[esp-r] Re: difference between TRNSYS and ESP-r load value

[Jon Hand]

A complex question about the equivalence of TRNSYS and ESP-r....
________________________________________

I have modeled a (50*15*27) 42 zone building in both TRNSYS 16 and ESP-r
11.6 but large differences are coming in cooling load values, for TRNSYS
it is around 1400 GJ annual load but for ESP-r it is around 2200 GJ annual
load, at first I thought I made some mistake in modeling and construction
parameters for both softwares, but I have checked many times and all the
details of geometry, heat transfer coefficients, glazing data,construction
data and weather data are exactly same and I need to come up with an
explanation as to why these differences are coming.
I have seen the ANSI/ASHRAE standard 140-2001 which compares bestcases for
different simulation softwares and I have tried to benchmark it and the
results match with those given in the standard but the problem is as I go
on increasing the zones,increase the size of the model add glazings the
total value of heating and cooling loads vary a lot.
Also I found one peculiar thing that in TRNSYS if I model a room with
three glazing orientations i.e west, east, north-south. I get maximum load
value for east side window and minimum for north-side but if same room I
model in ESP-r I get maximum load value for north-south side I dont why this
difference is there I am using the same weather file.
At last I want to know if some study has been done which gives the
mathematical differences between TRNSYS and ESP-r and tells the effect of
these mathematical differences through an example model, all I know is
ESP-r takes the rigorous heat balance approach and TRNSYS models its long
wave, short wave radiation and convection through star network system.
Sorry to bother you all with a long question
________________________________________

This is indeed a complex question as well as a long question. The mathematical
differences between ESP-r and TRNSYS will be considerable because they come
from such radically different backgrounds, they use very different solution techniques
and they treat the physics of buildings in (probably) very different ways.

The probably is related to how easy it is to look into the application and determine what is
happening inside. If you had FULL access to both applications and could step through
the calculations you might learn enough to identify the primary sources of the differences.

It is curious that less complex models exhibit smaller differences. That would suggest
that there is an issue of scaling to larger models.  In expert user of each of these
tools might be able to comment on how well the models had been designed and
implemented.

First 'red-flag' is that an attempt to benchmark with the complexity of model mentioned
is a substantial task and if possible I would look for a way to benchmark with one or
more less complex models. For example, ESP-r supports most of the BESTEST models
and automates the task of running these benchmarks.  If there were TRNSYS models
of BESTEST that could be accessed and run that might give clues.

There is the possibility that your assumption that both models are matched is not
actually true.  Also there could be errors in the model that you have not yet noticed
which are causing problems.

Are there some performance indicators for the two models which are similar?  Are there
some periods in the year when some performance is similar?

Regards, Jon Hand
cc Michael K.