Performance, microbial community evolution and neural network modeling of single-stage nitrogen removal by partial-nitritation/anammox process

Single-stage nitrogen removal by anammox/partial-nitritation (SNAP) process was proposed and explored in a packed-bed-EGSB reactor to treat nitrogen-rich wastewater. With dissolved oxygen (DO) maintained within 0.2-0.5 mg/L, reactor performance and microbial community dynamics were evaluated and reported. To ascertain whether control/prediction of the SNAP process was feasible with mathematical modeling, a novel 3-layered backpropagation-artificial-neural-network-(BANN) was also developed to model nitrogen removal efficiencies. When NLR of 300 gN/m(3).d and DO of < 0.3 mg/L was employed, the SNAP-process demonstrated autotrophic nitrogen removal pathways with NH (4) (+) -N and TN removal of 91.1% and 81.9%, respectively. Microbial community succession revealed by 16S rRNA high-throughput gene-sequencing indicated that Candidatus-Kuenenia-(33.83%), Nitrosomonas-(3.4%) Annatimonadetes_w5-(1.39%), Ignavibacterium-(1.80%), Thiobacillus-(1.33%), and Nitrospira(1.17%) were the most pronounced genera at steady-state. The proposed BANN-model demonstrated high-performance as computational results revealed smaller deviations (+/- 3%) and satisfactory coefficient of determination-(R-2 = 0.989), fractional variance-(FV = 0.0107), and index of agreement-(IA = 0.997). Thus, forecasting the efficiency of a SNAP-process with neural-network modeling was highly feasible.