Drinking water treatment plants are important in modern society. Their purpose is to supply safe drinking water to the communities they serve and thus it is critical that they operate well. Mixing is an important phenomenon in many systems, and it is vital in the efficient and consistent performance of drinking water treatment plants. It has been identified that drinking water treatment plants have issues with mixing (Heinicke, 2003) and the focus of this study is to diagnose the hydrodynamics within the system. The plant of interest is situated in Göteborg, Sweden, where a pilot scaled model has been created for use in modelling the hydrodynamics in the full scale plant.
Investigation and diagnosis of hydrodynamics have, to the author’s knowledge, not yet been performed on drinking water treatment plants and a simple procedure of determining mixing phenomena was sought after. A practical method that quantifies mixing phenomena is the compartmental model. This model has been utilised to model other water systems as this method is not computationally expensive while providing an accurate result. The aim of this thesis is to show that the application compartmental model is suitable in modelling drinking water treatment plants and can successfully diagnose the mixing properties of the system.
The compartmental model program (Jansons and Howes, 2004) used throughout the optimisation routine was altered to accommodate for the non-constant time step displayed in the drinking water treatment plant data. This model utilises the Nelder-Mead simplex method to search for a parameter set that minimises the sum of squares error (SSE) between the model and the experimental concentration values when given a set of initial parameters.
The application compartmental model was found suitable in modelling DWTPs and successfully indicated mixing properties present in the system. The hydrodynamics of the biofilter were determined by the 3 and 4 parameter models and resulted in similar diagnoses, while the mixing phenomena of the sandfilter was modelled and diagnosed using the 4 parameter model.
It was shown that a significant amount of non-ideal flow was present in the biofilter and sandfilter systems. This was shown by the large presence of ZDM, and small amounts of interchange between ZDM and main channel. This indicates that a considerable volume of both filters is not being utilised for mixing.
A number of adaptations to the model should be made to accommodate the application of data of this nature in the future. They include:
• The ability to weight points according to their importance in capturing RTD characteristics;
• Allowing the model to differentiate between a recycle and a short circuit;
• Modelling the tracer input as a CSTR.
As a result of the modelling of the hydrodynamics of two important drinking water treatment plant units, it has been determined that non ideal flow is significant in the system. Further analysis such as flow simulations and velocity experiments could be performed to gain full understanding of the hydrodynamics, but enough information is now known to understand that the efficiency of the system can be improved.