Extension for district heating systems

[mstriebs]

The SLPs by BDEW lead to erroneous results, when using them for district heating systems. In this publication we provide a set of SLPs to consider the influence of the district heating grid. Maybe this is of interest for the demandlib.

[joroeder]

Hi, thank you for starting this issue. I think this could be a nice contribution. It seems to me like a fast method to derive demand series of DHS based on existing DHS data, that's cool! Do you plan to start a PR? If you need help, please let us know ... we could also get in contact directly, if you want.

Some questions for better understanding of your method:

This method is intended for large "average" DHS with a mix of single-family, multi-family, BTS and industry? There is no input where you can select the share of buildings (or energy share) of the BTS sector or residential buildings or Industry? Do you know the share of the sectors of the DHS networks you analyzed? How huge were the differences between the data you analyzed? This might be helpful to know for the application of your method to other DHS. Maybe, in pure residential DHS, or DHS with much industry, the profiles should be used more carefully then.

The peak load seems to be a sensitive issue and is underestimated in all cases according your results. In your method, it results from the energy demand, the thermal loss, the weather, and the "average" profiles. Maybe, for the design case, the peak load could also be enforced for the design temperature (e.g. -12°C)? The total energy balance would not be largly impacted, but for dimensioning the energy converter, this could be a nice option ... what do you think?

You are saying that the SLP lead to wrong input for DHS - I mean nobody says that they can directly be used as DHS demand at the heat supply ;-) However, I am wondering if you add up the SLPs according to the shares of sectors in your DHS (PHH, BTS + individual industrial profiles (which are difficult to estimate)) and add the thermal losses of the grid, how big are the actual differences? (in your paper you used the SLP of MFH as basis for the case "No transients"?)

In the resulting profiles, are there any steps at midnight at 00:00 am? (I suppose this is negligible, or maybe depends on the method for calculating the daily energy demand?)

I am happy to hear from you!

[mstriebs]

Hey Johannes,
right now, I am quite busy with other obligations, but I could look into starting a PR probably in a couple of weeks.
Regarding your questions:
Unfortunately, I cannot go into more detail about the DHS I investigated, due to confidentiality. However, one of the networks in question is Flensburg for which I only used public data. Guessing from the information on their website and their annual report I'd say they have little to no clients from industry. However, public data in general is scarce in district heating.

Concerning the peak load: There would be different possibilities to set the peak load, if the “correct” peak load is known. Using a simple rule of three, would shift the entire heat supply curve upwards. Depending on how far off you are from the “correct” peak demand, this will lead to significant deviations in the energy balance. I’d rather suggest to check whether a better match could be achieved by a different option to consider the losses, or to if you run your model in capacity expansion mode, rather than simple dispatch, one could implement a capacity constraint to set a minimum value of installed heating capacity.

Concerning the last point using the SLPs Gas as input for DHS. True nobody said they apply, but in the lack of better data, I’d say it’s not farfetched, that it could be used anyway. I agree it would be interesting to see how the share of sectors and up to the final profile vs. using SLPs MFH as a basis. I did not test it. My guess would be that the detailed consideration of BTS-demand should have a higher impact than EFH, simply because I assume that in the larger DHS (i.e. city centers) only few EFH are actually connected.

Yes, in general there are steps. There is a different demand from 23:00 to 00:00 and from 00:00 to 01:00, but the difference is not huge for identical allocation temperatures. Changes in the allocation temperature from one day to the other and subsequently the daily heat demand have a higher impact.

Best regards,
Sebastian

[p-snft]

I am currently thinking how an implementation could be laid out. SLPs are already part of demandlib, thus an approximation of the effects of a heat network could be laid out as a post processing to generic demand profiles. If one happens to have real demand data at the building level, this could then be used instead of the SLPs. (For many locations, rough building models are available as open data, combined with some estimate for the insulation, you would get a demand curve for space heating. Also, smart meters are coming.)