[Add] mixed air in the building

[rubenbaetens]

@damienpicard which Medium do i need to pick for this: The "Incompressible moist air model with constant specific heat capacities" ?

[damienpicard]

I would indeed try with the Annex60 air. Within Annex60, Micheal also investigates the advantages of Buildings.Media.GasesPTDecoupled.MoistAir which might be numerically more efficient.

[rubenbaetens]

Continued in #34.

[damienpicard]

So far, the fluid ports of the ventilation always exist. It would be convenient to remove these equations conditionally to speed up calculations without ventilation.

[rubenbaetens]

@damienpicard i would not make the ports conditional, but their connection to boundary conditions conditional, e.g. by making a IDEAS.Interfaces.HeatingSystm 'none' which connects the mto fixed boundary conditions. Or would this slow the code down too much ?

[Mathadon]

This would introduce more equations but I think the overhead would be limited since the solution does not change. It may be worth trying this? I would add an option for conditional removal.

[rubenbaetens]

@damienpicard @Mathadon i made an issue27_MixedAir branch without conitional ports, but with a "None" ventilation system as is so far generally done in IDEAS. Could both of you review this ? The FixedAirFlowRecup ventilationsystem needs review then though.

[damienpicard]

@rubenbaetens I just reviewed the ventilation system. Important note: for a reason that is not clear to me, using a single source with sizeable ports to connect to the flowports cause an error. You can avoid this by making 1 source for each flowports.

I also renamed the model "none" as I added a parameter to set up a volume change rate (with default value = 0 ).

Finally, I fixed the ventilation with constant efficiency.

[damienpicard]

@rubenbaetens @Mathadon for one of my model, I made a ventilation system which does not use the air medium and could reduce the CPU by 20!

We need to investigate why there are non-linear systems even in the ventilation with zero mass flow rate!

[rubenbaetens]

@damienpicard @Mathadon

• With a default "fixed volume change rate with constant efficiency" you can indeed make most of teh systeem, including a "none"; but it will be my nature reflex then (anyway) to make formally a "none"-model with zero as "final" values to avoid possible errors at all times.
• What do you mean with "which does not use the air medium" ?

[damienpicard]

Ok, I will make an extension.

What I mean with "which does not use the air medium" is that I consider the ventilation only thermally. I do not use any Media then...

If in the future you can show that including a Media slows down the simulation, it might them be interesting to make the mixing volume conditional so that we can also handle the ventilation only thermal.

[rubenbaetens]

• I already made a "none" in the BESTEST directory.
• So, the Media-less implementatio nis actually just the old implementation on the master ?

[damienpicard]

I analyzed further the non-linear systems appearing with the ventilation system.

Here is a plot comparing: 1) None: zero ventilation using the IDEAS ventilation interface (so with electricity) 2) RecupZeroFlow: zero ventilation but with a model with a heat exchanger for recuperation (zero flow) 3) NoneNoElectricity: zero ventilation and no electricity calculations 4) Recup: non-zero ventilation with recuperation 5) NoneNoElectricitySimpleAir: same as NoneNoElectricity but Modelica.Media.Air.SimpleAir is used instead of IDEAS.Media.Air.

From this and from the statistics I conclude the following: 1) the electricity introduce numerical Jacobians (as we already knew) 2) Using ventilation with recuperator and zero mass flow rate is not very efficient. So the ventilation None should no be an extension of the general case with recuparator 3) Using the simplified dry-air medium of modelica is more efficient that the complex IDEAS.Media.Air (if no moist air is needed) 4) Even the use of Modelica.Media.Air.SimpleAir(T_min = 273.15-50) introduces non linear equations:

der(structure.gF.vol.dynBal.medium.d) := -287.0512249529787_( structure.gF.vol.dynBal.medium.p_der(structure.gF.vol.dynBal.medium.T))/( 287.0512249529787*structure.gF.vol.dynBal.medium.T)^2;

[rubenbaetens]

@damienpicard what's on the y-axis ? If it is CPU-time, it only indicates an initialisation problem (of half a second) as the gradients of all 5 plots is rather equal.

[damienpicard]

Indeed, CPU time. But the gradients are not the same...

[rubenbaetens]

@damienpicard then i suggest to run an entire year as example ...

[damienpicard]

To illustrate it beter, I runned an on-off ventilation (period = 36000s) for the case of a recuperator, without recup and media from IDEAS, without recup and with Modelica SimpleAir:

[damienpicard]

@rubenbaetens I added some examples for the ventilation systems and made a None and a ConstantAirFlowRecup. For the example I updated the buiding structure example.

I think we can merge that to the A60 branch and continue the discussion on CPU in this issue or a new issue.

[rubenbaetens]

@damienpicard i suggest we don't merge it to Annex 60, but that i'll merge it to issue51_BESTEST and once that is solved, i push my branch to Annex60.

[Mathadon]

The temperature-dependend density is probably an important cause. It may be worth to make a model with constant density for this..

[Mathadon]

Actually using Buildings.Media.GasesConstantDensity.SimpleAir does not give any significant improvement.

[damienpicard]

@rubenbaetens see branch issue27_MixedAir for last update.

[rubenbaetens]

OK, looks ok now and which solves this issue. I moved the remaining CPU-efficient implementation of a MixedAir medium to #108 ...

[Mathadon]

@damienpicard did you check my emails?