing temperature (as predicted by the Arthenius equation) (2) Decrease in reaction rate with increasing temperature due to shift away from ‘optimum (240 for mesophilic and >65 for thermophilic). ACETATE OXIDATION Acetate oxidation isthe first step of a two-step reaction in which acetate is first oxidized to HYCO; (eq 1), and subsequently converted to CH. (Eq 2) (Zinder and Koch 1984), ‘This reaction is’ performed by a homo-acetogenic bacterium in co-culture with a hydrogenoteophie methanogen CHy-COO'+ 4H,09 4H, +2HC0, +H AG*=+104KYimol (1) 4H, + HCO, +H > CH + 34.0, AG" 135 KW/mol 2 The high Gibbs free eneruy for the acetate oxidation reaction (AG"= + 104 ki/mol) might suggest that the contribution of syntrophie acetate conversion to the overa digestion process is not very important compared to acetoclastic methanogenesis. Under specific stress conditions, oF those that favour acetate oxidation over other forms of ‘acetate removal, such as high temperature, its importance is considerably magnified. Petersen and Abring (1991) demonstrated that syntrophic acetate oxidation might contribute to up t0 14% of total acetotraphie methanogenesis in a thermophilic (60°C) digester. ‘At temperatures between 50°C and 65°C the predominant degradation pathway for ‘acetate depends largely on the acetate concentration, At low acetate concentrations the facetate oxidation pathway is important whereas at high acetate concentrations the aceticlastic reaction (Equation 3.1) is the preferential pathway (Zinder and Koch, 1984; Petersen and Ahring, 1991). At temperatures higher than 65°C the syntrophie acetate ‘oxidation is the predominant pathway as it is beyond the temperature range of the aceticlastic methanogens (Lepisto and Rintal 1999). In the ADMI it is considered that the majority of acetate will be degraded via the ‘aceticlastic reaction, Nevertheless, when regarding extreme thermophilic processes (T>65°C) or thermophilic treatment (45°C>1H3CO,] and [COs] can be taken as the effective acid. The two step reaction STR Draft 18/09/01 BuSOLIDS PRECIPITATION Solids precipitation isthe complexing of cations and anions in neural inorganie solid form. Potentially important solid precipitans in anacrobie digesters include calcium carbonate (CaCOs, pKy=8.2-8:5), calcium phosphate (CaPO,) magnesium carbonate (MgCOs, pK..=7.5-8.2), the metal sulfide precipitates (particularly FeS and FesS3) and in rater cases, magnesium-phosphate complexes such as struvite (MgNH,PO,) and newberyite (MgHPO,) (Musvoto etal. 2000a). Modelling metal sulfide precipitation is mainly of importance when sulfate reduction is modelled, and when Fe is added to precipitate the resulting, sulfide, an expensive option that is becoming less popular. This is therefore not addressed here. ‘The most important form of precipitate is CaCO (Van Langeraak, 1998), because of the commonly large amounts of Ca" in pulp and paper wastewaters, for Which anaerobie technologies are commonly used for treatment. The magnesium precipitates are of particular importance when the influent is high in Mz", ‘or Ma(OH)s is used t0 raise pH. The acidrase system Mz™/MgOH"/Mg(OH)s, and magnesium-phosphate derivatives must also be considered under these circumstances. Because of the complexity of the Mz” system, and because the CaCO system can be used t illustrate common complications, this is addressed specifically from here. Formation of the solid phase is a complicated process that depends heavily on Kineties as well as thermodynamies. The three phases are nucleation, crystallization, and ripening (Van Langeraak, 1998). ‘The second two processes are surface related, which ‘causes the rate at which they oecur to be dependent on the surface area (and hence concentration) of the solid phase. Additionally, beeause a number of additives and ean affect these processes: for example, phosphate inhibits the formation of CaCOs precipitates (Van Langeraak, 1998), There may be a number of different precipitates, ‘withthe same molecular formula, depending on precipitation rat, thermodynamics, and temperature. In particular, CaCO, ean generally be in the amorphous (forms faster and ‘at poorer thermodynamics, but fess stale) oF the exystalline form. ‘The main reason for excluding precipitation kinetics was the complexity of the process and range of different precipitating cations (and larger number of products), and hecause systems which have high levels of Ca°* and Mg"* are relatively restricted (see above). However, in order to model the physicochemical characteristics of these systems effectively, some form of precipitation mechanism should be included in the model. Excluding the precipitation process, while including the Ca’* ions as cations would cause the model to: (1) Over predict pH; as precipitation effectively removes 1 Ca” ion per HCO," ion, at around pH 7 (2) Under predict methane composition and over predict liguid inorganic earbon, as inorganic carbon is complexed during precipitation and (3) Under-damp physico-chemical dynamics generally, as the system is dynamically buffered by the pool of CxCOs, with slow precipitation Kinetics, These are particulatly important in high-rate anaerobic digesters, a the precipitated solids are retained together With biological solids. Additionally, it may be of value to inelude a precipitation process in order to assess the extent of ion precipitation, evaluate inorganie solids inventory and design new processes (e.g. Van Langeraak, 1997), Methods for including precipitation are given by Musvoto (2000, 20006), and Van Langeraak (1997), but the simplest method, for a single precipitant, isto include the precipitation process as an equilibrium reaction, or simple frst order kinetics. When a number of precipitant products are present, or the research or operational question is heavily oriented towards the kinetics of preeipitation and the physicochemical, a more complication precipitation kinetic system should be implemented. STR Draft 18/09/01 iyalso means that the HCO,!COsj4) is the slowest liquid-liquid transtormation with a time constant of 2500 «'" (25°C, Stumm and Morgan, 1996), which is still orders of magnitude faster than the most rapid biological process. Because association/dissociation processes are so rapid, they are often referred 10 as equilibrium processes and can be solved algebraically. The species considered important and their constants are shown in Table 4.1 Implementation ‘The equations used depend on whether the system of equations is solved dynamically or algebraically. Two approaches can be taken; the charge balance or tableau method (Morel and Hering, 1993). The taskgroup recommends the charge balance be used because itis easier to comprehend and has a better educational impact. However, the tableau method can produce he wsed to improve mumeric structure of the algebraic equations for implicit or dynamic solution. The charge balance can be expressed as Lse -LS, ESc- represents al cations and ES. represents all anions, ay Implemented using the TWA anaerobic model, the charge balance is as follows. (denominators for organic acids represent the g COD content per charge mole): SoS Sy Sear * Saye +80 — Fae ae seu *Suays +8 “Sueos “Se TIE 160 7 208 Sem Sau =o (42) Where S,,.. and S,_- represent metallic ions such as Na* and CI* and are included to represent strong bases and acids respectively (as well as, for example the saline partners of NH," and HCO, in added NH,CI and NaHCO). Sey aid Say ea be treated as otherwise inert compounds with no consumption o¢ reaction terms. LCFA'S were not included in the acid-base system, because the number of charged sites per COD is so small. However. if free LCFA inhibition is tobe used, or the LCA concentration is high, they must be included ina similar manner to VEA, ‘The acid/base equilibrium equations can be expressed in two ways depending on whether they are solved as dynamic state equations or as algebraic equations. If solved algebraically, the combined concentration ofthe acid/base pair can be expressed as a state variable. For the inorganic carbon (CO, ,/HCO, pai, this state is as follows: Se ~Sco, “Syeos = 43) Bicarbonate can be calculated using acid-base equilibria equations: Keo Si 7 0 44) Hoe Koco, #8) where Koco: is tne dissociation comsan. Likewise, for organi acids, and Inorganic nog, he inorgane fom ean be eleultd as follows KuvraSversout s,,,. ~ SevesSwrawet WN Kava +8 as) STR Draft 18/09/01 BalSySwv — so Me RS, / and hydroxide: a7 ‘Therefore, implemented as a DAE system, the fee form (@8 Scon and ionie form (e48s Syeos) aw lumped as a single dynamic stae variable (e, SyaS,). However, if the liquid phase physicochemical equations are implemented as dynamic equations, the free and total forms ane implemented as dynamie state variables, the lamped dynamic form (e.g. Si) is redundant, and an ‘additional dynamic rate equation is used for acid-base. This is specially addressed in section 5.3. ‘The biological production rates inthe Tables 3.1 and 3.2 can be either in the acid state variable or the ‘base state variable (but not both), though we recommend having the rate equations in the free form (ie, COs, HA6, 6). Sy. isthe only algebraic variable when the system is implemented as purely {dynamic equations, and iti calculated from the charge balance. Therefore the algebraic equation sc is explicit. A mixed solution method can also be used, with a number of the equations solved dynamically, and the rest solved algebraically. This approach was used by the ADM. taskgroup in order to provide the simulation package with a state variable in Scos for gas stripping (a specific limitation ofthe simulation package used). 4.2 GAS-LIQUID TRANSFER ‘The folowing three main gas components were considered significant as intermediates and as having a sirong effect on biological processes and on outputs (solubility at 25°C) + CO~ relatively high solubility (0.027 Myf") # CH, relatively low solubility (0.0014 Myg barge") Hy “relatively fow solubility (0.0010 Mya") Other potentially important gases included HS, which was not included because sulfate reduction was also not included as a biological process, and ammonia, which is so soluble (50 Mycbatya', Stumm and Morgan, 1996) that the mass flux to gas is negligible compared to that in the effluent. Gas-Liquid Transfer Equations Gas and liquid phases in contact will reach steady state with respect to each other. When the liquid phase is relatively dilute, Henry’ s law can be used to describe theeguilibrium relationship. Henry’ s law is commonly expressed as the concentration in the liquid phase duc to a gas phase partial pressure: o as) KuPpnin “Siqioe Where: Syqiseis the steady state liquid phase concentration for component i (M). Poni is the steady state gas phase partial pressure of component i (bat) Kir is the dimensionless Henry’ s law constant (Mba) Resistance to transfer of relatively insoluble gases such as carhon-dioxide, methane and hydrogen is mainly in the liquid phase (Coulson and Richardson, 1993; Pauss er al, 1990), Pauss et at. (1990) showed that gases in anaerobic digesters may be supersaturated to a STR Draft 18/09/01 i)significant degree in relation to effluent organics and total COD balance. Therefore, dynamic gas transfer equations should be used to describe gas-liquid transfer. The most common equations follow the two-film theory of Whitman (1923). Derivation is described ‘by Stumm and Morgan (1996) and the mass flux is a combination of the driving foree and the rate equation (4.10), Pos =RLAK Py “Sto (4.10) Where ka is the overall transfer coefficient multiplied by the specific transfer area (d") and p, i the specific transfer rate of gas i. tis necessary to correct Ky for Hy and CH, by a factor of 16 and 64 respectively to account for the COD basis of Syncyw a8 compared to the molar basis of Ky. Because transfer of all three gases are liquid film controlled, and the diffusivities are similar they should have kya values ofa similar order of magnitude. Values for kya vary a great deal depending on mixing, temperature and liquid properties, and for simplicity, we recommend using the same ka value for all three gases. This can be estimated using relationships to estimate k,a of O, in aerobic systems (which is also liquid {lm controlled; Metcalf and Eddy. 1991), or in systems producing medium to large amounts of gas, set an order higher than the fastest biological process for pseudo-equi 4.3 VARIATION OF PHYSICOCHEMICAL CONSTANTS WITH TEMPERATURE Changes in temperature have a fundamental influence on the physicochemical system, mainly because of changes in equilibrium constants, The overall effect om the system is due to changes in physicochemical constants with temperature is generally more important than that of changes in biochemical constants, The van" Hoff equation describes variation of cquilibria constants with temperature, and the taskgroup chose it hecause of its fundamental basis. Derivation and further details are given in Stumm and Morgan (1996) and Puigdomenech et al. (1997). If AH (heat of reaction) is assumed independent of temperature, the van‘t Hoff equation ean be integrated to equation 4.11, where AH” is standard enthalpy. R, the gas law constant, Ky is the known equilibrium constant at reference temperature T, (K) and K- is the unknown constant at TK}. Sa AN (tL aap kK, R(T TF, Alternatively if tis assumed that T)-T:#T)? and @ is substituted for (AH"/RT)*), equation 4.11 resolves (0 the following commonly used form (Siegrist er a., 1993: Angelidaki er al, 1909; Vavilin etal, 1994): K, =K,exp(-(T, - 1) ay) Between (°C and 60°C, equation 4.11 is effective forall equilibria constants in Tables 4.1 and 4.2 except the organic acids. However, the K, values for the organic acids vary by a stall amount inthis temperature range, and can be assumed constant STR Draft 18/09/01 Bs)Table 4.1: Acid/Base dissociation Constants (PKs) Reid Base Pa pK @SC) AN? Uimole")_ WAHT) T=29K COyHCO; 6.80" 7648 0.010 NHS"INHe 9.20' 51965, 0.070 HeSIHS. 6.05! 21670 0.029 HeO(OH'H') 14.00" 55800 0.078 HAciAc 4.76 — Maximum of 4.81 at 60°C? NA Herr 4.88" Maximum of 494 at 60°C? NA s-HBUBu 482° Maximum of 492 at 60°C" WA FHBuBur 426° ‘No other data NA asHVaNVai 426° No other data NA iHVava 478 No other data NA T. Snoeyink, V.L. and Jenkins, D. (1980). 21 Sill, L.@. and Martel AE. (1964). 8. Does not it constant enthalpy form of vant Hoff (equation 4.11) Table 4.2: Gas transfer parameters Gas KSC) AIP (Tmole") HVAT) T1=296K —— Difugily at 208K Mis batgn* ims"xi04) * Fe 0.0010" “#180 ~0.00866 485 cH. 0.0018? 4240 0.01929 187 co, 0.027, 39410 0.02628 1.98 1. Perry and Green (1984) 21 From Pauss ef al. (1990) 3. Correct by a actor of 16 (Hz) and 64 (CH,) to change Ky rom Mar to kg COD-m*-bar" STR Draft 18/09/01 165 Example: Model Implementation for a Single Stage System Implementation of an anaerobic model depends on whether the physicochemical equations are imp! nti equations. In the Fitst ease, solution requires a differential and algebraic equation (DAB) solver. Tn the second, only a differential (DE) solver is required, but the differential equation set is stiffer, and an increased number of errors are introduced, ‘This section mainly deals with implementation as a DAE system, with notes to implementation as a DE system. For testing and analysis, the taskgroup implemented the ‘model as a DAE system, with CO,-HCO,” acid-base processes implemented as a differential equation. The system implemented here is 4 consiant volume completely mixed system (Figure 5.1). ‘There is a simple, but limited method for accounting for biofilm systems, a different implementation is required for more complex hydraulic systems such as solid phase digestion, This is assisted by the high portability of the rate matrices (Tables 3.1 and 3.2, We also often included control of pH by addition of IM caustic in an additional input stream, though this is not included here for simplicity tated as algebraic or di > dn Sp Prat Sos Pas (Gas Phase sree Pons APoar Vow | Sou Pens Liguid Phase Vu 5, don >| a $ Biochemical and — Sigs Physico-chemical ‘Conversion processes. * Kase x Figure 6.1: Schematic of a typical, singla-tank digester (q = flow, m®d" V = volume, mé; Simin = concentvation of guid components; Xx concentration of particulate ‘components; alin kg COD.m "is the component index, see Tables 3.1 and 32) STR Draft 18/09/01 ba5.1 LIQUID PHASE REACTIONS FPor each state component, the mass balance can be writ as in equation 5. VS, a Sas Saidar¥ DMs a where the term Jp,Vay isthe sum of the specie rates for process j mailed by, (Table 3.1 and 3.2), If « constant volume is assumed, the expression can be in Sigs a in equation 5.2, If the volume is not constant with time, iis also a dynamic state variable, and the chain rule must be used to express the concentration dynamic state equations in dS, as, Ifthe residence time of the concentration state is variable, for example, solids in biofilm ‘or high-rate reactors, the retention time ean be extended by replacing the second term (mass ‘low out) as in equation 5.3. Kags — GaXins x Vig tex FV ou Yow, a) Where tex is the residence time of solids components above hydraulic retention (Ge. iF x=0, the overall sludge retention time, or SRT is Viq/dya) 0 simulate separate solids retention (@). This is not a perfect implementation, as biofilm systems are very complex, and more fundamental solids retention models have been published by Bolle et al (1986) and Buffiere et al. (1998). However, implementation of these models is beyond the scope of this por. Tn addition to the rate equations in Tables 3.1 and 3.2, the following liquid/gas transfer equations for Sy, Sco and Sic (oF Sco, depending on implementation) must be added: Prt, SKLAR 1,Pyasit, “Sigit,) (549) Pr.cn, =K1a{K yen Pomc Sige) (sb) Pra L4(Ky.co,Pgasico, ~Siaco,) (S.4e) where Pr, is the transfer rate of gas i and Syqco: isthe fraction of inorganic carbon as co, 5.2 GAS PHASE EQUATIONS ‘The gas phase rate equations are very similar to the Liquid phase equations, except with no aadvective influent flow, and with only dynamic state components. ‘The dynamic states can be either in pressure (bar), oF concentration (M or kg COD-m"). In our implementation, we used gas concentration, with pressure ealculated from concentration with the ideal gas law SIRT where § is the concentration in M. The differential equations for the gas phase ate of the form in equation 5.5 STR Draft 18/09/01 18)Vie = +P 65) ot Vow Vue ‘The teem Vug/Vou is required as the gas transfer rate is liquid volume specific. The pressure of each gas component can be calculated using the ideal gas law for the three gases (nar, factors in denominators are COD equivalents of the gases): Pont, =Spusn, RTS (36a) Powctt, = Spor RT/O+ (5.60) Pysco, =Seuco,.RT (5.60) If it is assumed that the headspace is water vapour saturated, the following relationship cean be used (Himmelblau, 1991) for water pressure: Poamo 323-107-200 on) ‘where T is the temperature in K. The most common way to ealewlate the gas flow, isto set it equal to foal gas transfer, corrected for water vapour (equation 5.8) RT ee Mains Pr Prec) ow) So where Ppa is the set headspace total pressure (normally 1.013 bar). If the headspace pressure is variable, there is downstream processing of the gas, or the modelling package requires it (asim our case), the gas flow can be calculated by a control loop in pressure. To 4o this, the gas phase pressure must he calculated from partial pressures (eq. 5.9), and the ‘low ealeulated by a control loop (in this ease proportional control eq. 5.10). Poa = Past, +Pymenn, +Prcco, * Ppt. 69) yn hy (Pax ~ Pan) 6.10) ‘where kp proportional gain, nominally set in our case 1 10,000 m'xt"-atm", This function also has an advantage in tall reactors, where the gas pressure duc to hydraulies may be significant higher than laboratory reactor. 5.3 SPECIFIC EXAMPLE: INORGANIC CARBON ‘This is an example of implementation of the inorganic carbon states in liquid and gas phase for both DAE and DE systems. (DAE System Substituting the inorganic earbon (S,c) into equation 5.2 gives the liquid mass balance: 61D STR Draft 18/09/01 139)where Siq28iqjc> In addition, the free and ionic fractions of Sie (Seas and Sycos respectively) are caleulated as part of the algebraic set of equations: SsuurShuse Sisco, 6.12) Koco, #8 K,co,Suue ow Koco, *Siqu ” ‘where K,con isthe acid-base equilibria constant, The gas phase equations are exactly as shown in section $.2. All S, variables except Sic are algebraic. (ii) DE System ‘When the inorganic carbon system is implemented as a DE system, the variables S, Sjicos are dynamic and the Sic variable isnot used. ‘The two dynamic equations for Seon and ShicasFespectively ae! isco 5 Siseos uN ‘i eu and Syxnco bos 5.15) ss 15 where S3Siycom Tere is lo am adnate equation for ai hase etons Paroco2 =ka/nc026 co, Sigur? ~Kaco,Suyco,) 65.16) where Paco: is the production rate of CO, from HCO,, (ie, base t0 acid, Mat), Kyo the dynamic constant, sominally set to an order of magnitude higher than the highest biological rate ceonstant alter adjustment of units to reduce model sitfness (Md") and Kycex isthe COJHCO; equilibrium constant (ie. acidity constant, MD. If all acids and bases are implemented in this ‘manner, the only algebraic equation required is the charge balance and there is an explicit solution as itis a single equation, STR Draft 18/09/01 40)6 Suggested Biochemical Parameter Values, Sensitivity and Estimation ‘The key criticism of structured models is that because of their complenity, there are « large ‘number of parameters. Fortunately, the kinetic parameters are generally less variable than in ge systems and a wide Variety of parameters are shown in Appendix A. Some hiometric coefficients are given in Table 6.1. A list of suggested kinet parameter values for mesophilic high-rate and solids, and thermophilic solids digesters as ‘well as qualitative variation and sensitivity are given in Table 6.2. ‘These parameters have been tested on data sets for consistency, and will give reasonable prediction of syster response under a number of conditions, but are largely arbitrary, based on our experience. For more specific values from a number of studies, see those in Appendix A. Physicochemical parameters are given in Chapter 4, and (with the exeeption of kya) are independent of application. ‘Table 6.1: Suggested Stoichiometric Parameters and Qualitative Sensitivity, Vaiablty Parameter (no ai Description Vaue——V Souxe ‘Soluble ners from compex 0.1 24 Sine Particulate inerts from complex 0.35 2 Sevee Carbohyarates trom complex om 24 Sue Proteins from complex om 24 Sie Lipids trom complex om 24 Fai Falty acids rom pide 0.95 + 2 Foes Hydrogen trom sugars 019 + 3 Fours Butyrate from sugars om 33 Sooas Propionate rom sugars oz 33 Soon Acetate from sugars ost 303 Tras Hydrogen trom proteins oo 8623 Foose Inorganic mitogen fom proteins 0.00723 Srase Valerate irom proteins oz 023 Suse Butyrate rom proteins 02 82 3 Soom Propionate from proteins 00 23 a Acetate trom proteins 04 23 v"Wahanaiy of parameter" Vas vary Tie botwoun ocean. 2 Vara Donwoon processes and subevates, 5, Varies dyamaly tin process, “hls 1 See Gossett ane Belsr (192) testinate ners in acated te 2 Based here on palate tgiycerde, Vates benveen 81% and 86% depending on LCFA chain ong 2. See Appr to extmae products rom sugars and amine aid, STR Draft 18/09/01 mnMethods for biological parameter estimation for anaerobic models was one of the areas in which the literature is less developed, and (as stated in the foreword), we believe that a ‘number of valuable conteibutions can be made to facilitate systematic and repeatable parameter estimation, and just as importantly, parameter sensitivity and identifiability Associated with this, we found numerical methods for parameter eptimisation and parameter ‘identifiability to be very valuable during internal testing and analysis of systems Gn particular, the Simplex method; Nelder and Mead, 1965; was used extensively for parameter optimisation). Our strategy for parameter estimation was generally to minimise the number ‘of biological parameters to be optimised namerically by (a) taking those with @ low variability, such as Kj, and Y from literature values, (b) taking more variable parameters from studies with similar reactor design and feed matrix (if available) and (c) reducing parameters by numerical analysis of sensitivity, correlation and identifiability, This reduction process generally resulted in requiring numerial estimation of 1-2 parameters each for the acetogenic and acetoclastic biological groups, or hydrolysis for optimal prediction, Reasonable prediction ofall variables was in most cases achieved by (a) or (b). Parameters requiring numerical optimisation were generally those in Table 6.2 that have both a high variability and cause significant changes in model outputs, and are addressed further below. Hydrolysis Parameters. In many cases there are one or two significant parameters. In solids digesters fed with a relatively homogeneous substrate such as primary or activated sludge, the important kinetic parameter is disintegration of composites, as subsequent hydrolysis is commonly much faster. The most important stoichiometric parameter is the inert fraction (Pavlostathis and Gossett, 1986). To estimate incts fractions (I-D) and disintegration constants, Pavlostatis and Gossett (1986) and Gossett and Belser (1982) are recommended. In systems fed a heterogeneous mixture of particulate protein and fat, the influent matrix ‘would reflect this, and the important parameters are protein and fat hydrolysis. In this case, the disintegration provess is only used for recyeling of decayed biomass, which isa relatively small fraction of the COD flux. Fitting of parameters in these systems i8 assisted as the feed ‘normally contains either protein and fat (mainly animal or food- processing), or carbohydrate particulates (carbohydrate food processing, brewery, starch etc). Parameters associated with propionate. ‘The key parameters inorder of variability are Ks yo, ka yo ad Kun These ean generally be fited by either a number of steady state conditions, or a single dynamic experiment Under higher loading conditions, hydrogen inhibition becomes important. Parameters associated with acetate. ‘The critical parameters here are again Ks and Kg These ean normally be estimated as for propionateunder medium and low loading conditions. When close to the inhibitory pH, pH inhibition can also have an effect on fitting of the above parameters. We have also observed some variation of the decay rate for aceticlastic methanogens between solids digesters and high-rate systems, with the decay inereasing in solids digesters. Kiyo is also very important in systems with high ammonia concentrations but we have observed a low variability in this parameter (Siegrist and Batstone, 2001). STR Draft 18/09/01 payAppendix A: Review of Parameters Te A Dlragatan anya Parana eee RG) Bye) Bc) Remini 7 Tem 2 om on inom eagiermiss = 2 te ‘och binary susce = ‘ ow cen stesso pmarotage 514 = 1 oat oa Seroj pmarycge 5 = ‘ oa ose Gmoyaste—pimyskage a? = ' 2 ak ceain ro = . ‘ots ‘inane ost = * Coxe ‘inane Fonsi 5 10¢man 194 030 on sien sepia alge = ‘one oa os ccm ayatte may ge = = 270 inane elutes = ® 08 inamc— inwyage 7297 = ‘ oe Siname mw cge 72.77 =ee (condi) wp ohbny eoobem shy a wetide> a aaa STE +e tae bol Gio accom areas $US GSE Sh Ge Scam ees Soom SP Ge oh ce Sas ee fee aeet 228 ame ne ap sam eso acme cate RIN ag te tot t8 ca cute pte 887 Sh uy toh de tone tar me Paya oy Se Gis unre So TSB a eg cecodio'e naclon co0teo) oh 3s uci waco ea aaa ae ee a ‘i Ar Poem Bag ee a FFF ge Ena ag bn coo co i ee 7 gS ng SS 7 keer cunt vem Mama Sonn Suet ce Spat st feo the Setar ix mo Ae Sika Saas S38 lemme ok eran Sf tae ae oe i Sa aes can oane a7 ne SNRs oto ao onah_oyoo ox! omnesbo") agcbbm} icopcoo') ht oe F a Sore ome ann PRR ay firms) sehen” at toe iste = oss ede ton Som Site penton Fear tt tes tom Soro ligeeson soe ‘aman tes tumors ome tan "Kw" hay a837 on tse sont Sone ome Oar eee. nets BEST a ‘Tomes Gm oe oat) gues" oo Got Gos oe in mum aiayaige 387 ak ouroty one Os 8m} oee 13) Symmes pomuyoue Gy $8 buen ta Gass coms one om) ior Saar cae Sune ceomte 97 ai asm onan oan dye RBs te og cam aie sunystte “Rote (Gujrand Zeer 943 Lobia Vasiin (1, Novak and Carbon (1970 1b Othe (8 1. Para lo (999 12 Pvt ang GaAppendix B: Integration with the ASM ‘This appendix details methods of integration in common oF linked models with the ASM. series of activated (aerobic) sludge models (Henze et al., 1985, Henze et al, 2000). There are two major differences in implementation of the ADMI, compared to ASMs, excluding structure, states in general and the physicochemical system: (1) Units: kg COD” instead of gam” (mgt); M instead of mM for HCO, M instead of g Nem? (mg N* (2) Kinetics: rein substrate uptake rather than growth rat ‘The reasons for these differences were vatied and are covered in detail in the report but cean be summarised as: (1) to accommodate the physicochemical system effectively and in agreement with the majority of anaerobic kinetic studies and (2) to allow for a beter representation of inhibition kinetics when simulating lower yielding anaerobic processes, However, when modelling anaerobie processes together with aerobic processes (either in the same model, linked models, oF distributed models), it may be desirable to use the same implementation for the ASM and ADM models in order to compare kinetic parameters, and allow use of common states, and we recommend use of ASM protocols in these circumstances, ‘There are four major scenarios for implementation of the ASM and ADMI models together (Figure B.1), 1. | Vessels ‘Vessels 2. | Vessels vessels modeled >> mocotos mogaies }—> mecaloa DyASM 5, x, BY ADM. By ADM g, x, by ASM 9g, setvated/atmary 2g, protommamers ‘sodge digesters 3. Vessels. > Vessels 4 Vessels modeled inednod mnadoled brash & ‘bya by abit ons ASM ached primary 44g, anaerobic zoe n EBPR, stig digesters wh ‘Suarobc smmona oaaton veevee steam Figure 6.1: Scenarios for integration of the ASM and ADM showing outputs trom ene mode! (Ss. being used as inputs tothe other model ‘The first three cases can be relatively simply modelled by using the ASMs and ADML together. However, the fourth ease is more complicated, and is not addressed in this report Anaerobie polyphosphate release for example, is modelled by the ASM2d (though not as an anaerobic process, Henze ef al, 2000), and the ASM structure is more appropriate for this process, at least in activated sludge treatment systems, When modelling other special processes normally addressed partly by the ASM in anaerobic zones (or visa-versa) itis probably better to adapt cither the ADM or ASM structure of equations to incorporate those processes, or produce a specialised model. (sa)Tyody “Has Yel, B53) Pirowin Kos YoY Ky+s Where 1). are the Secondary substrate, egulation or inhibition functions. Values for Hs Cam be calculated from Haq =ky/Y, and are also given in Appendix A. Outputs from the ASMs as inputs in ADMI Table 8.2: Non-zer0 outputs trom the ASMs as inputs to the ADM "AON name ADH input Valve rom ASH outputs Complex parcuates X EX, wxchcing % Particulate inerts x x Soluble inert 5 s, Inorganic carson Se Sai Inorganic nitrogen Sn Set Cations or anions San Sun none, estate om measured pH Nato: normal thor shou bo no nivale (Syms) or oxygen (Sa), 0” RBCOD (Ss) In tho feed to an anaerobe digester (nominally ftom the clarleruderion) I here i, ihe frst {wo cases, an equivalent amount af X. GOD should be removed as aed (or see insert on nitrate reduction), while Se can be nominally added as amino acids (Su). Outputs from the ADM as inpuis in ASMs Table 8.3: Non-zero outputs trom the ADM as inputs to the ASMs ——=8iirane_____ASWinpat_____"Valts fom AD outputs ‘Soluble ners Ss. Ss 1s, used in ASM Recoo Sue 2.08, ,exctuding S, Swand Sec and Sy i Sais used in ASM cluding 8, Swand Si, if Sais not \VFAs (acetate) Ss DSve Particulate inerts! x x sBcoD Xs LX; excluding x: NH'sNHs nitrogen Ste Sn Slowly brodegradable Ske hone, tt ‘organic nitrogen Patlculate biodegradable organic Xo none, fe nitogen Inorganie earbon Sux So Note: 1. Some organics not Bode pradable via anaerobic digestion are degradable aerobically It this Is observed empirically (via low levels of inert COD in the effluent), or found experimentally, by biodegradabily analysis, some S\ or X; could be diverted to Ss and Xs, respectivelyAppendix C: Estimating stoichiometric coefficients for fermentation Stoichiometric coefficients from monosaccharide fermentation ‘The ratios of products from monosaccharide fermentation can he largely simplified by the necessary balancing of elements, and assuming that fermentation proceeds by the three key and propionic acid may result from the second reaction, but for estimation of all four COD products, the different ratios are implicit in the frst reaction, ed in Table 3.4 as summarised in Table C.1. Different ratios of acetate/ ‘Table C.1; Monosaccharide equations as implemented in the ADM Products Reaction 1 Acetate (CsHizOs+2H,0-92CH.COOH+2COn4He 2 Acetate, Propionate __SCsH1,0y-> 4CHsCH,COOH+2CH:COOH+260:+2H:0 3 Butyrate CeHigOe-9 CH CHsCHCOOH+2C0:+2Hs ‘The fraction of monosaccharide, which degrades via the frst, second and third reaetions cean be expressed a Mor Meu ANd Mau respectively, where Maat Moa Ma cH ‘Therefore, the stoichiometric coefficients of each product can be calculated from these fractions, and the relative COD of each product in Table C.1 (Table C.2). Based on this, the four stoichiometric parameters can be reduced to 2 parameters, and COD balance assured. We coded these functions into the model. ‘Table C.2; Stoichiometric costicents from monosaccharides uptake, based on Table C.1 Products Goeticent Acotato Foon= O67 Mi 40-22-The Propionate Soagn 0.78 Bulyrale Sous sar th Hydrogen 33 pect0.17(1~ Tone Mh Stoichiometric coefficients from amino acid fermentation An initial estimate ofthe stoichiometric coefficients from amino seid fermentation can be made from the amino acid mix of the primary protein(s) i's Stickland! donor or acceptor status, and the characteristics of Stickland reactions as described in Section 3.4. ‘The donorlacceptor status of the amino acids are shown in Table C.3, Ramsay (1997) compiled the reactions based on this table into a spreadsheet that can be used to predict the products of amino acids (Table C.4). This ean be used together with the amino acid content of a protein fr protein mix (from analysis or FAO, UN, 1970; two shown in % mole total amino acids) to predict the stoichiometrie yield from a mixture of proteins. While it should not be necessary, the result can be normalised in COD. C,, (the earbon content of the protein) can be easily analysed by TOC/COD analysis, or ealewlated from the amino acid content. (37)‘Table C2: Amino acid Stickland acceploridonoruncoupled status (Rameay, 1997) Hydrogen, ‘Acceptor ‘Ay Donor AKyl Donor Alkyl Donor/Acceptor Isoleucine Alkyl Donor ‘Serine Alcohol Donor Threonine ‘Aloohol Donor/Acceptor cysteine ‘Sulphur containing Donor Methionine Sulphur containing Donor Praline Forms ring with amino Acceptor Phenylalanine ‘Aromatic Donor/Acceptor Tyrosine ‘aromatic Donor/Acceptor Tryptophan ‘Aromatic Donor/Acceptor “Aspartic acid Carboxyt Donor Glutamic acid Carboxyt Donor Nitrogen containing Donor igen containing Donor Histaine Nitogen containing Uncoupled ‘Table C4: Products from Amina acids based on Stickland reactions. (Ramsay, 1997) Hanne Reid Molecular HAc HP HBu AVa IN IG Other He ATP Beat, Cacoin” Formula Flesh’ Arginine CeHOsNe 0S OS 0 05 4 1 0 1 1 5.40% 2BO% Histidine GH.ON; 1 0 05 0 3 1 1 0 2 2.40% 260% Lysine GrlwOsNz2 1 0 7 0 2 0 0 0 1 720% 6.40% Tyrosine GOuHNN 1 0 0 0 1 1 0882 1 1 270% 4.90% Tryptophan CyHzON 0 0 0 0 1 1 1471 2 1 0.90% 0.80% Phenylalanine CHiN 9 0 9 0 1 1 1478 2 1 380% 4.00% Cysteine CsHsONS 7 0 0 0 1 1 0 05 1 150% 010% Methionine GsHinONS 0 1 0 0 1 1 0 1 14 2.00% 250% Threonine GHLON 1 0 08 0 1 G O 4 1 480% 3.90% Serine CHON 1 0 9 0 1 1 0 1 1 5.70% 7.60% Loucinoy GiHyON 0 0 9 1 1 1 0 2 1 14.40% 14.20% Isoleucine Valine G:HO:N 0 0 1 0 1 1 0 2 1 850% 6.70% Glutarine G:HO.N 1 0 08 0 1 1 0 0 2 18.80% 19.20% Aspartate CON 1 0 0 0 1 2 0 2 2 8.80% 6.40% Giycine G:HON 1 0 0 0 1 0 0 + 0 8.40% 3.00%. Alanine GHON 7 0 0 0 1 1 0 2 1 S40% 400% rol GHON 05 05 0 05 1 0 0 1 0 400% 11.40% ‘asic units of mole procuetimole amino acid. 1. Units of mole% fetal amino acids, ‘Table C5: Calculated products from amino acids, Source Propionate Butyrate Valerate He feos Suuss fins Susa__ fran Casein 007 027026 0.008 0.07 Bee! fash 07 019 ox 0011002