Co-Digestion of Extended Aeration Sewage Sludge with Whey, Grease and Septage: Experimental and Modeling Determination

: The potential of co-digestion mixing thickened secondary sludge (TS) from extended aeration wastewater treatment plant and locally available substrates (whey, grease and septage) has been studied in this work, using three steps. The ﬁrst step was a batch test to determine the biological methane potential (BMP) of different mixtures of the three co-substrates with TS. The second step was carried out with lab-scale reactors (20 L), simulating anaerobic continuous stirred tank reactors, fed by three mixtures of co-substrates that were determined according to the previous step results. Modeling was applied in the third step, using ADM1 as a mechanistic model to help understand the co-digestion process. According to the BMP step, septage used as a co-substrate has a negative effect on performance, and the addition of 10–30% grease or whey would lead to a gain of around 60–70% in the production of methane. The results from the reactor tests did not validate the positive effects observed with the BMP assay but conﬁrmed good biodegradation efﬁciency (>85%). The main purpose of co-digestion in this scenario is to recover energy from waste and efﬂuents that would require even more energy for their treatment. The protein and lipid percentages of particulate biodegradable COD are important variables for digester stability and methane production, as predicted by modeling. The results of simulations with the ADM1 model, adapted to co-digestion, conﬁrmed that this model is a powerful tool to optimize the process of biogas production.


Introduction
The activated sludge process is the one most widely applied to treat domestic wastewater, due to its well-known facility design and operation parameters. Among diverse activated sludge processes, the extended aeration activated sludge system (EAASS) is the most widely employed technique for the treatment of domestic wastewater in France, with more than 60% of domestic wastewater treatment plants (WWTP) and a capacity greater than the 2000 person-equivalent (PE) volume [1]. The extended aeration process is often better at handling organic loading and flow fluctuations and has a relatively low sludge production due to a long sludge age, as there is a greater detention time for the substrate to be assimilated by microorganisms [2]. Anaerobic digestion is an efficient method to reduce and valorize waste activated sludge [3]. In France, 97 WWTP were equipped with anaerobic digestion units in 2018, out of the nearly 20,000 WWTP in France at the time [4]. Almost half of the 97 installations equipped with digesters are activated sludge units with extended aeration. These facilities treat the effluent of about 30 million PE, close to 30% of the total PE in France. The biogas produced was mainly used on-site (boiler) and for the cogeneration of electricity. Sewage sludge from an activated sludge extended aeration treatment is less fermentable than primary sludge and, therefore, has a naturally lower methanogenic power [5]. However, we feel that it is important to characterize the performance achieved in digestion with this kind of sludge.
compounds. The added volume of inoculum was 30 mL/bottle. The dilution of substrates was realized with a buffered nutrient solution (pH 7) [19]. Biogas production has been determined by pressure variation over time using a manometric method with classical thermodynamic conversion in volume. At the beginning and end of the test (end of biogas production), Dry matter (DM), volatile solids (VS), and chemical oxygen demand COD were measured using standardized methods [20]. Biogas composition (CO 2 , N 2 and CH 4 ) was determined by gas chromatography, using a katharometer as a detector (Shimadzu GC-2014, with a Supelco stainless steel column: 60/80 Carboxen 1000, 15ft x 1/8in).
The kinetic parameters of biogas production have been calculated using a modified Gompertz equation, which is useful when the lag phase is present [21]. This equation (Equation (1)), which represents sigmoidal curves, allows us to assess the lag time. To take into account two successive phases that may correspond to the consumption of two substrates sequentially, Equation (1) has been modified (Equation (2)): (1) where Xg is biogas production (Lbiogas gVS −1 or COD −1 ); K i is ultimate biogas production (L biogas gVS −1 or gCOD −1 ); k i is the maximum rate of biogas production (t −1 ); λ i is the lag phase (t). Three parameters are significant for the kinetics of biogas production (Xg): the ultimate potential for the production of biogas (K i ); the maximum rate of biogas production (k i ); and the lag phase (λ i ).
The ultimate potentials were estimated from the average of the last values of the cumulative biogas production curves. The kinetic parameters were estimated using the solver function of the spreadsheet by maximizing the coefficient of determination (R 2 ) between experimental data and calculated ones.
In order to quantify the synergy in anaerobic co-digestion, kinetics parameters obtained for the different mixture assays were normalized by the parameters from the monodigestion test with TS according to Equation (3): where N p represents the normalized parameter; P mix , the parameter value of the BMP co-digestion assay; and P TS , the parameter value of the reference TS BMP assay. Four co-substrates have been tested: extended aeration thickened sludge (TS), grease (G), septage from a septic tank (SP) and whey (W) ( Table 1). The results were compared to a blank (inoculum + acetate or cellulose). Thickened sludge and grease (produced during the pre-treatment for the process) had previously been collected at the local wastewater treatment plant, septage was collected from the septic tanks of individual houses near the local wastewater treatment plant, and the whey came from the cheese factory of the agricultural school of La Motte-Servolex, (Savoie, France). The experimental plan adopted was determined by the proportion of COD, expressed as a percentage of thickened sludge/co-substrate: 90/10; 70/30; 50/50. For each proportion, the test was performed in triplicate. The nominal COD content was equal to 6.4 g for each bottle.
The initial substrate-over-inoculum ratio (S/I), expressed in COD, was equal to 4.

Bioreactor's Test
The reactors were cylindrical tanks with a volume of 15 L and a headspace volume of 5 L ( Figure S1 in Supplementary Materials). Bioreactors were run as continuously stirred tank reactors (CSTR S ) without biomass recycling so that the sludge retention time (SRT) was equal to the hydraulic retention time (HRT).
The HRT was fixed at 15 days, providing a minimum acceptable SRT, according to the results of previous BMP tests and recommendations in the literature [3,22,23]. Reactors were mixed by continuous pumping (2 L/min). Reactors were heated by an external thermostatic water bath, maintained to 35 • C. Sludge was fed at regular intervals every day to help maintain steady-state conditions in the digester.
The substrates and co-substrates had the same origin as for the BMP test. Biogas production was determined by a volumetric pulse meter comprising 2 chambers filled with water in communication, regulated by a solenoid valve. Each pulse corresponded to a volume of 20 mL of water, displaced by the increase in pressure inside the reactor. Biogas production was also determined by measuring the pressure inside the reactor, using an accurate 0.1 mbar pressure gauge.
The pH and temperature were measured continuously using a specific probe inside the reactor.
The VFA contents and the composition of the biogas were measured by gas chromatography, approximately once every ten days. The analysis conditions are detailed in Table S1 (cf. Supplementary Materials).

Statistical Analysis
Statistical analysis was performed using the StatEL®software (ad Science, Paris, France).
Statistical analysis was conducted using parametric (one-way and two-way ANOVA) or non-parametric (Friedman analysis) tests, and the difference between reactors was expressed as significant, at a significance level of p < 0.05. The p-value is the estimated probability of rejecting the null hypothesis (hypothesis of "no difference" of the test when that hypothesis is true).
Post-hoc tests (Newman and Keuls tests) were conducted in order to decide which groups were significantly different from each other, based on the F-test for Ryan's test (ANOVA), and based upon the average rank differences of the groups for the Friedman test.
The non-parametric statistical test was carried out when data did not meet the criteria for a parametric test (normally distributed, equal variance).

Modeling and Simulation
The ADM1 model takes into account two extracellular steps (the disintegration and hydrolysis of substrates) and three steps of microbial anaerobic metabolism (acidogenesis, acetogenesis and methanogenesis). These reactions are carried out by seven groups of microorganisms and three types of enzymes. The model assumes constant stoichiometry coefficients, with the inhibition of the degradation of volatile fatty acids by hydrogen. The model also considers the inhibition of the growth of microorganisms by pH, and the inhibition of acetoclastic methanogenesis by hydrogen and ammonium [19]. The ionic balance and the gas-liquid exchanges are also integrated. This model considers the mortality of bacterial cells and their recycling into biomass.
The ADM1 model has been implemented in AQUASIM 2.0 by the IWA Task Group for the Mathematical Modeling of Anaerobic Digestion Processes. AQUASIM 2.0 is an open-source software developed in 1998 for the identification and simulation of aquatic systems [24]. Furthermore, AQUASIM allows researchers to conduct sensitivity analysis and establish estimation parameters, being divided into variables, processes, compartments and link interfaces.
The modeling approach is based on the material balance, assuming a continuous stirred tank reactor with output levels equal to those at all areas inside the reactor.
For this study, the ADM1 model was implemented in AQUASIM as a differential and algebraic equation (DAE) and used an ideal CSTR compartment.
The model was solved as a system, with the biochemical reactions integrated explicitly as a system of 26 differential equations (7 groups of bacteria and archaea, 19 catalyzing processes), 3 physico-chemical liquid phase processes, 15 equations for the mechanisms of inhibition, and 8 implicit algebraic variables (per reactor). This approach is a tool to overcome the inherent "stiffness" problem, due to the fast hydrogen and pH dynamics (now explicitly algebraically calculated), which results in some numerical solvers encountering difficulty when solving the system of ODEs [17].
The values and the sensitivity of the parameters were determined by statistical fitting of the experimental data, observed in steady-state using AQUASIM 2.0, which is able to perform linear sensitivity analysis with respect to a set of selected parameters and to perform parameter estimation using the weighted least-squares method [24,25].
Our approach to adapting the ADM1 model to co-digestion was to characterize the actual feed mix with the stoichiometric composition of composite particulate chemical oxygen demand (CODp), i.e., carbohydrates (Xch), proteins (Xpr), lipids (Xli), and inerts (Xi). This approach was successful in terms of model predictions [26] and was similar to the input model for fractionation of COD, described by Arnell et al. [27]. As described by Feucken et al. [28], ADM1 was modified by introducing three state variables for fractionated particulate COD (Xch, Xpr, and Xli), and three corresponding hydrolysis processes (firstorder hydrolysis). In AQUASIM, the stoichiometric composition of substrate mix has been determined by repeating the corresponding characterization using dynamic inputs.
The biodegradable part of particulate COD has been considered, based on COD analysis during the experiment. The ultimate methane potential and the hydrolysis parameter (k hyd ) are estimated by fitting to BMP data using the defined model (Equation (1) or (2)).
Specifically, to take into account the mixture composition feeding, the fractionation of COD, based on substrates proportion, has been applied to the inputs. For example, with a TS/W mixture at 70/30, the soluble amino acids input has been calculated from the following equation: (0.7f aa-TS + 0.3f aa-W ). CODs (4) where f aa-TS is the percentage of amino acids in thickened sludge; f aa-W is the percentage of amino acids in whey; and CODs is soluble COD in g/L. The missing input state variables required for ADM1 have been found from a dataset containing the characterization and fractionation of organic substrates, taken from the literature [29].
Certain kinetic rate equations have been modified to take into account the inhibition linked to substrate concentration. Non-competitive substrate inhibition has been modeled using the Andrews-Haldane equation (Equation (5)).
where mu maxX is the degradation rate constant in kg COD kg COD −1 d −1 ; S x is the substrate concentration in kg COD m −3 ; K s is the half-saturation constant in kg COD m −3 ; and K i is the inhibition constant in kg COD m −3 .

Thickened Sludge and Grease Mixtures
The first phase of biogas production, with a very visible plateau, was observed with 100% grease and became less and less visible, depending on the decrease in the proportion of grease in the mixture ( Figure 1A). This behavior has been observed in BMP tests carried out with the solid fat waste, with a comparable S/I ratio [30], or with grease trap sludge mixed with sewage sludge [31]. The presence of 10% grease accelerated the production of biogas and greatly increased the methane potential by a factor of approximately 1.5, compared to thickened sludge alone ( Table 2).
The mixture of thickened sludge and grease in equal parts led to the same biogas production as the fat alone, but with a shorter lag phase. Note that the increase in the ultimate production of biogas was not directly proportional to the increase of the share of grease, which translates as a phenomenon of inhibition.

Thickened Sludge and Septage Mixtures
With more than 10% of sewage matter in the mixture of sludge and septage, a very strong negative effect on the production of biogas was observed ( Figure 1B). The mixture with 30% of septage led to an inhibition of about 36% of the potential of the sludge alone, and of about 60% for the mixture with 50% of septage.
With 10% septage, there was a positive effect on the first phase of production but not for the second phase ( Table 2). As modeling two-phase kinetics is not adequate to describe the shape of the curve exactly, a kinetic with three successive phases was tested, the last with a lag phase of about 25 days. Table 2. Kinetic parameters and biogas production rates per removed VS of BMP assays.

Thickened Sludge and Whey Mixtures
The mixtures of 10% and 30% whey accelerated and increased the potential for biogas production per gram of VS ( Figure 1C) compared to sludge alone.
The equal-part mixing led to the same potential as the sludge alone, but the second phase started later, with a latency time of more than 14 days ( Table 2). The production of biogas was lower with pure whey as the substrate.
Modeling by the modified Gompertz model was satisfactory, with coefficients of determination between the observed and estimated values of around or above 0.99 (Table 2). This allowed us to quantify the parameters of the specific production kinetics of biogas per gram of volatile matter in the mixtures. The mixtures of 10% grease and 10-30% whey with the sludge thus appear to be very interesting in terms of co-digestion.
The values of kinetic parameters and biogas production per removed VS have been standardized, taking into account the TS test as a reference and applying Equation (3) ( Table 3). A score was determined, corresponding to the calculated average value of NP, taking a positive value for lag phases. This score is used to compare and rank the results. Taking into account the importance of some parameters, and considering the residence time in the digester, the parameter lag phase 2, K 1 and k 1 must be considered as essential, and should be assigned a greater weight, i.e., lag phase 2, K 1 and k 2 multiplied by 2.
Looking at the different scores, the addition of 10-30% grease or 10% whey would be the most preferable. Septage used as a co-substrate had a negative effect.

VS Removal Ratio
Looking at the removal of volatile solids, for all the tests, they were on average higher than 60% and were in line with the observed kinetics ( Figure S2 in Supplementary Materials).

Methane Production
Significant variability in the proportion of methane in the biogas produced was observed for most of the mixtures tested (Table 4). This variability was very low for mixtures with 10% co-substrates for grease and septage.
Grease and whey alone produced low quantities of methane. This was due to an inhibition of the methanogenic flora, and only the hydrolysis-acetogenesis phase was carried out.
The 10% mixture of grease produced more methane than the thickened sludge alone and confirms its significance for co-digestion. These results show the possibility of using grease and whey as substrates for the anaerobic co-digestion of thickened WWTP sludge. The level of 10-30% COD of the total load should not be exceeded, according to the S/I (in VS) ratios tested. These proportions would lead to a higher production of biogas and an increased methane content.
However, it is not advisable to mix more than 10% of septage as a proportion of the total COD load. For this proportion, there would be no increase in biogas production, but the methane concentration would be higher.
The second phase of the study with the CSTR reactors was carried out according to these conclusions concerning used ratios. The co-substrate septage from the septic tank has not been tested in this experiment.

Bioreactor Test
The volumetric loading rates before mixing with the co-substrates were 1.1 ± 0.22 g VS L −1 d −1 and 1.8 g ± 0.7 COD L −1 d −1 . During the period of co-substrate addition, the loading rates were in the same range of values (Table 5). The ratios expressed in COD between the substrate and the co-substrate were nominally fixed in VS at 80/20 for the TS/W mixture and 60/40 for the TS/G mixture.
The COD, VS and DM measurements made it possible to calculate the real average ratios between the substrate and co-substrate. The nominal ratios have been respected, but only for DM. For VS and COD, the ratios are 70/30 for TS/W and are close to 50/50 for TS/G.

Biogas Production
The observed production of biogas was of the same order as expected values for a well-conducted digester with waste activated sludge, i.e., 0.8 to 1.1 L g VS removed −1 d −1 , under the same conditions of loadings [32].
By expressing the daily production of biogas per gram of biomass (VS) in the reactor, the results show that it was the highest for the mixture with whey and the lowest for the mixture with grease. Mean values were, respectively, equal to 0.43 ± 0.11 L d −1 g VS

Qualitative Analysis of Biogas
The proportions of CH 4 and CO 2 were similar between the three feeding conditions ( Table 6). The Kruskal-Wallis test does not allow the H0 hypothesis to be rejected, with the risk of error p < 0.05. No reactor stood out significantly from the others.

VS and COD Removal
Analysis of the data over the entire period of the test showed a smaller reduction of volatiles and COD between the input and output for mixing with grease, while removal was greater than 80% for the other reactors (Table S2 in Supplementary Materials). This difference is statistically highly significant, at p < 0.000001. The dispersion of the data around the mean value was also greater for the mixture with grease, with a coefficient of variation of nearly 30%.
It can be observed that soluble COD had a lower removal yield than particulate COD, as total COD yield was higher than soluble COD.
The VS removal efficiency has evolved linearly over the applied load for TS and TS/W mixtures but is less clear for the TS/G mixture ( Figure S3 in Supplementary Materials).
The mixture with grease was visually inhomogeneous, and deposits of grease were observed in various areas of the reactor; this could explain the poorer efficiency of this mixture for co-digestion.

VFA Measurements
We observed a tendency for the accumulation of volatile fatty acids over time, and more particularly for acetate ( Figure 2). The total VFA concentration between day 35 and day 70 did not exceed 0.6 g L −1 , except for the TS reactor at day 70, due to an accumulation of acetate. The Kruskal-Wallis test does not allow hypothesis H0 to be rejected, at the risk of error of p < 0.05, and there were, therefore, no significant differences.

Biogas Production Per Reactor
One of the greatest areas of interest in modeling is to simulate biogas production, according to variations of feeding, with the estimation of the biogas composition. The comparison of biogas production per reactor between the simulated and experimental values shows that the simulation with thickened sludge presents a satisfactory adjustment (R 2 = 0.704) but includes a slight underestimation. With the whey mixture, the amplitude of the variations observed experimentally was not correctly simulated (R 2 = 0.321), and for the mixture with grease, there is a slight underestimation (R 2 = 0.643) (Figure 3).
The fractionation of the organic matter into its elementary substrate (amino acids, long-chain fatty acids, sugars), adjusted to the different mixtures and consideration of biodegradability of particulate matter, is insufficient to achieve a complete simulation of the real phenomenon. With the grease mixture, Monod's equations were modified by the inhibition equation (Equation (5)) using parameter estimation to determine the value of the inhibition coefficient.
Before the addition of the co-substrate, real biogas production per reactor was slightly higher (statistically different at p < 0.00005) for the reactor with TS_W (Table 7). This higher production was shown in the simulation but was without a statistically significant difference.
The addition of whey as a co-substrate did not influence the average biogas production per reactor significantly, both in the experiment and in the simulation. The addition of grease showed a similar behavior between experiment and simulation. This influence was statistically different at p < 0.00001 for the experiment, and p < 0.005 for the simulation. In the case of the experiment, the average production was lower (9.82 L d −1 VS 11.6 L d −1 for TS) and was also lower regarding the simulation (9.11 L d −1 VS 10.01 L d −1 for TS).

Biogas Composition
The simulated composition of the biogas did not show any effect from the addition of the co-substrate ( Figure S4 in Supplementary Materials). This confirms the observations of the experimental values.

pH
Simulations showed the relative stability of the pH (Figure 4), with correct adjustment to experimental values for thickened sludge, an underestimation for the mixture with grease, and an overestimation for the mixture with whey.
It was not possible to adjust the modeling using realistic parameters that were responsible for the change in pH.

Microbial Degraders and Population Distribution Profiles Obtained from the Simulation with ADM1
Experimental microbial population analysis has not been realized in this study, but it is possible to simulate with ADM1 the growth rate and decay of the different microorganisms involved in anaerobic digestion.
The microbial degrader population distribution obtained from the simulation shows that on each day corresponding to the addition of co-substrates, profiles were similar between the three reactors ( Figure 5). On day 70, at the end of the simulation period for both experiment and simulation, the profiles were different. The main differences were observed for long-chain fatty acid degraders, sugar degraders and acetic acid degraders.

Discussion
Among the process parameters impacting heavily on the quantity and quality of biogas production, co-digestion that is directly linked to feedstock type is one of the main variables [33].
The results obtained with the BMP tests confirmed the positive effects of co-digestion, as has been reported in the literature, with compounds that do not inhibit anaerobic digestion [34,35].
These positive effects, especially those on biogas production and methane, are dependent on their proportions when mixed with thickened extended-aeration sludge. The results of the reactor dynamic test showed less evidence of the positive effects observed with the BMP tests and indicate mainly that co-digestion in this case had no adverse effects.
The feed-degradable COD percentage is very important, as previously shown [27,36,37], and can explain the differences observed in the BMP test and reactor test, especially for less biodegradable substrates like septage and grease.
The protein and lipid percentages of particulate biodegradable COD are important stated variables for digester stability and methane production, as predicted by the simulation [27] and confirmed in this study.
Concerning lipid percentages, laboratory-and pilot-scale anaerobic digesters have shown larger increases in gas production with the addition of readily available highstrength organic wastes, such as fats, oils, and grease (FOG) [34,38]. Co-digestion of mixed municipal sludge (primary and thickened sludge) with concentrated external organic wastes, such as FOG, can result in significantly high methane production at higher organic loading rates, and lower digester HRT/SRT values, as long as COD and VS destruction do not drop below an acceptable lower value [39]. However, the anaerobic digestion of high lipid wastes has been reported to cause the inhibition of acetoclastic and methanogenic bacteria, substrate, and product transport limitation, sludge flotation, digester foaming, blockages of pipes and pumps, and clogging of gas collection and handling systems.
The limit for FOG (grease interceptor waste) addition, leading to an inhibited digestion process, was identified to be between 20 and 40% (v/v) or 65.5 and 83.5% (w/w) of VS as added [40]. This inhibition has been observed with the BMP test and reactor test in this experiment.
Increased FOG levels extended the lag phase and eventually displayed strong or complete inhibition. Operating anaerobic digesters with FOG levels in the concentration range of 0.1-1.5% (v/v) enhanced biomethane production, resulting in a 2-to 19-fold increase. The highest biomethane production was observed at 1% FOG (19-fold increase), with a lag phase of 25 days. LCFAs were degraded by 80-90% during anaerobic digestion, following a 0.5-1.5% FOG loading. The addition of either 2% or 3% FOG permanently inhibited biomethane production, due to high VFA accumulation (17-19 g/L) and low LCFA reduction (29 and 18%), respectively. Under optimum biomethane production (1-1.5% FOG), Syntrophomonas and Fermentimonas were abundant, indicating their role in LCFA degradation and acetogenesis [41].
Concerning whey addition, and dairy wastewater by extension, it is important to assume an organic matter load that is compatible with a CSTR process to avoid any failure of this non-adapted model in a high-strength load process [44,45]. Indeed, whey has a high C/N ratio, a low pH, and is rich in easily degradable sugars (mainly lactose), which could lead to rapid acidification in the digester and, thereby, process inhibition [46,47]. A correlation study was made on a whey/TS mixture reactor, among pH, COD load, and the production of biogas, in an attempt to explain the significant variations observed. We did not obtain any significant correlations between these parameters. The highest level shown is between the pH and the production of biogas, with a coefficient of determination R 2 of only 0.44 (r = 0.66 at p = 0.00001). The pH reduction observed with this reactor seems not due to the accumulation of VFAs from the acidogenesis step, as shown by chemical analysis, but instead is probably because of the weak buffering capacity of the mixture due to acidification, as whey has a low pH. Indeed, if the change in pH reflects the excess VFA, it may also be a sign of the high production of NH 4 (protein hydrolysis). It is therefore necessary to adapt the organic load to the buffering capacity of the reaction medium; the recommended VFA/alkalinity ratio is 0.2-0.5 [48].
To enhance the nutrient balance with improved pH buffering, and to optimize the production of methane, a C/N ratio of about 15-30/1 is required [12,33], and this can be reached by deploying co-digestion. Co-digestion of 10% to 30% whey with thickened sludge has shown a synergistic effect on biogas production in this study, as also observed when mixing with poorly degradable substrates [49,50].
The results of simulations with the ADM1 model adapted to co-digestion confirm that this model is a powerful tool to optimize the process of biogas production. Numerous publications have been noted, indicating that the disintegration and hydrolysis steps are limiting factors of the co-digestion process [51].
To improve ADM1 simulation with whey co-substrate, it may be possible to consider the uptake of lactate and incorporate it in the biochemical process [52]. This does not seem necessary in our case, both because of the simulation results obtained and also an overly low concentration of lactic acid in the digester when the proportions of the mixture of whey is less than 20-30%. However, in the presence of significant lactate concentration, it would be advisable to add the metabolic pathway of the transformation into methane, which can lead to complete conversion [53]: 2 lactate + 3 CO 2 3 acetate 3 CH 4. Therefore, it seems important overall to respect the C/N ratios recommended for the whey/thickened sludge mixture, or other co-substrates that are rich in nitrogen to avoid excess nitrogen causing ammonia formation, together with an increase in pH, and consequently a drop in methane yield or, at the opposite end of the spectrum, nitrogen availability depletion, resulting in incomplete methanogenesis and reduced biogas productivity.
If the results obtained with the BMP test could lead to a significant increase in production, as observed in the literature for dairy whey and grease sludge [54], biomethane and, therefore, energy, the reactor tests did not validate this hypothesis satisfactorily. Process instability was observed in semi-continuous mode, with beverage wastewater as the co-substrate, which was not predicted by the BMP test [55]. BMP can be used to screen an array of substrates but is not suitable for predicting continuous performance.
As extended aeration plants necessitate high operational costs, due to high energy needs, 0.903 ± 0.509 kWh m −3 [56] additional power via co-digestion should not be considered too much to cover the energy needs. However, reactor tests confirmed good biodegradation efficiency, and the main purpose of co-digestion, in this case, is to recover energy from waste and effluents that would require even more energy to treat with this type of extended aeration process.

Conclusions
The results obtained confirm that it is possible to add local organic waste to the anaerobic digesters of an extended aeration wastewater treatment plant to treat the thickened sludge. This addition is even positive, as a way to valorize these effluents into biogas. To identify good practice for the procedure of co-digestion, it is necessary to abide by the designed daily volume or mass loads, as well as the residence time required for good degradation of the co-substrates. Adapted modeling using ADM1 makes it possible to estimate the proportions of the mixture to be applied, based on data from BMP tests from the literature or from specific tests, if necessary. This type of approach, combining batch testing (BMP), dynamic laboratory study, and mechanistic modeling, is to be favored. It can also be easily adapted for the study of pretreatment if necessary. It is also possible with this approach to generate the relevant data to test different technical, environmental and economic scenarios, in order to define competitive processes for industrial adoption.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/su13169199/s1, Table S1: Main characteristics for VFA and biogas analysis; Table S2: VS and COD removal; Figure S1: Diagram of a reactor; Figure S2: VS removal ratios with BMP assays; Figure  S3: VS removed vs VS loads for TS, TS/W and TS/G; Figure S3: Comparison between experiments and modelling for biogas composition.