Service Hot Water Demand: Too High

Geo_curious · November 8, 2022, 3:33am

Hi all,

I have a multi-family building that I am using to compare several different domestic hot water demand models. The LBT dragonfly version is about 7.6 times as large as other sources, ASHRAE norms, Goldner and Price 1994, other simulation tools, etc. This must be a geometry or unit error. I have been staring at this for a few hours without relevant progress. If someone else could give it a second look, I would appreciate it.
building_compare.gh (156.1 KB)

MohammadHamza · February 8, 2023, 7:21pm

Hi @Geo_curious,

I am not sure if you managed to figure out the problem or not.

Is it 7.6x in terms of annual demand or peak load? I suspect it to be the latter, which would be caused by not accounting for simultaneity/diversity of the load (i.e. not everyone consumes the water at the exact same time in the multi-family building).

DF includes a component to help diversify the load profiles called “DiversifyBldg”, which could help solve your issue.

In fact @chris I am surprised to see that this component is not included in any of the example workflows that DF provides. I would advocate that it should / must be included in all of the examples, lest people run into similar problems like the OP.

And while we are on the subject, @chris could you please also point me toward some documentation of the logic/algorithm that this component uses? I find the approach of using Gaussian distribution interesting and would be great if you could share some additional information about it to help understand what is exactly happening to “diversify” the load profiles. Is there a paper specifically on the topic of how this is implemented in UrbanOpt / DF ?

Many thanks for all your inspiring work with LBT and especially DF !

chris · June 30, 2023, 2:10am

Hi @Geo_curious ,

Hopefully, you figured this out but you tried to assign three programs to a single Dragonfly Building:

This basically means that you assigned the “MidriseApartment::Apartment” Program to the entirety of your building, assuming there’s no corridors or any other spaces in the building. The current recommended workflow is to just plug the whole “MidriseApartment” into the Dragonfly component and this will give you a blended program that is better representative of a full building. You also selected the “pre-1980” template, which is going to be far less efficient than more recent templates.

Apart from this, I can say that the underlying assumptions come from ASHRAE, the DoE commercial reference buildings, and the openstudio standards gem so they should align with the sources you cite there.

@MohammadHamza ,

If you are curios about the code used in the diversification process, you can find it here:

github.com

ladybug-tools/honeybee-energy/blob/master/honeybee_energy/programtype.py#L397-L486


      
          def diversify(self, program_count, occupancy_stdev=20, lighting_stdev=20,
                        electric_equip_stdev=20, gas_equip_stdev=20, hot_water_stdev=20,
                        infiltration_stdev=20, schedule_offset=1, timestep=1):
              """Get an array of diversified ProgramTypes derived from this "average" one.
          
          
    This method is useful when attempting to account for the fact that not
              all rooms within a building will be used by occupants according to a
              strict regimen. Some rooms will be used more than expected and others less.
          
          
    This method uses a random number generator and gaussian distribution to
              generate loads that vary about the mean program. Note that the randomly
              generated values can be set to something predictable by using the native
              Python random.seed() method before running this method.
          
          
    In addition to diversifying load values, approximately 2/3 of the schedules
              in the output programs will be offset from the mean by the input
              schedule_offset (1/3 ahead and another 1/3 behind).
          
          
    Args:
                  program_count: An positive integer for the number of diversified programs

This file has been truncated. show original

I have no paper for it because it’s relatively simple and I am not one of the researchers who has published a paper on the topic. Needless to say, if someone who has researched this has any suggestions about the current implementation, I am happy to assist. Also, if anyone wants to make a code contribution of their own method for diversifying loads, I am happy to review PRs and merge them.