Get Complete Project Material File(s) Now! »

Predicting methane production: the analytical way

The main aim of AD modelling is to predict methane production from an organic matter source defined by its own characterization.
Traditionally the performance of AD in wastewater treatment was evaluated using parameters such as chemical oxygen demand (COD), total organic carbon (TOC) and biological oxygen demand (BOD). In order to optimize plant design and operation, Raunjkaer et al. (1994) proposed to link COD fractions and biodegradability (useful for modelling purposes). Kayhanian et al. (1995) showed that the content of biodegradable volatile solids (VS) impacted the prediction of biogas production rate, the computation of the organic loading rate and the Carbon/Nitrogen (C/N) ratio. However, since the seventies, the most widely used indicator to assess the performance of the digesters is the amount of methane produced per unit of Total Solid (TS) or Volatile Solids (VS) of any given substrate (Chynoweth et al., 1993). The most commonly used method to measure anaerobic biodegradability is the biochemical methane potential (BMP) (ISO EN 11734, 1995).
The BMP assay is a procedure developed to determine the methane yield of an organic material during its anaerobic decomposition by a mixed microbial flora in a defined medium. The procedure was developed for a serum-bottle technique by Owen et al. (1979). Angelidaki et al. (2004) described the procedure and the calculations. The test ends when the cumulative biogas curve reaches an asymptote, usually after 30 days of incubation but it may be much longer for non-easily degradable material such as fibers e.g. 200 days for cardboard (Abassi-Guendouz et al., 2012). Therefore, the main inconvenient of the test is the time consumed. Chynoweth and Isaacson (1987) wrote that maximum theoretical methane yield determination was useful to evaluate digester performance and to provide basis for experimental work. However, the literature reports different analytical conditions for the test and many factors may influence the anaerobic biodegradability of organic matter. Enhancements of this method led to different parameters studies: substrate/biomass from inoculum ratio (S/X), pressure biogas measurement, macro and micronutrients additions, etc. (Owen et al., 1979; Gledhill et al., 1979; Shelton et al., 1984; Battersby et al., 1989; Kameya et al., 1995). More recently, a specific group from IWA (i.e. the specialist Group on Anaerobic Biodegradation, Activity and Inhibition Assays) has been set up in order to discuss about BMP methodologies and to propose a standard protocol (Angelidaki et al., 2009). Thus, first guidelines for a definition of a new international standard protocol were defined.
In the same way, an international interlaboratory study has been conducted in order to compare the BMP test with substrates such as starch, cellulose, gelatine and biomass material (Raposo et al., 2011). Nineteen laboratories participated in the study, using different protocols. Except for the gelatine, a small number of outliers were obtained. The relative standard deviation ranged between 15% and 24% and decrease to 10% when the outliers were not considered. The influence of inoculum, temperatures, volume, and headspace gas appeared to be insignificant. However, kinetic rates were widely different (standard deviations ranged from 57% to 68%) and they were impacted by substrate/inoculum ratio.
In order to reduce time consumption, other ways to determine an equivalent of the BMP value have been investigated using several kinds of organic matter characterization techniques.

Predicting methane production: predicting tools

According to Buffiere et al., (2006), “methane productivity not only depends on the amount of degraded volatile solids, but also on the nature of the solid: carbohydrates, proteins or fats have different methane potential.
Consequently, the biochemical composition has become an important descriptor for anaerobic digestion, both for production prediction and for kinetics assessment”. In other words, biochemical composition is required for the use of integrative tools such as models (static or dynamic) and to achieve an accurate prediction of digester performance. As Angelidaki et al. (2004) concluded, methane yield depends strongly on the nature of each biochemical family in addition of the COD content. Integrative tools, in this review, are the implementation of different relationships between the organic matter composition and the methane production or the anaerobic biodegradability. Static models are correlations (obtained by linear regression or partial least square regression) where the variable of interest is explained by one or more variables based on some analytical composition of the given substrate. Static implies neither kinetic equation nor variation over time. Dynamic models include these variations and are usually more complex: biological reactions are explained by kinetic equations such as the Monod type and included in differential equations representing mass balance in the process. In the following paragraphs, an overview of the different integrative tools found in literature is presented.

Static models

• Correlations between organic matter composition and anaerobic biodegradability
Theoretical BMP has been calculated since 1930 with the Buswell formula (Buswell, 1930). The stoichiometric equation is based on elemental composition (CnHaOb) where organic matter is reduced to methane and oxidised to carbon dioxide (equation 2.1), with the assumption of total conversion.
Derived from the Buswell formula, another existing relationship (equation 2.2) is based on the knowledge of the main biochemical composition of a substrate, carbohydrate, protein and lipids, and based on the stoichiometric conversion of model compounds in COD (Raposo et al., 2011).
Shanmugan et al. (2009) calculated the empirical formula for each waste based on the results of the chemical analysis. The formula was used to estimate the COD equivalent and the stoichiometric methane potential with the Buswell equation (Buswell, 1930). The measurement of elemental composition (carbon, hydrogen, nitrogen and sulphur) was used to characterize different types of sludge and municipal solid waste. The methane production potential calculated overestimated the experimental one. Lesteur et al. (2010) explained that measuring elemental composition is very fast but the obtained value takes into account all the organic matter, without any differentiation between biodegradable and non-biodegradable organic matter. Moreover, part of the biodegradable organic matter used for bacterial growth is not taken into account by the Buswell formula. Additionally, when applied on municipal solid wastes, Davidsson et al. (2007) showed that theoretical methane potential is more realistic when calculation is based on biochemical composition (lipids, carbohydrates, proteins) rather than on elemental composition analysis.
During the last two decades, several authors tried to build other static integrative tools based on organic matter characterization but they are mainly applied to municipal solid waste (Buffiere et al. 2006), kitchen, fruits and vegetables wastes (Gunaseelan, 2007 and 2009). Few studies dealt with municipal sludge although the methodologies used on solid waste can be transposed to sludge. The most recent publication has been made by Mottet et al. (2010) and Appels et al. (2011).
Seeking an indicator of biodegradability, Mottet et al. (2010) proposed to link Van Sœst fractionation with biodegradability of sludge, using partial least square regression. Extraction mainly occurs with the first neutral detergent (50% to 80% of TS). Following detergents, targeting hemicelluloses, cellulose and lignin, extract little material (5-20% of TS). Thus, this method, adapted for vegetable wastes, was not suitable for municipal sludge (mainly proteinaceous). Previously to Mottet et al. (2010), Chandler et al. (1980) showed that the anaerobic biodegradability was inversely proportional to the lignin content (equation 2.3). Buffiere et al. (2006) found an interesting relationship between the sum of cellulose and lignin percentage of VS to the biodegradability of kitchen waste. In the same way, Gunaseelaan et al. (2009) showed that there was a correlation between biodegradability and carbohydrate, proteins, lipids, acid detergent fibres, cellulose and ash concentrations obtained with Van Soest method. An accuracy of 94% was obtained when applied to fruit and vegetables. That approach was validated on real scale plants (equation 2.4)
Contrary to previous studies, Mottet et al. (2010) observed that the Van Sœst fractionation cannot be used as a tool for biodegradability prediction. Applied on municipal sludge, the error for the validation model is about 35%. These authors highlighted that it would be interesting to develop a new method based on successive extractions more adapted to this substrate.
In the second part of their work, the authors found a better correlation between anaerobic biodegradability and specific fractions of organic matter (equation 2.5).
The oxidation degree (i.e. COD/TOC), the proteins, carbohydrates, lipids and soluble organic carbon percentages of VS were input variables of a PLS model. The validation step gave an error of 11% and the model regression coefficient was 0.938. However, the number of used secondary sludge used was small (6 sludge used for calibration and 4 substrates used for validation, including cellulose) and the biodegradability range was narrow (35% to 66%).
In the same way, Appels et al. (2011) developed a PLS model to predict the BMP of waste activated sludge with 19 characterization parameters (soluble and total COD, soluble and total carbohydrates, soluble and total proteins, TS, VS, pH, heavy metals, detailed VFA). They showed a strong positive correlation is established with VFA, carbohydrates and proteins whereas soluble organic matter is not influential for this kind of sludge.

READ  The unconditional and seasonal aspect of cash transfers for the prevention of acute malnutrition

Indirect correlations

Correlations between aerobic activity tests and anaerobic tests such as BMP are often proposed. Aerobic tests are less time consuming than anaerobic ones and they are easier from a practical point of view (e.g. no need of anaerobic conditions).
Cossu et al. (2008) showed a good correlation (r² = 0.80) between respiration index (RI4) (mgO2/gTS), which represents the oxygen consumption cumulated in 4 days (Sapromat® apparatus used), and the biogas produced in 21 days GB21 (Nl/kgTS) on municipal solid waste from landfills. Scaglia et al. (2010) found similar results with a correlation between dynamic respiration index (DRI) and anaerobic biogas potential (equation 2.6) with a regression coefficient of 0.89.
Another kind of commonly established correlations is between the initial reaction rate of the BMP assay and the final production value. Donoso et al. (2010) developed an experimental procedure to estimate kinetic parameters from sewage sludge based on the initial reaction rate method. Batch experiments were performed for 3 to 4 days and methane production was monitored. The maximal slope (linear regression) represents the initial reaction rate. S/X ratio is also investigated in order to evaluate the specific effect of the substrate. The optimum ratio went from 0.51 to 1.11 gVSFed gVSInoculum-1. The set of data of initial methane production rate at different initial substrate concentrations was used to estimate the maximal production rate of methane and the affinity constant. An optimization of the experimental data with the simulated data was performed. Authors succeeded in predicting methane production with Monod kinetics. However, the simplified model did not allow accounting for overloads, temperature, inhibitions on continuous digesters modelling and the model underestimated CH4 production by 20% with the parameters obtained in batch tests. Moreover, the inoculum adaptation to the substrate is crucial for this kind of analysis.

Table of contents :

I.1. Statement in methane production prediction from municipal wastewater sludge
I.1.1. Municipal wastewater treatment sludge: definition and composition
I.1.2. Predicting methane production: the analytical way
I.1.3. Predicting methane production: predicting tools
I.1.3.1. Static models
I.1.3.2. Dynamic models and evolution of substrate complexity
I.1.3.3. ADM1 and influent characterization
I.2. Critical review
I.3. Advanced techniques for organic matter characterization
I.3.1. Near infrared reflectance spectroscopy (NIRS)
I.3.2. 3D Excitation Emission fluorescence spectroscopy
I.4. Conclusions and perspectives
I.5. Problematic definition and scientific strategy
II.1. Sludge characterization: analytical methods
II.1.1. Total organic matter analysis
II.1.1.1. Total solids and volatile solids
II.1.1.2. Chemical Oxygen Demand
II.1.1.3. Total carbon analysis
II.1.1.4. Nitrogen analysis
II.1.2. Biochemical characterization
II.2. Biodegradability and bioaccessibility : definition of quantitative variables
II.2.1. Biochemical Methane Potential tests
II.2.2. Interpretation and calculation of BMP
II.2.3. Interpretation and calculation of XRC/XSC
II.3. Sludge samples
II.4. Chemical sequential extraction protocol
II.4.1. Definitions
II.4.2. Sequential extraction Protocol
II.4.2.1. Laboratory material
II.4.2.2. DOM
II.4.2.3. S-EPS
II.4.2.4. REPS
II.4.2.5. HSL
II.4.2.6. COD mass balance and organic matter extraction yield calculations
II.5. 3D-EEM fluorescence spectroscopy
II.5.1. Fluorescence Spectrometer
II.5.2. Fluorescence Spectra
II.5.3. Dilution and linearity for quantification
II.5.4. Spectra interpretation
II.6. Anaerobic digestion laboratory scale reactors
II.7. Mathematical modeling : modified ADM1 and statistical methods
II.7.1. Modified ADM1
II.7.2. Input variables modifications
II.7.3. Kinetic modifications
II.7.4. Liquid/Gas transfer modification
II.7.5. Modified ADM1 input implementation
II.8. Statistical tools
II.8.1. Partial Least Square Regression
II.8.1.1. Definition
II.8.1.2. Interpretation
II.8.2. Other statistical tests
II.9. Conclusion
III.1. Preliminary results : Biochemical methane potential (BMP) test and sequential extractions
III.1.1. Biochemical Methane Potential: S/X ratio investigation
III.1.2. Sequential extraction and sludge profile
III.1.2.1. Validation extractions number
III.1.2.2. Fractions extraction repartition and sludge type
III.1.2.3. Effect of size particle distribution on chemical extractions protocol
III.1.3. Biochemical nature of sludge and extracted organic matter
III.1.3.1. Non-fractionated sludge characterization
III.1.3.2. Fractionated sludge characterization
III.1.4. Sludge fractionation conclusion
III.2. Correlation between chemical and biochemical accessibility investigation
III.2.1. Material flow investigation : anaerobic stabilization test
III.2.2. Biodegradability and bioaccessibility investigation of sequential extraction fractions
III.2.3. Methane production curve and correlation of fractions extracted
III.3. Conclusions
IV.1.1. Fluorescence spectroscopy and organic matter complexity
IV.1.1.1. Sequential extractions fractions fluorescence
IV.1.1.2. Evolution of fractions during anaerobic treatment
IV.1.1.3. Thermally treated sludge
IV.2. Definition of indicators from sequential extractions coupled with 3D-EEM liquid phase fluorescence spectroscopy (3D-SE-LPF) results
IV.2.1. General complexity indicator
IV.2.2. Zone-specific biodegradability indicator
IV.3. Correlations between 3D-SE-LPF indicators and biodegradability
IV.3.1. Exploratory PLS regressions
IV.3.2. PLS regression model set up for biodegradability prediction
IV.3.2.1. Calibration and validation datasets
IV.3.2.2. Validation results of PLS regression
IV.4. Correlations between 3D-SE-LPF indicators and bioaccessibility
IV.4.1. Exploratory PLS regressions
IV.4.2. Validation PLS regression for XRC prediction
IV.5. Identification of recalcitrant molecules to biodegradation: sensitivity analysis
IV.5.1. Sensitivity analysis of PLS models: definition
IV.5.1.1. Sensitivity analysis of PLS models: fractionation variables
IV.5.1.2. Sensitivity analysis of PLS models: fluorescence zones variables
IV.5.1.3. Sensitivity analysis of PLS models: scenario analysis
IV.6. Conclusions
V.1. Continuous lab pilots performances
V.1.1. Reference period
V.1.2. Disturbing period
V.2. Modified ADM1 modeling
V.2.1. Modeling procedure
V.2.2. Input implementation and parameters
V.2.3. Model calibration
V.2.3.1. Steady state calibration
V.2.3.2. Dynamic state calibration
V.3. Model validation
V.3.1. Dynamic validation with disturbing data
V.3.1.1. Input data during both references and disturbing period
V.3.1.2. Dynamic validation
V.3.1.3. Sensitivity analysis
V.4. Bioaccessibility variables and impact on hydraulic retention time
V.4.1. Scenarii analysis
V.4.2. Correlations found between bioaccessibility variables and HRT
V.5. Conclusions and discussion


Related Posts