WindowProperty:FrameAndDivider fails to describe ~75% of frame heat transfer

[sullivj1]

Issue overview

Window frames in WindowProperty:FrameAndDivider do not engage in infrared heat exchange. This leads to two issues:

1. Documentation: that this is the case is buried in the section on the window output variables (Section 1.10.47.39 Surface Window Net Heat Transfer Energy [J] in the version 22.1 reference), with no mention of it as an issue in the main inputs section for WindowProperty:FrameAndDivider as far as I can see. Indeed, the descriptions of the frame internal and external projections actually say that those inputs should affect IR emission and absorption, which would imply the opposite. This creates a problem in terms of transparency.
2. Accuracy: testing this indicates that this simplification may not be a safe one, as infrared heat exchange may be as much as ¾ of the heat losses. Failing to capture them may lead to significant errors in some cases, as well as misrepresenting comparisons of different frame types.

Initial tests were done with a test model that was completely surrounded by shading surfaces in order to eliminate the confounding effects of shading and solar gains, and isolate heat losses. Running this model in Wellington (New Zealand) resulted in the addition of aluminium window frames decreasing the heating load, despite the fact that aluminium window frames should have significantly higher heat loss than the double glazing in the model:

70mm aluminium frame 3,286.93 kWh 140mm (70x2) dividers 3,261.66 kWh No frame or dividers 3,560.13 kWh

This issue had been seen before (https://unmethours.com/question/82668/frame-and-dividers-unexpected-behavior/), but was suggested to be a bug due to not giving the window frames any depth. I suspect however that conclusion was an error due to not removing the sun – the shading from the window frame projections was obscuring the heat loss difference and making it appear like it had been resolved. All testing I have done finds no issue with depthless window frames.

When we look at the heat transfer through the window, we see that the heat loss through the frame is just much lower than that through the glass:

If we focus purely on the convection losses, they all line up. An aluminium frame has the same losses as single glazing, and roughly twice as much as double glazing. The problem is the lack of infrared heat exchange, which when we look at the breakdown of heat losses in the glass, is the majority of the surface heat losses:

To examine the potential impact of this, and explore workarounds, I approximated the frames manually by adding opaque subsurfaces (“doors”) around the windows as illustrated below:

When we look at the resulting heat losses per m2 for the different elements here, we now get a much more sensible result. Theoretically, if we plot the U-values of the walls, glazing, and frame against their normalised heat losses we should get a straight line as this is essentially the definition of a U-value (W/m2.K). When we do this for a “correctly” modelled window this does not happen, as shown on the left (Figure 7). When we manually model the window frame, however, we get the result we’d expect. The shift in frame heat loss is also consistent with the above analysis of heat loss components – the manual modelling produces heat losses almost four times as high, consistent with infrared heat transfer comprising roughly ¾ of the heat losses.

The flip side of this lack of infrared heat exchange appears to be that the window frames also do not respond to the sun the way we would expect. Here are the external surface temperatures of a north facing external wall (southern hemisphere) and the frame of a window on it. The wall, if given some absorptance, will heat up significantly under the sun, and its surface temperature changes significantly based on reflectance. In contrast, the surface temperatures of the window frame are low (similar to the wall with very high reflectance), and do not increase when we paint them black. This may then have implications for cooling loads.

Potential impact To get a feel for the potential impact of this issue on modelling results, I ran the above manual window frame process over a couple of different models I had doing window comparisons.

Looking at an example of a range of different windows in a ~155m2 standalone house, I find very significant differences in predicted energy use:

Comparing EnergyPlus’s predictions of how different windows compare to each other in a simple 10x10m test cell to the predictions of AccuRateNZ (using CSIRO’s Chenath engine), we see that modelling the window frames manually in EnergyPlus lines them up much better with the trends AccuRateNZ predicts. In particular, if we focus on comparing the differences between window frames we see that modelling them normally in E+ using FrameAndDivider results in it predicting significantly smaller differences.

To sum up: it appears to me that an obscure assumption made by WindowProperty:FrameAndDivider objects results in them drastically underestimating the frame heat transfer. Manually modelling the frames as opaque subsurfaces looks like a reasonable workaround that produces results more consistent with comparisons to other tools, how we’d expect the sun to affect surface temperatures, and how we’d expect heat loss to vary between surfaces. It is, unfortunately, more computationally expensive, and to be practical needs the user to have scripts to create the window frames.

Details

Some additional details for this issue (if relevant):

• Platform: Windows 11
• Energyplus version 22.1.0

Checklist

Add to this list or remove from it as applicable. This is a simple templated set of guidelines.

• [ ] Defect file added (list location of defect file here)
• [ ] Ticket added to Pivotal for defect (development team task)
• [ ] Pull request created (the pull request will have additional tasks related to reviewing changes that fix this defect)

[mjwitte]

@sullivj1 Please attach a zip of the idf files used to generate these various comparison.

[sullivj1]

windowframe_tests.zip Here's a set of models with my simple test model. The first set (no sun) is the model completely surrounded by shading in order to focus purely on heat loss through the surfaces. I've included versions with with just glass, with frames, with dividers instead of frames, and with frames manually added as opaque subsurfaces. The main models are double glazing, but I also added a couple of single glazed versions since comparing the heat loss of the glass and aluminium frames can be informative.

The second set (reflectance comps) has the shading removed in order to illustrate the response of the outside surface temperatures to the sun. It's a bunch of different combinations of the external wall and frame surface being either 10% or 90% reflectance.

For normalising the heat loss, the wall is 5.1m2, and a window is 2m2, with 1.6m2 of glass and 0.4m2 of frame.