Modelling of Primary Sludge and RAS Fermentation for Carbon Management Optimization at WRRFs

Introduction Conventional wastewater treatment plants are transitioning to integrated resource recovery and upcycling, which requires effective management of biodegradable organic carbon. This involves developing approaches to carbon diversion, such as anaerobic digestion of primary sludge and waste activated sludge for energy and fertilizer (Gavala et al 2003). Alongside the push to enhance resource recovery, many WRRFs are also under pressure to meet strict effluent nutrient (N and P) discharge targets. While new nitrogen management strategies that do not rely on carbon are emerging, the majority of nutrient removal systems still follow a conventional, optimized nitrification-denitrification pathway that relies on biodegradable organic carbon to support denitrification. Moreover, biodegradable organic carbon is crucial for biological phosphorus removal. Years of experience have demonstrated that successful biological nutrient removal depends not only on sufficient organic substrate, but also on the right type of bioavailable carbon supplied at the appropriate location and time for optimal utilization (Wu et al, 2009; Huang et al, 2018). Our research aims to determine the best way to utilize the variety of organic-carbon rich feedstocks available in a WRRF and how fermentation can be used as a platform process to manage the logistics of biodegradable organic carbon supply and demand. To address these questions, we conducted a pilot-scale assessment of the fermentation of a blend of wastewater-derived primary sludge (PS) and waste activated sludge (RAS) from two treatment plants. Plant 1, which only performs BOD removal, has a short solid retention time (SRT) of 2 days, while Plant 2 performs biological nutrient removal (both TN and TP) and has a longer SRT of 10 days. We conducted batch fermentation experiments on different blends of RAS and PS over a two-week period to analyze biomass kinetics and the products formed during fermentation. We used the SUMO simulator software to model each batch experiment, calibrate the two-week experimental data, and analyze the hydrolysis rates and influent fractionation variations based on the feedstock characteristics. Material and Methods Experimental Set-up Pilot-scale fermentation experiments were performed using a purpose built two-tank pilot system (Figure 1). Each bioreactor have a capacity of 75 L (20 gal). The temperature was controlled at 20 degrees C in each bioreactor. Each fermentation experiment was run in duplicate for 14 days. Three blend scenarios of RAS and PS were targeted, for both Plant 1 and Plant 2 feedstocks: 1)RAS only 2)50% RAS and 50% PS in VSS mass 3)100% PS Modelling Data from each batch experiment were modelled using Sumo Dynamita to validate the observed results as well as elucidate the biological processes. Full plant models were configured for Plant 1 and Plant 2 (see Figure 2) and an SBR process unit was used to represent the pilot reactor. Default influent parameters and fractionation for the full plant models were initially used and were tuned in order to match the TCOD, TSS, VSS, sCOD, NHx and OP data. After the initial conditions were matched, the kinetics and biology were calibrated until the experimental curves were fitted. The reduction factor of anaerobic hydrolysis was also tuned from the 0.5 default to the 0.05-0.15 range, as observed in the Water Research Foundation 4975 report. Results and Discussion Figure 3 shows the timeseries of the soluble COD yield, expressed as mass of sCOD produced per unit mass of VSS (mgCOD/gVSS) throughout the experiment for the different RAS and PS blends, for both short SRT combinations (left) and long SRT combinations (right). For each test, the experimental data was simulated on SUMO and the results of the calibrated models are shown in the dashed line. It can be noticed that the sCOD yield is different between the different PS and RAS blend combinations and also between the long and short SRT RAS. Overall, the highest fermentation rate is achieved around 3 days from the start of the experiment for all the tests. Higher sCOD yields are shown in the first week, while the sCOD production largely slows down in the second week of the test. Figure 4, left) shows the 3-day fermentation rate comparison for the different RAS and PS blends, for both short SRT and long SRT RAS biomass. The reduction factor of the anaerobic hydrolysis kinetic parameter was calibrated for each test and varied between the different feedstock blends (Figure 4, right), showing higher hydrolysis rates with higher fermentation rates and higher PS content, as reported by Ozyildiz et al 2023. The average 3 day fermentation rates ranged from 0.25+/-0.1 mg sCOD/gVSS/h for long RAS only to 3.1+/-0.2 mg sCOD/gVSS/h for 50% PS and 50% RAS, long SRT. Overall, the 50/50 ratio was noticed to be the optimum ratio to maximize soluble COD production, with both RAS types, however a higher fermentation rate (around 50% higher) was achieved when longer SRT RAS was combined with PS compared to the short SRT/PS blends. The fermentation performance is driven by the ratio between the concentration of hydrolysing microorganisms, mainly deriving from RAS, and the fermentable substrate, which is more available in primary sludge. Our research outcomes will be of value to utilities aiming to integrate the stringent effluent nutrient (N and P) discharge targets with energy and resource