Improvement on thermostatic valve.

[hscheipe]

The temperature targeted in the simulations of the IDEAS.Fluid.Actuators.Valves.TwoWayTRV block does not correspond with the setpoint setting. The smooth control signal makes that the lower limit of the bandwidth is taken as temperature targeted instead of the setpoint setting itself. As consequence, the comfort is lower than expected.

This phenomenon is the result of this part of the code:
Modelica.Blocks.Sources.RealExpression yExp(y=
IDEAS.Utilities.Math.Functions.smoothHeaviside(
(TSet - T)/P - 0.5,
0.5))
"Smooth control signal"

and can be solved by changing it to this, to make the use of the TwoWayTRV more intuitive and avoid confusion on setpoints:

Modelica.Blocks.Sources.RealExpression yExp(y=
IDEAS.Utilities.Math.Functions.smoothHeaviside(
(TeSet - T)/(P/2),
0.5))
"Smooth control signal"

Further, it would be useful if the setpoint temperature could be entered through a RealInput instead of a fixed parameter to be able to couple this thermostatic valve with the occupant template in an easy manner.

[annadellisola]

Thank you @hscheipe for raising the issue.

The opening characteristic is indeed a Heaviside function based on the temperature difference between the setpoint and the zone temperature, scaled by a proportional band P (in K) that the user can configure.

In the current implementation of the TwoWayTRV, the valve opening becomes zero at T=TSet when the proportional bandwidth is chosen smaller than P<0.5. As a consequence, the system effectively targets a temperature of TSet - P/2 rather than the setpoint itself.
To ensure that the controller actually targets the setpoint, the implementation can be modified as follows:

IDEAS.Utilities.Math.Functions.smoothHeaviside((TSet - T)/P, 0.5))

In the plot below, the current opening characteristic is shown in green, and the proposed modification is shown in red for a setpoint of TSet=20°C.

The change you suggested would also reduce the proportional bandwidth P, but this can already be achieved directly by choosing a smaller value for P.

[annadellisola]

@Mathadon, do you have better insight on the reasoning behind the original implementation?

IDEAS.Utilities.Math.Functions.smoothHeaviside((TSet - T)/P - 0.5, 0.5))

[hscheipe]

@annadellisola thank you for looking into this!
I forgot to mention, but I proposed to divide by P/2 instead of P as commonly a bandwith is chosen around a reference signal, going from (TSet - P/2) to (TSet + P/2), as it is also implemented in e.g. Modelica.Blocks.Logical.OnOffController. Dividing by P is of course also an option, but maybe the naming of the parameter should be adjusted then to avoid confusing, because now the band around the reference actually equals 2P?

[annadellisola]

@annadellisola thank you for looking into this! I forgot to mention, but I proposed to divide by P/2 instead of P as commonly a bandwith is chosen around a reference signal, going from (TSet - P/2) to (TSet + P/2), as it is also implemented in e.g. Modelica.Blocks.Logical.OnOffController. Dividing by P is of course also an option, but maybe the naming of the parameter should be adjusted then to avoid confusing, because now the band around the reference actually equals 2P?

Thank you for the clarification.

I think the confusion comes from how the proportional band P maps to the transition region of smoothHeaviside.

The function smoothHeaviside(x, delta) transitions smoothly from 0 to 1 over approximately x in [-delta, delta].
With the current implementation:
x = (TSet - T)/P - 0.5
and delta = 0.5.

So the transition occurs when:
-0.5 <= (TSet - T)/P - 0.5 <= 0.5
which becomes:
0 <= (TSet - T)/P <= 1
and therefore:
TSet - P <= T <= TSet

So in the current implementation, the proportional band corresponds to the temperature interval [TSet - P, TSet], not a band of width 2P.

This means:

valve approximately fully open at T = TSet - P

valve approximately fully closed at T = TSet

midpoint of the transition at T = TSet - P/2

This is why the effective regulated temperature is shifted below the setpoint.

If instead the implementation is changed to:

IDEAS.Utilities.Math.Functions.smoothHeaviside((TSet - T)/P, 0.5)

then the transition occurs for:
-0.5 <= (TSet - T)/P <= 0.5
which gives:
TSet - P/2 <= T <= TSet + P/2
In that case, the proportional band is centered around the setpoint, and T = TSet corresponds to approximately 50% valve opening.