How to predict COD- and SS-removal efficiency at different SRT and aeration?

I'm modeling a high-loaded activated sludge plant mainly for COD-removal, operated at low SRT. I'm performing simulations at various SRT in order to predict effluent quality for the real plant. For the simulations, COD and SS-removal efficiency increases with decreasing SRT in the range of 1.2 - 2.0 days - that is the effluent quality improves with decreased SRT. For my full-scale experiments at the real plant, I observe the opposite. The only way I can achive a decreased removal efficiency for the simulations is by changing the parameters for the Takacs sedmintation model each time I lower the SRT. This is however laborious and also I can't accuratly determine sedimentation parameters for each SRT in the full-scale plant.  I would rather like the sedimentation characteristics to adapt to changes in SRT automatically and in a realistic way.

The reason to why the effluent quality improves at lower SRT seems to be due to the fact that the loading in terms of suspended solids and COD decreases as the reactor concentration decreases. The decrease in reactor concentration appear realistic but the sedimentation model does not recognize the change in sludge settling characteristics or reactor composition (particle distribution, different particulate COD-fractions etc). Is there any sedimentation model which  more intimately couples sedimentation characteristics to the sludge composition and conditions of the reactor? My friend have the same problem in predicting effluent quality for different aeration strategies. I've seen examples of others having succesfully simulated the effect on effluent quality at different sludge age using BSM1 in Matlab/Simulink.

Best regards
Per

I cannot advice on the most appropriate settling model for low SRT / loading rates - hopefully other users will be able to assist you on this.

What I can suggest though is a practical solution to the issue you describe, i.e. "[i]how to automatically have the settling properties of the sludge vary with the sludge age[/i]".

What you actually need is to control (via a custom control model) the settling properties based on e.g. SRT.
So, the first thing you need is to implement such control model in MSL.

But, in order for the settling properties to be "controllable", they are to be defined as [b]manipulated variables[/b] (= input interface variables, flagged as [i]manip[/i]) - and this is generally NOT the case, if I recall well: e.g. 'v0', 'v00' and 'r_P' are all [b]parameters[/b] in the Takacs model.
So, you need to modify the base implementation by: i) moving the definitions of these OBJ's to the interface section of the model (and adapting the definition, by e.g. indicating the causality, terminal, etc.); and ii) modifying the equations in which they are used by replacing "parameters.[NAME]" by "interface.[NAME]"

Thank you, this was very interesting!

I have done something similar using the Scenario Analysis function for which I ran several steady state simulations at different SRT set-points (I've implemented a SRT-controller). For each SRT-set point I had different values for the takacs settling parameter f_ns, which I derived from an estimation of my full-scale test. Your option seems more convinient though. The challege is probably to find the right relation between the composition and dynamics of the reactor and the settling parameters..

Very true: you can indeed do sort of the same, by running a Scenario Analysis.
You see, what a Virtual Experiment does is varying a set of model [b]parameters[/b] in-between simulations. In other words, it does [b]automatically[/b] and sequentially, what you would do manually (plus, of course, evaluate the overall outcome of the experiment). And it acts on model parameters, that are otherwise constant during the course of a simulation.
This is very appropriate, if you are to compare the outcome at different sludge ages.

The solution I was suggesting is considerably more complex (it requires that you write some MSL) but would allow you to evaluate [b]dynamically[/b] the effect of varying sludge ages. If this is not necessary, then your solution is certainly a better option.