Evaluation of thermal steam-explosion key operation factors to optimize biogas production from biological sludge

Sludge production minimization and resource recovery are nowadays priority issues in modern wastewater treatment plants (WWTPs) due to stringent environmental laws on wastewater treatment and sludge disposal routes. In spite of the suitability of anaerobic digestion as the key management and valorization option, the biological nature of sludge, especially secondary, limits both digestion and dewatering. Thermal hydrolysis has proven to optimize the anaerobic digestion of biological sludge by accelerating the rate-limiting hydrolysis step, improving biogas production, volatile solids removal and sludge dewatering, preventing foam formation, and removing pathogens.

However, the reported results on performance are difficult to summarize and compare, varying depending on the sludge source and treatment conditions (Carrère et al. 2010).

According to the reviews by Carrère et al. (2010) and Hii et al. (2014) most of the laboratory- or pilot-scale studies report optimum treatment temperature in the range of 160–180°C, 180 °C being the generally accepted temperature limit to avoid inhibition by formation of refractory compounds at high temperatures (Dwyer et al. 2008; Ariunbaatar et al. 2014).

With regard to the treatment time, the range 30–60 min is mostly accepted, but recent studies (Donoso-Bravo et al. 2010) found that for the pretreatment at 170 °C temperature, the time (ranging from 5 to 30 min) did not influence sludge digestion, suggesting that long operation carried out at full scale could be unnecessary.

Regarding the effect of the flash, none of the bibliographic information analyses whether the re-flashing (repetition of sudden decompression without extra heat consumption) of sludge can enhance even further the solubilization of organic matter, thereby increasing the methane production.

From a full-scale implementation point of view, the available information on the operation conditions of the different thermal hydrolysis commercial processes (Cambi^®, Biothelys^®, Exelys^®, TPH^®, Lysotherm^®, Turbotec^®) shows that the operation conditions are in the generally accepted range 150–230 °C, 20–60 min and one single sudden decompression (flash). As key factors for process design and economics, the revision of these values is a matter of major interest.

This study aims at evaluating different temperature–time-flash combinations (ranging from 110 to 180 °C and 5 to 60 min) on the biochemical methane potential (BMP) of steam-exploded secondary sludge, compared to untreated samples. The results are analyzed in terms of methane yield, kinetic parameters and severity factor.

According to Pérez-Elvira et al. (2008), the study was performed for waste activated sludge (WAS) from the municipal WWTP of Valladolid (Spain), operated at 13 day solids retention time. A single sample of sludge was thickened without polyelectrolyte to 14%TS (79%VS) before being fed to the thermal hydrolysis unit in all the batch experiments performed.

The anaerobic inoculum for the BMP tests was sampled from the anaerobic digester in the WWTP treating mixed sludge, and pre-incubated for 2 days at 35 °C in a thermostated chamber prior to use in order to activate the micro-organisms and to deplete most of the residual organic matter.

The thermal hydrolysis pilot plant operated (Figure 1) consists of a 20 L hydrolysis reactor heated with live steam (12 bar) from a boiler, and connected to an atmospheric flash tank (100 L) by a decompression valve that opens in a steam-explosion effect (sudden decompression). The operation is batch, and automatically controlled by fixing both temperature and hydrolysis time.

To obtain the experimental plan (Table 1), a response surface methodology with Box–Behnken experimental design was used in this work. Three factors at two levels were considered: temperature (110° and 180 °C), time (5 and 50 min) and number of flashes (1 and 3). This last variable was evaluated by re-flashing the treated sludge twice or three times. The experimental plan consisted of 15 runs, including three repetitions at the center point of the experimental design.

BMP assays at 35 °C were conducted in triplicate in 160 mL serum bottles filled with 50 mL of a mixture of anaerobic inoculum and the corresponding substrate (untreated or treated secondary sludge) at a substrate to inoculum ratio of 0.5 g/g (on volatile solids (VS) basis). In this test, micronutrients and macronutrients were used for optimal function of anaerobic micro-organisms. Moreover, NaHCO₃ and Na₂S were added to provide a buffer capacity and avoid aerobic conditions, respectively. The methodology used was the one suggested by Angelidaki et al. (2009).

The bottles were closed with butyl septa, sealed with aluminum caps, purged with helium for 5 min and incubated in a thermostated chamber at 35 °C in an orbital shaker at 150 rpm/min. Methane production in the BMP assays was determined by periodic measurements of pressure and biogas composition in the headspace of the bottles. Reference tests containing only anaerobic inoculum were prepared to determine the endogenous methane production of the inoculum, which was subtracted from the total methane production in the BMP tests to obtain the real methane production of the substrate. The experimental values obtained are always referred to average values, with the corresponding standard deviation.

Table 2 summarizes the target parameters calculated.

It is worth mentioning that both solubilization and degradation factors (SF and DF) are calculated with respect to the particulate fraction of the chemical oxygen demand (COD), in contrast to most of the references that express these parameters with respect to the total COD. These proposed expressions are more accurate as the particulate matter is the potentially hydrolyzable fraction during the pretreatment.

And sludge biodegradability (BD) was calculated as the ratio of the experimental (mL CH₄/gCOD) to the theoretical methane production (350 mLCH₄/gCOD_removed).

Four models were considered to fine-tune the experimental data from BMP tests to theoretical equations in order to estimate kinetic parameters with a certain degree of confidence.

Based on similar studies with solid wastes (Cano Herranz 2014), the models considered were (see Table 3): first order equation (FO), Modified Gompertz (MG) equation, transference function (TF), and logistic function (LF).

Despite differing mathematically from each other, the four models have common features: a kinetic parameter (R_m or μ_max) which indicates the maximum slope of the curve (mL CH₄/gVS/d), a maximum biogas production parameter (P) expressed as mL CH₄/gVS, and a lag-phase parameter (λ), in days. B is the calculated methane production (mL CH₄/gVS) for each time t. The correlation factor (R²) was also calculated to assess the accuracy of each model with respect to the experimental data.

Total solids (TS), VS, total chemical oxygen demand (TCOD) and soluble chemical oxygen demand (SCOD) concentrations were determined according to Standard Methods (Eaton et al. 2005). The soluble phase for SCOD was obtained by centrifugation at 5,000 rpm for 10 min. The pressure in the headspace of the BMP bottles was measured with a pressure sensor PN 5007 (IFM, Germany), and biogas composition was determined using a gas chromatograph coupled with a thermal conductivity detector (Varian CP-3800, USA).

Figure 2 presents the digestion curves obtained in the BMP tests: Figure 2(a), (b) and (c) the influence of temperature and time, whereas Figure 2(d) the influence of re-flashing.

The first evidence of these experimental results obtained is that all the conditions tested presented higher methane production in the final values for day 30 compared to the control (untreated sample), from 20% improvement (at 110 °C) to 40% (at 180 °C). Surprisingly, time was not a relevant parameter, the results for the different temperatures tested being rather similar for pretreatments at 110 and 145 °C (Figure 1(a) and 1(b)). Only at 180 °C did the pretreatment time influence negatively the lag phase (see kinetic parameters in Table 2).

Therefore, thermal steam-explosion at short operation times (5 min) and moderate temperatures (145 °C) seems to be very attractive from a degradation point of view, thus presenting a methane production enhancement similar to the one obtained at 180°C and without negative influence of the lag phase.

Figure 2(d)) clearly exhibits that neither time nor re-flashing influenced the methane production.

The modeling approach was performed for all the samples, showing generally a good accuracy except for the tests with lag-phase. To summarize and focus the discussion, the results presented below correspond only to three samples: the untreated (CONTROL), TH-1 (pretreated, with no lag-phase) and TH-15 (pretreated, with lag-phase).

Table 4 summarizes the results of the estimated parameters obtained with the four models, and Figure 2 presents the model fit to the experimental BMP curves.

Table 4 shows that in most cases the estimated parameters were determined with a high degree of confidence (R² = 0.98). Regarding the maximum methane production, the four models estimate similar values in all the tests. However, some differences can be observed among the different models. FO estimates the micro-organisms' growth velocity (μ_max), which cannot be compared to the maximum methane production rate (R_m), which is estimated by the other models. While TF tends to overestimate this parameter, LF and MG estimate more similar values. Donoso-Bravo et al. (2010) also found this coincidence between LF and MG models.

Regarding the lag-phase, it is only determined by all the tri-parametrical models (MG, TF and LF), but only MG and LF fine-tune correctly this kind of kinetic. Table 4 shows that in the test with 3–4 day lag-phase (TH-15), TF and FO exhibited a poor correlation (R² = 0.948 and 0.943, respectively), showing that these models do not fit lag-phase kinetics. This experimental evidence is very clear in Figure 3, where only MG and LF follow accurately the experimental points in TH-15, and is consistent with the experimental results of Cano Herranz (2014) performed with grease waste.

Comparing the accuracy of the four models, the final conclusion is that the MG equation results to be in general the most appropriate to fine-tune thermal hydrolyzed secondary sludge kinetics, showing an average regression coefficient R² of 0.989. Similar accuracy values were obtained by Donoso-Bravo et al. (2010) and Cano Herranz (2014) with solid substrates and thermally pretreated sludge, respectively.

Most of the studies on thermal hydrolysis of secondary sludge report that it is an effective pretreatment method to improve anaerobic digestion kinetics and methane production from sludge (Wilson & Novak 2009; Oosterhuis et al. 2014; Zhang et al. 2014). However, the quantification of this improvement is difficult to measure by the sole observation of BMP curves and the information that could be extracted can be inaccurate.

Therefore, Table 5 summarizes for the different operation conditions the prediction parameters calculated from the experimental curves (SF, methane, BD and DF), together with the parameters obtained with the application of the MG (which was the most accurate).

As previously stated, thermal pretreatment improved the anaerobic digestion of the sewage sludge evaluated (‘control’: 47% biodegradable, 250 mL CH₄/g VS_fed) by increasing its BD and methane potential for all the conditions tested.

When correlating the main parameters with respect to the severity factor (Figure 4), some interesting behaviors can be observed.

First, the SF increased linearly with the severity as a direct consequence of the cell disruption that takes place during the thermal pretreatment (Dwyer et al. 2008), obtaining a solubilization of 40% of the particulate matter at the highest severity factor evaluated (log R₀ = 4.1 in TH-14&15). The same results were obtained for the DF (maximum 40% degradation of the particulate matter in TH-14&15), thus meaning that all the organic matter solubilized was degraded to methane. Correspondingly, the BD increased linearly, from 47 to 64% at 180 °C–50 min pretreatment (TH-14). This linear link between COD solubilization and methane production is consistent with the results obtained by Carrère et al. (2008).

The lag phase increased dramatically from λ = 0 in the fresh control up to 3.5 days for severity log R₀ > 3 (180 °C, t > 30 min), as previously presented in Figure 1(c)). It is generally considered that the extreme thermal hydrolysis conditions could lead to producing some slowly biodegradable or non-biodegradable recalcitrant compounds (in most TH cases melanoidins), making slower the multiplication of necessary bacteria and decreasing soluble phase consumption (Dwyer et al. 2008; Ariunbaatar et al. 2014).

When comparing the improvement in the methane production (mL CH₄/g VSfed) and in the production rate (mL CH₄/gVSfed.d), it can be observed that although the methane production increased with the severity (as previously commented), the methane production rate did not show the same trend, and exhibited an optimum at a severity factor of 3, then decreasing sharply, again pointing to the fact that the most severe pretreatment could lead to the formation of recalcitrant compounds.

An analysis of variance (ANOVA) was done for methane production in order to test the model significance and suitability. The significance of each coefficient was determined using the F-value test, at a 95% confidence level.

The results of the variance analysis are presented in Table 6. From there it can be concluded that coefficient for the linear effect of the temperature (A) on methane production is a statistically significant model term at 95% confidence level (P < 0.05), thus confirming that only temperature affected the increment of methane production.

Although neither linear effects of the time (B) nor of the flash (C) on the methane production are statistically significant, the linear effect of the time (B) on methane production is more valid than the linear effect of the flash (C).

Based on the model, a maximum methane production 345 mL CH₄/g VS (40% increase) was predicted at conditions: 180 °C, 45 min and 1 flash, with a desirability of 0.882.

This paper assesses through BMP tests of the influence of different temperature–time–flash thermal hydrolysis pretreatment conditions and combinations on the anaerobic degradation of secondary sludge. All the conditions tested (110–180 °C, 10–50 min, 1–3 flashes) presented higher methane production, exhibiting a maximum improvement of 40% solubilization and subsequent degradation of the particulate matter. Generally, the correlation between the severity of the pretreatment and the performance of the subsequent digestion was linear. However, only temperature showed a positive influence on the methane production, although at extreme thermal hydrolysis conditions, the lag phase increased dramatically, probably due to the formation of recalcitrant compounds. Time and re-flashing exhibited no significant influence.