Using Mathematical Modeling to Identify Causes of Souring During Food Waste Anaerobic Co-Digestion

Abstract Anaerobic co-digestion (AcoD) offers an opportunity to treat food waste (FW) using extra capacity of municipal digesters but is susceptible to reactor souring. We develop a novel model to explore the conditions where FW AcoD reactors sour based on varying organic loading, hydraulic retention time (HRT), and feeding mode (continuous vs. semi-batch). The model includes typical biomass present in AcoD, common FW components (carbohydrates, proteins, lipids) and their hydrolysis products, pH and chemical speciation, alkalinity and pH, and select fatty acid inhibitions. With a long enough HRT and especially with continuous feeding, increasing food waste loading improved pH stability, biogas yield, and methanogen accumulation. AcoD operated with semi-continuous feeding was more sensitive to lower HRT than AcoD with continuous feeding. For example, feeding food waste having 200 g/L TCOD required a minimum HRT of 10 days for continuous feeding, but 20 days for 2-days in the semi-continuous mode. The best indicators of the onset of souring were bicarbonate total alkalinity below 1500 mg CaCO3/L and VFA greater than 500 mg/L as acetate. The model and its results will help designers and operators determine and monitor crucial parameters for industrial application of AcoD. Introduction: The Food and Agriculture Organization (FAO) reports 130 million tons of food waste (FW) are produced annually, accounting for one-third of the global food production (Gustavsson, Cederberg, Sonesson, Otterdijk, & Mcybeck, 2011). In anaerobic co-digestion (AcoD), FW is digested with municipal sludge, increasing energy production as methane (CH4) and reducing materials sent to landfills. However, the steady operation is a big challenge for FW co-digestion plants, restricted by the OLR and hydraulic retention time (HRT). In this work, we utilize mathematical modeling to explore causes of AcoD reactor souring (i.e., pH < 6.0, washout of methanogens, and no CH4 production) taking into consideration FW loadings, HRT, and feeding mode. Material and Methods: Our food waste co-digestion model (FWDM) was based on models by Young et al. (2013) and Batstone et al. (2002), as illustrated in Figure 1. Reactor simulations first involved feeding thickened wastewater sludge (ThS) alone to establish steady state conditions. Then FW was co-fed at the same volumetric flow rate as ThS. ThS and FW composition were based on Young et al. (2013) and Kupferer (2020). The ThS feed has 50 g TCOD/L, 43 g/L total solids, and a pH of 6.55. The FW has 50-200 g TCOD/L, 10-32 g/L total solids with 41% carbohydrates, 32% proteins, and 27% lipids, and a pH of 5.2. Chemical speciation and pH were calculated using the proton condition (PC). The model also included inhibition functions for pH, VFAs, and H2. Operational parameters including flow rates, HRT, and volumes of the digester and gas phase were based on conditions for AD reactors we used to evaluate AcoD in the lab. The model was implemented in MATLAB using an ordinary differential equation solver, ODE15s. Mass balances on each element in the system achieved < 0.00001% error. Results and Discussion: We explored the impact of FW organic input concentrations from 50-200 g TCOD/L, HRTs ranging from 8 to 30 days, and system operation mode (continuous or semi-continuous). For system operation, the reactors were simulated as continuously fed, fed once per day instantaneously, fed once every two days instantaneously, and fed for 8 hours per day. Similar trends for souring were observed for a wide range of operating conditions. In brief, souring was directly caused by depletion of bicarbonate alkalinity with the addition of FW. As an example, we present simulations with reactors being fed semi-continuously every 2 days with ThS and 200 g/L-TCOD FW at HRTs ranging from 15 to 30 d. With addition of FW, the organic loading increased to 19 and 21 g TCOD/L-d in the 18-d and 20-d (HRT) reactors, respectively. Figure 2a shows that the 18-d HRT reactor soured (< 6.0) at about 47 d after FW feeding began, but the 20-d reactor remained maintained a pH of 6.5-7.0 for the duration of the simulation. The added food waste and rapid hydrolysis of carbohydrates and proteins in it caused a rapid increase in fermenter and methanogenic concentration in both reactors (Figure 2b); consequently, the fermenting bacteria produce higher concentrations of VFAs, mostly acetate (Figure 2c). VFAs averaged 1300 mg COD/L in the 18-d reactor between feedings, versus 400 mg/L in the 20-d HRT. At the same time, bacteria consumed NH4+ and HCO3- for metabolic purposes, reducing HCO3- concentration in the 18-d reactor from an average of 3000 mg CaCO3/L to 830 mg CaCO3/L between feedings; for comparison, the 20-d HRT reactor showed HCO3- from 3200 to 2200 mg CaCO3/L between feedings. The combination of significant reduction in NH4+ and HCO3- buffering capacity and the formation of significant amounts of VFAs depleted system alkalinity in the 18-d HRT reactor, causing the pH to decrease to < 6.0 on day 47 after FW feeding began. Our results also showed (not in a figure) that continuous feeding avoided souring at HRTs as low as 10-d HRT, because spreading out the feeding process prevented fermenters from rapidly fermenting carbohydrates and proteins to VFAs that destroyed HCO3- alkalinity. Conclusions: Our modeling predicts that souring occurs when HCO3- alkalinity decreases to <1500 mgCaCO3/L with a commensurate increase in VFAs to >500 mg CaCO3/L. These parameters must be monitored as predictive indicators for FW AcoD souring. Our results also show that continuous feeding avoided souring making, it an operational advantage over feeding semi-continuously for several hours every one or two days.