Abstract

In this dissertation, the application of mechanistic and data-driven models in nitrogen removal systems including nitrification and deammonification processes was evaluated. In particular, the influential parameters on the activity of the Nitrospira activity were assessed using response surface methodology (RSM). Various long-term biomass washout experiments were operated in two parallel sequencing batch reactor (SBR) with a different temperature, aeration mode and substrates and were used for the modelling data source. In the next step, three extension models for the complete ammonia oxidation (comammox) process were developed using the GPS-X simulation software. The extensions were incorporated in the conventional two-step nitrification model. The developed comammox model accurately predicted nitrogen species, biomass concentrations and microbiological indexes. In addition, the contribution of the comammox in nitrogen conversion was generated using Sankey graphs under different operational conditions. Moreover, prediction of the N2O emission in the liquid phase during the nitrification systems was evaluated using hybrid mechanistic/machine learning (ML) method (GPS-X and python programming). In addition, various feature selections (FS) was applied to figure out the effective factors on the production of the N2O emission in the SBR nitrification systems. Finally, a model-based optimization of aeration was performed using GPS-X on the mainstream deammonification system was carried out, in which the DO value and on/off ratio were the variables and N removal rate (NRR) and N removal efficiency (NRE) were the target of optimization.

Author (1)