Changing adm1init_ADMtest changes the initial values. This change will not affect the ultimate outcome of your simulation - it might, if the ADM1 is a 'chaotic' system, but I haven't seen anything to this effect.

Changing the initial values can have two useful effects: it can speed up getting to the dynamic euilibrium; and it can improve the ease with which the integration algorithm can start the simulation. But usually you leave the initial conditions alone.

What you normally want to change are:

PHysical parameters - also held in the init file;
Calibration parameters - ditto

and

the sludge flow & concentration, held in the MAT file.

In my MAT file ALL values are constant with time. The only value that changes is the 1st variable, which is TIME, not S_Su. The value of S_su is the second variable, and has the value in the file of 0.01.

In other words, the MAT file has a CONSTANT set of conditions, defined at time 0 and time 10,000. I have not looked at how the model implements influent handling - does it linear interplation or does it hold the values constant until the next time change. Either way, with the defualt MAT file, the result will be the same. I think that it holds the value constant until the next time change, but, as I wrote, I have not looked at this.

Initial values are in the adm1init_ADMtest.m file.

Input (i.e. sludge feed, not initial values) are in input_ADMtest.mat

To change the digester temperature you want to change the value for T_op, not T_D.

The value for Q_D will set the flow for the period just before the start of the simulation; mathematically, at t = 0-. As soon as the simulation begins, Q_D is dropped for the Q value read from the input data file, and appllies for t = 0+ and beyond.

When a heat balance is added to the ADM1 model then T_D will be of use, and will set the initial temperature; at the moment it is ignored completely. The temperature specified in the influent file is currently ignored.

If you are operating at a temperature other than 35C you need to adjust the model parameters to compensate for the temperature.

If you wish to have temperature vary then you need, as a minimum, a temperature-varying function for kinetic and solubility parameters. At present this appears to be the 'state of the art'.
A functional form for growth parameters of mu(T) = a exp(-b T) - c exp(-d T) is common; and for death, kd(T) = f exp(-g T) should be acceptable. But this would mean writing your own version, or, at least, modifying and recompiling the C code that you get from Ulf.

If you do make such changes Ulf would appreciate being sent them.

Initial conditions don't affect the final result.

But they do affect how quickly you stabilise; they do affect the speed of the simulation at the start; they can affect being able to start in the first place.

Many simulations assume a sterile feed - not knowing what bioamss is present in the sludge/wastewater, set it to zero, and rely entriely upon the performance of the tank to maintain a healthy biomass concentration. But then the initial conditions for the biomass must ensure that you have enough to grow, and to avoid loss while stabilising.

adm1init_ADMtest.m   ---> refers to initial reactor environment and
input_ADMtest.mat    ----> refers to wastewater which is newly introduced to reactor

Yes.

So, in other words if my tank is empty and I introduce for a first time wastewater it will take some time, if ever, to grow biomass a.k.a initial conditions (adm1init_ADMtest.m   concentration etc  = 0 ).

While, in another situation a reactor is already running and I just alter the wastewater quality (when I have initial conditions of the reactor adm1init_ADMtest.m   and new wastewater input_ADMtest.mat)

Yes.

Hi all,
I have some problems with the implementation of ADM1 and I found your forum very very useful so I registered. Thank you for all the posts in the first place, they have helped me to understand Ulf's implementation better. I have a question: I am trying to use ADM1 for another size of digester and feed rate so I have changes Vliquid, Vgas and Q_D in "adm1init_ADMtest" and column 28 in the vector "input" that is created in the variable list in Matlab when you load "input_ADMtest.mat". What I wanted to see is what would happen if I treat a sludge with the same characteristics of Ulf's but with another size of digester and feed rate.

The steady state value seems similar to that found in the literature but the model is producing an enormous peak in the response at the beginning of the simulation (near time 0, approx at 5e-3)which seems to have a numerical origin because it is not the typical biogas peak from quick hydrolysis and I don't know why. As I told you, I've only modified flow rate and volume of liquid and gas and left the rest of Ulf's parameters. Do you know the reason for this?

Also I would like to simulate the reponse of the digester for a year but I have read in Matlab help that the simulation time is in seconds so if instead of 1000 seconds as it is by default in Ulf's file I put 31536000 which are the seconds in a year, I overload the system, as it's logical (or is the simulation time in days and would it be enough to type 365 instead of 1000?) However, I have seen that many other authors (Bajzelj, 2009 "A comparative energy balance model of the anaerobic digestion of domestic food waste both with and without domestic sewage sludge" for example) have simulated the digester response for a whole year and I wonder how...
I've attached images of the two graphs that I obtain.

Thank you very much for your help.

Best regards,

Alba

Yes, in effect I have reduced the working volume, head space and flow rate. I hadn't realized that the initial concentrations were in moles. I'll change that and see what happens, thank you very much for that advice!!

While trying to understand what was wrong, I developed some files to run Ulf's much quicker. They just need to be copied to the same folder as Ulf's files and only by running the "setup" one the rest of the steps are done (load the initial state vector, the influent vector, the simulink program...) so I tried to attach them here with a readme file in case someone finds them useful, but the file is too big and it doesn't let me..if anyone wants them, just let me know. I think it's easier to work with ADM1 with these since all the parameters of the model can be changed in an Excel file I've created, following the format of Zaher et al in their paper about the transformer to ADM1 "GISCOD". Cheers!

Done! I wanted to ask again however, about the issue I had in the simulation. I still get a peak in the response after having set the methane, hydrogen and carbon dioxides in the gas phase to zero. I even set to zero the dissolved methane and hydrogen concentrations. The peak reduces its value (now at around 60 m3/day with a steady state value of 1,6 m3/day more or less) but still appears. I am trying to simulate co-digestion of sludge and food waste and I used Zaher's GISCOD interface and a mass balance to find out the inputs for ADM1. My mass balance is quite simple: I just multiply each one of the parameters given by the transformer to the fraction of the flow. For example, if I have 1 m3 food waste and 2 m3 sludge then I multiply the carbohydrate fraction of food waste by 1/3 and add it to the carbohydrate fraction of sludge times 2/3. I'm telling you this because I'm thinking if it could be an issue of wrong values taken for the influent...Any other ideas?thank you very much!

oh, ok I hadn't thought about using the obtained values as initial conditions, yes, that is what I get too, a peak on biogas production like 4 times higher than the steady state value...I'll definitely try that. And thank you for the two documents, I had already seen them but I will revise them again...I even tried reproducing the graphs on the second one (Bajzelj, 2009) but I didn't get the same form of the curves, that is why I think something is wrong with my simulations...however, when I run it with a sludge with the same characteristics as Rosen and Jeppsson (2006) I don't get that peak, even reducing the digester volume so it must be the initial conditions, as you say. I'll try your suggestion, thanks again for the help!

Oh, no, I wrote an example on how to use the functions in the readme file. You have to pass as an argument the name of the matrix you want to plot, so, for example
>>plot_VFA(tout,DAE2out)
where tout is the vector that saves the time and DAE2out is the matrix that stores the solutions of the system of equations (those two variables are created by Jeppsson's files). In the same way, if you want to use another matrix, like to see what the ODE or the DAE1 system are giving then you just type:
>>plot_VFA(tout,ODEout)
In the case of the biogas production you have to type it like this:
>>plot_biogas_prod(tout,adm_in,DAE2out) because it also calls Jeppsson's function to print the results.
Hope this helps, cheers!

Ok, I think I know why my simulations were so weird.. I thought the mat file was for the initial conditions, I was reading the posts now and realized this. I have corrected the files in case someone is interested. (If you have the old ones, where it says inoculum you should type the characteristics of the influent so you can correct the excel spreadsheet your selfs really). Sorry for that!
Cheers.

Alba wrote on May 31^st, 2012, 4:43am:


Unfortunately were wasnt any readme file in zip, but I got it working as per your suggestion.
Also few remarks, as per matlab syntax, function name should be the same as file names, and in 1st line you dont need semicolon at the end.
function []=plot_VFA(tout,matrix)

I've managed to run simulation without any problem.
Also, since I am running simulation with your input data I cant be sure that the calculated output is correct.

I have benchmark results from Ulf, where a steady input and the expected output, is given. I can provide by e-mail if you ask - I won't be in a position to send until the middle of the next week.

Hi,
Thanks to Alba I have really nice input excel file, which make the whole experience more pleasurable.
My next question would be : How do you fraction and characterize the influent to match model input variables. I am in particular interest of slaughterhouse wastes, but it seems that it will be similar to any kind of waste (food, municipal etc)

Hello

I'm starting to work with the ADM1 model, I've read several articles but have not got the model in Matlab.

I saw the posts that you work with the model provided by Ulf, I was wondering if I could have these files in matlab.

My goal is to do a job optimization and simulations comparing with another model developed here for the treatment of stillage from the alcohol distillation.