Dynamic modelling of anaerobic digestion of Fischer-Tropsch reaction water.

Abstract

Fischer-Tropsch Reaction Water (FTRW) is a high organic strength wastewater produced as a by-product in Sasol’s Fischer-Tropsch Reactors. Typically it has an organic load of 18000 mgCOD/L and is highly acidic with a pH of approximately 3.8. It is deficient in nutrients (N and P and other micronutrients). This dissertation deals with the biological and physico-chemical model development of a dynamic anaerobic digestion model, and explores two different approaches to representing the physico-chemical processes that complement and interact with the bioprocesses. The performances of the resultant two dynamic models (ADFTRW1 & AD-FTRW2) were compared in order to assess to what extent the more detailed and rigorous ionic speciation modeling in AD-FTRW2 addressed the shortcomings attributed to the simplified physicochemical modeling in AD-FTRW1. The ionic speciation model used in AD-FTRW2 uses a classic equilibrium formulation along the same lines as in the UCTADM2 model for anaerobic digestion of municipal wastewater sludges (Brouckaert et al., 2010), while AD-FTRW1 uses a simplification of the approach developed by Musvoto et al. (2000) in order to represent short chain fatty acid (SCFA) dissociation and the weak acid base chemistry of the inorganic carbon system. A 44 day extract from a 700 day laboratory-scale dataset (Van Zyl et al. 2008) was used as the basis for comparing the models. During this period the membrane bio-reactor was subjected to varying flow and load conditions. To validate the models, the experimentally measured and model predicted process variables of reactor alkalinity, reactor pH, biogas production and effluent SCFA concentration were compared. It was found that AD-FTRW2 provided superior agreement with pH data, but predictions of alkalinity, gas production rate and effluent short-chain fatty acids were not significantly improved in AD-FTRW2 relative to AD-FTRW1. This outcome was hypothesized since pH is strongly dependent on physico-chemical processes such as ionic interactions in solution and gas exchange which were the components to the models (AD-FTRW1 versus AD-FTRW2) which differed most significantly. Alkalinity, which is also highly influenced by physico-chemical model representations showed substantial improvement however statistical analysis could not show this improvement to be significant. The other two variables that were compared, biogas production and effluent SCFA concentration, displayed very similar agreement with experimental data. These variables depend more on mass balance effects and biological kinetics and were therefore not significantly altered by the more rigorous handling of aqueous chemistry in AD-FTRW2. It was concluded that AD-FTRW2 constitutes an improvement in model predictive power over AD-FTRW1 at a small cost in computing time.