Anaerobic treatment of food waste in pilot scale

Problems arising from the lack of solid waste management (SWM) have reached alarming proportions and have become a complex challenge, particularly in developing countries. Municipalities are facing serious problems concerning SWM due to the prioritizing government actions to street sweeping and waste collection; also dumps are still being substituted by landfills, and recovering materials or energy from the waste is not even considered. In some European countries, waste disposal in landfill sites is the last option within the hierarchy of SWM (Council Directive 2008/98/EC) and European legislation strongly encourages renewable energy plants based on biogas production by anaerobic digestion (AD).

Considering that over 50% of the municipal solid waste produced in developing countries consists of organic matter, any treatment aiming to use this type of waste must satisfy the principles of the waste management hierarchy: reduction, reuse, recycling, energy recovery, before landfilling. AD is one method of minimizing the consequences of the disposal of organic matter in landfills and dumps.

There is a growing worldwide demand for renewable energy and alternatives to oil, since the use of the biogas produced in replacement of fossil fuels is considered as a clean development mechanism. AD technology can be used for energy production different types of organic waste including: farmwastes; the organic fraction of municipal waste; sewage sludge and commercial organic wastes. This makes AD ideally placed to reduce CO₂ emissions and to contribute to waste management, focusing on the production of renewable energy.

The recovery of bioenergy and bio products from wastes has attracted interest from researchers in recent years (Labatut et al. 2011). Several systems have been developed for the production of methane, hydrogen, electricity, bioplastics, biopesticides, biosurfactants and other value-added products (Li & Yu 2011). Food waste diversion from landfills is a reality in some European countries and is an emerging trend in the United States (U.S.) (Robbins & Sharvelle 2013). However, when food wastes as well as fruit and vegetable wastes are anaerobically digested, sometimes low stability and low efficiency may happen due to its low C/N ratio. Leung & Wang (2016) pointed out the large challenge of using food waste as sole substrate due to its low buffer capacity. Failure in AD of food waste is so common and hard to predict that many researchers are working with early warning indicators for monitoring the process failure; although the indicators suggested by each study are different, as for the same warning parameter there even existed multi-thresholds (Li et al. 2014).

According to Zaman & Reynolds (2015), the size of AD plants varies with local needs, but in developing countries like Bangladesh, India, Nepal and China, small scale plants are quite used in order to produce biogas for cooking purposes. Nevertheless, there is a need to focus on the improvement of domestic scale plants, which could make a significant contribution to sustainable waste management, mainly in developing countries. Biomethanation at small scale is important to minimize transport costs and provides on-site treatment without aesthetic and sanitation issues.

Given this context and in view of contributing to the development and applicability of AD for domestic use, an anaerobic digester (biodigester) was used in the present study to treat food waste from a university canteen. The following aspects were assessed: (i) organic matter and volatile suspended solids efficiency removal; (ii) suitability of the treated effluent for reuse or disposal; and (iii) characteristics of the biogas produced.

The biodigester was fed with low solids substrate, with an average total solid (TS) content of 4.8%. The digester was inoculated with 30 kg of manure and 150 L of synthetic domestic wastewater prepared in the laboratory (De Souza & Foresti 1996).

The food waste used for feeding the digester was taken from the university canteen and consisted of cooked and raw leftovers (fruit, vegetables, etc.). Once collected, the solid waste was ground using an industrial blender (frequency - 50 Hz/power - 0.5 CV/speed - 3,500 rpm) and then diluted. The initial feed ratio was 0.2 kg of ground waste to 10 L of tap water. In order to avoid accumulation of long chain fatty acids and consequent inhibition of digestion process, the organic fraction was diluted before being fed into the digester for stable digestion process. The amount of waste was increased gradually until reach 2 kg of waste to 10 L of tap water. Aiming to avoid additional buffering costs, the gradual load increase was the only startup strategy adopted (Phase I, described below). After stabilization of the digester, the amount of 2 kg of waste was maintained, and the effluent generated by the digester was used for partial replacement of tap water. All these steps resulted in different operational phases:

• Phase I (PI) - Failure (139 days): the digester was fed daily with a gradual increase from 0.2 to 2 kg of waste for every 10 L of tap water. Acidification of the biodigester was observed in this stage.

• Phase II (PII) - Buffering (274 days): supplementation of substrate using NaHCO₃. Feeding was maintained at a rate of 2 kg of waste/10 L of tap water. The buffering rate was reduced gradually according to the following proportions of bicarbonate alkalinity (kg): COD (kg):1:1 (maintained for 71 days); 0.5:1 (maintained for 18 days); 0.25:1 (maintained for 185 days).

• Phase III (PIII) - Balance (139 days): referred to the feeding period with 2 kg of waste/10 L of tap water, without alkalinity supplementation.

• Phase IV (PIV) - Recirculation (168 days): consisted of the addition of effluent overflowed by the digester to replace some of the tap water to form the substrate. At this phase, the anaerobic digester was operated with the following water: effluent proportions to form 10 L of substrate: 6.5:3.5 (maintained for 115 days) and 3:7 (maintained for 53 days).

The digester performance was assessed by monitoring the parameters shown in Table 1, in samples collected from the influent (substrate), biogas and effluent. Analyses were performed in accordance with the methodologies described in the Standard Methods for the Examination of Water and Wastewater (APHA 2005), as well as those described by Dilallo & Albertson (1961).

The biogas generated in the biodigester was characterized in terms of volume (on a daily basis) and composition (at every change of phase). The alkalinity was measured, according to Dilallo & Albertson (1961), twice during PI and weekly at the other phases.

Table 2 contains the characterizing parameters of the influent substrate: TS, volatile solids (VS), and chemical oxygen demand (COD). The large variety between the parameters, particularly COD and solids, resulted in different values of organic load ratio (OLR) applied to the biodigester.

In order to overcome this limitation, some authors relate the use of strategies, which can approach the choice of a proper inoculum (Forster-Carneiro et al. 2007), co-digestion with other wastes like meat residues (Garcia-Peña et al. 2011), changing reactor design (Bouallagui et al. 2005), among others. During or after a failure, remedial measures as alkali addition, feed interruption and mixing with a nitrogen-rich supplement can also be adopted (Jiang et al. 2012). The strategy of gradually increase the OLR in the present work did not result in the biomass adaptation, as previously observed by other authors (Bolzonella et al. 2003).

In this work, the failure was probably consequence of the type of food waste used. Much of the waste consisted of leftover fruit and vegetables that were rapidly degraded and converted into VFA. In general, pH values between 6.5 and 7.5 indicate that the AD process is stable. As the methanogenesis can be inhibited when the substrate is not adequately buffered, the influent was buffered with NaHCO₃ in PII. The failure experienced indicated need of an experienced operator to identify the problem and solve it timely.

After 50 days of buffering, the effluent exhibited a pH of approximately 7.0. Therefore, NaHCO₃ was gradually reduced until a balance was reached, and the pH of the effluent was naturally maintained between 7.0 and 8.0 as a result of the conversion of organic nitrogen to ammonia nitrogen, which is present in organic waste proteins. These results indicate that buffering supplementation and its gradually removal was better than increasing the OLR for the digester startup.

Total alkalinity (TA) of the influent ranged from 108 to 320 mg CaCO₃ L⁻¹ in all phases. Effluent TA ranged from 432 to 632 mg CaCO₃ L⁻¹. Thirty-five days after starting the buffering, the effluent exhibited TA values in the range of 4,000 mg CaCO₃ L⁻¹ (Figure 2(b)). In PIV, the average TA in effluent was 7,300 mg CaCO₃ L⁻¹. Felizola et al. (2006) obtained TA values ranging from 5,400 to 7,100 g CaCO₃ L⁻¹ in a baffled reactor, with a daily organic loading of 15.9 kg m⁻³ d⁻¹ of organic solid waste (80%) and sewage sludge (20%), and an average TS content of 5%.

The VFA/TA ratio from all phases is presented in Table 3. This ratio (average value of 7.41) was quite high in PI, confirming acid production without sufficient alkalinity to buffer. In PII, the ratio decreased significantly (average value of 1.03), indicating that stability has been reached, the acids produced had started to be consumed, and the buffering system had worked well. In PIV, there was practically no accumulation of acids, indicating that the process was well established and the conversion of organic matter was directed to methane production. This does not indicate that the acids were not being produced, but rather that there was no accumulation in PIV. Thus, careful attention must be paid to alkalinity levels during the digestion of food waste.

After the buffering phase, the VFA/TA ratio started to decrease, becoming stable at the end of the process (PIV), with values ranging from 0.27 to 0.43. Ghanimeh et al. (2012) reached stability with a VFA/TA ratio of 0.07 by operating a thermophilic digester. Digesters that operate with VFA/TA values under 0.3 can be considered as stable (Sanchez et al. 2005).

Influent TS content ranged from 13,566 to 47,084 mg L⁻¹, whereas in the effluent, TS content ranged from 4,978 to 18,754 mg L⁻¹ (Figure 3(c)) in all phases. The maximum removal efficiency of COD and TS was 71% and 87% respectively, during PIV. Despite high percentages of removal efficiency for COD and TS, the average concentration of these constituents in the effluent was still high: 9,930 mg COD L⁻¹ and 8,600 mg TS L⁻¹.

The results of ammonia nitrogen analysis from PII to PIV are shown in Table 4. The corresponding values of ammonia nitrogen in the effluent were 239.5, 497 and 438.3 mg N-NH₄ L⁻¹. These data indicate that there was a greater supply of nitrogen to the digester during PIV, which must have contributed to the improvement in digester performance during that phase.

In PIV, the theoretic bicarbonate alkalinity (Equation (1)) from the ammonification of 247.6 mg N (average value of nitrogen converted to ammonium in PIV) was 1,758 mg CaCO₃ L⁻¹. This value changed to 5,655 mg CaCO₃ L⁻¹ when the theoretic bicarbonate alkalinity from weak organic acids salts (Equation (2)) such as acetate (average content in PIV of 2,294 mg L⁻¹) was added. The average value of alkalinity in the digester effluent was 6,574 mg CaCO₃ L⁻¹ in PIV. This was higher than the theoretical value, due to the alkalinity from other sources not considered in these calculations. These results confirm the balance of the system observed in PIV.

Gas production rate (GPR) and specific gas production (SGP) are displayed in Table 5.

Average biogas production was 1.19 m³ per week in PIV. Despite the increase, the volume of biogas generated in PIV (0.17 m³ d⁻¹) is still considered low, since 0.34 to 0.42 m³ d⁻¹ of biogas per person is necessary to cook food (Singh & Sooch 2004). In PIV, the biogas produced had average methane content of 60% and the maximum value reached was 85% on the 500th day. The maximum SGP reached in PIV was 0.76 m³ (kg VS)⁻¹. This value is close to that obtained by other authors: Rao et al. (2000) obtained a total biogas yield of 0.564 m³ (kg VS)⁻¹, with batch digestion of food wastes, in laboratory scale; Pavan et al. (2000) obtained a SGP of approximately 0.6 m³ (kg VS)⁻¹ in a two-phase system with a mesophilic temperature of the hydrolytic reactor and a thermophilic temperature in the methanogenic reactor; Angelidaki et al. (2012) obtained an average biogas yield of between 0.5 and 0.8 m³ (kg SV)⁻¹ when operating continuously stirred tank reactors at 55 °C. According to Cecchi et al. (2003), organic solid waste that comes from kitchens, markets, restaurants, and cafeterias generally exhibits SGP values between 0.67 and 0.89 m³ (kg VS)⁻¹. Therefore, the digestion almost reached maximum biogas production, indicating that the energy from biogas generated in the anaerobic digester can partly contribute to domestic use, reducing costs of waste disposal and energy acquisition.

The operation of the biodigester in 4 phases indicated the possibility of using it to treat waste generated in cafeterias and kitchens, with good potential for biogas generation (SGP = 0.76 m³ (kg SV)⁻¹). Alkalinity buffering was a satisfactory strategy in terms of promoting the biodigester's stability, which failed after 60 days of operation. Stability was reached after 40 days of buffering with sodium bicarbonate at an initial proportion of 1 kg of sodium bicarbonate: 1 kg COD. The best COD removal efficiency (72%) was reached when recycling the effluent generated in the digester. With regards to TS reduction, an efficiency of 87% was achieved in the same phase. Despite satisfactory removal efficiency achieved after recirculation, the effluent generated in the digester still contained a high concentration of COD (9,930 mg O₂ L⁻¹), TS (TS 8,600 mg L⁻¹), and ammonia (474 8 mg L⁻¹), which hinders its reuse and disposal in the environment.

The authors would like to thank FACEPE (Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco – Pernambuco State Foundation for Science and Technology) for the financial support provided, and CNPq (Conselho Nacional de DesenvolvimentoCientífico e Tecnológico – National Council for Science and Technology Development) for the scholarship awarded to the first author.