Talk:Tech:Multizone ETP Linearization

Massless node reduction
The discussion only presents how to perform series reductions. Parallel node reductions need to be discussed also. In addition, computing the temperatures for delta/wye node configurations needs to be discussed because no combination of series/parallel reductions is possible for these, and they are not solved by the general method. --Dchassin 21:41, 24 December 2012 (UTC) What happens when the massless nodes have a non-zero Q? --Dchassin 17:03, 26 December 2012 (UTC)
 * Added parallel and delta-wye transformations and method to calculate temperature of any newly created nodes for wye configurations. --Dchassin 16:53, 26 December 2012 (UTC)

PD Controls
Someone needs to check that the PD control expression in Step 2 is correct. --Dchassin 05:01, 26 December 2012 (UTC)

Maximum timestep
The method needs to discuss how to calculate the maximum acceptable timestep to remain within a given uncertainty. --Dchassin 05:01, 26 December 2012 (UTC)
 * A brief discussion of this is introduced in the beginning of the Methodology section.

Proportional control gain
The method describes the proportional-differential control gain $$k$$ as being unique for the entire/all HVAC system(s) in the building.
 * 1) Should the PD gains be different for each zone?
 * 2) Should there be two gains, one for the proportional and one for the differential control?

Jcfuller 18:52, 27 December 2012 (UTC)
The methodology example has some inconsistencies:
 * T1 and T2 are not THE air and mass temperature, but an example of one air and mass temperature.
 * I'm confused by Figure 2. Why do you only reduce the configuration by Node 5?  Why not all of the other series or parallel massless components?
 * Series/Parallel/etc.: While it makes some sense to reduce the model complexity if the model becomes very large, I'm not sure of the value of reducing an inherently linear system that is relatively small. Especially when you are going to have to back out the answers eventually anyway.  Won't you eventually need T4?  Typically you only reduce order a linear model if you don't need to know the values of the internal nodes.  But, in this case you will need to them.  With that in mind, it seems like it would be faster and more efficient to directly solve the linear equation in its entirety rather than trying to decipher the best reductions to use.

It would also be helpful to formulate an example of your matrix solutions, rather than just discussing the equations that go into it.