Technologies using the different resolution

[Carlos Mantilla]

Hi!
We are modelling solar thermal heat with storage in the medium-temperature heat industrial sector. The solar technologies and thermal storage are modeled with an hourly resolution at DAYNITE  level across four representative days for each season, each with different availability factors (AFs).
However, these technologies compete with biomass and electricity, which are modeled at an ANNUAL resolution, for supplying industrial medium temperature heat.
We’re trying to understand how the model handles this mix of timeslice resolutions:

• How is hourly heat demand met when some technologies operate at DAYNIGHT resolution and others at ANNUAL resolution?
  


• Which technology does the model prioritize or select first under this setup?
  


• Is it recommended (or necessary) that all competing technologies use the same timeslice resolution for consistency and accurate competition?
  

We’d appreciate insights on best practices or experiences working with mixed timeslice resolutions in TIMES.

Thanks in advance

[Antti-L]

As there have been thus far no responses, perhaps I can share just some basic thoughts mainly from the technical viewpoint, although I am not a true modeling expert.

As to ANNUAL level technologies in the energy end-use sectors, I think they are typically used for those energy services, where each consumer would (in real-world) usually have only a single technology installed for meeting the demand for the energy service, and therefore each technology should follow the typical load profile of the consumer.  Such energy services include especially household and tertiary end-uses (space heating, water heating, cooking, refrigerators etc.).  When modelled on the ANNUAL level in TIMES, regular technologies will indeed produce the energy service according to the predefined load profile, when that is defined for the output commodity by using COM_FR. Storage processes can likewise be modelled with a predefined load profile, if defined as NST and the output commodity has the COM_FR defined and the commodity is tracked on the ANNUAL level.

Industrial demand for medium temperature heat (say 'INDMTH') may perhaps not be best assumed to be such an energy service, because the relevant industries may be large enough, or the industry sites may typically have several consumers of INDMTH, making it suitable for installing several technologies serving the total INDMTH demand on each typical site. In such cases the individual technologies need not be imposed with a pre-defined load curve, but the load profile of each technology can be left optimized. In this case one should therefore usually not define the technologies on the ANNUAL level, but instead, the INDMTH commodity should probably be tracked on DAYNITE / WEEKLY / SEASON level (depending on the demand variability), and the technologies likewise.  In that way the technologies can compete in the heat market while their individual load profiles are optimized.  The aggregate load profile would in that case usually be defined downstream of INDMTH.

[[email protected]]

As there have been thus far no responses, perhaps I can share just some basic thoughts mainly from the technical viewpoint, although I am not a true modeling expert.

As to ANNUAL level technologies in the energy end-use sectors, I think they are typically used for those energy services, where each consumer would (in real-world) usually have only a single technology installed for meeting the demand for the energy service, and therefore each technology should follow the typical load profile of the consumer.  Such energy services include especially household and tertiary end-uses (space heating, water heating, cooking, refrigerators etc.).  When modelled on the ANNUAL level in TIMES, regular technologies will indeed produce the energy service according to the predefined load profile, when that is defined for the output commodity by using COM_FR. Storage processes can likewise be modelled with a predefined load profile, if defined as NST and the output commodity has the COM_FR defined and the commodity is tracked on the ANNUAL level.

Industrial demand for medium temperature heat (say 'INDMTH') may perhaps not be best assumed to be such an energy service, because the relevant industries may be large enough, or the industry sites may typically have several consumers of INDMTH, making it suitable for installing several technologies serving the total INDMTH demand on each typical site. In such cases the individual technologies need not be imposed with a pre-defined load curve, but the load profile of each technology can be left optimized. In this case one should therefore usually not define the technologies on the ANNUAL level, but instead, the INDMTH commodity should probably be tracked on DAYNITE / WEEKLY / SEASON level (depending on the demand variability), and the technologies likewise.  In that way the technologies can compete in the heat market while their individual load profiles are optimized.  The aggregate load profile would in that case usually be defined downstream of INDMTH.

Hi Antti,

I have a follow-up question regarding your previous response. I assume it's generally better to model all these technologies using DAYNITE time slices rather than a mixed or inconsistent time structure, in order to avoid underestimating the competitiveness of each technology. 

However, this approach places greater pressure on upstream flows—especially during peak demand hours. For example, if natural gas mining is modeled with ANNUAL time slices—forced to distribute activity uniformly across seasons and hours—it cannot respond to hourly variation in downstream gas demand. This may trigger dummy flows given the annual up limit of upstreams. On the other hand, modeling all upstream processes with DAYNITE would significantly increase computational burden. Given this trade-off, would it be better to keep upstream flows as SEASON, or do you have other recommendations?

----BELOW IS A FURTHER TEST---

I set all upstream processes to SEASON, but they can't ramp up or down across hours to match the variability in downstream demand. As a result, there's often unused gas or reliance on IMPNRG/IMPDMD, which is understandable. However, what's puzzling is that the seasonal supply doesn’t seem to align with the expected peak demand (the peak hour) multiplied by the total hours in each season [bcz for my understanding, if upstream process A is SEASON level and downstream B is DAYNITE level, the seasonal distribution decision for A should be made based on peak hour demand of B--to avoid penalty of IMPDMD or IMPNRG]. Why is that?

Best,
Xiao

[Antti-L]

Well, if you are asking me, I already described my viewpoints above, and so I tend to disagree with your broad “generally better” statement.  And I also disagree with your statement that modeling with ANNUAL timeslices would force activity to be distributed uniformly across seasons and hours: it doesn’t force such unless the commodity representing the activity is modelled on the DAYNITE level (and without a load curve specification). But I have no doubts that you are the supreme expert with respect to what is best for your own models.

> the seasonal supply doesn’t seem to align with the expected peak demand (the peak hour) multiplied by the total hours in each season [bcz for my understanding, if upstream process A is SEASON level and downstream B is DAYNITE level, the seasonal distribution decision for A should be made based on peak hour demand of B--to avoid penalty of IMPDMD or IMPNRG]. Why is that?

Well, it isn't so in general, it's just in your model. I just tested such with a small test model, and it certainly wasn't so.  The upstream process at SEASON level was even able to ramp up or down across hours to match the variability in downstream demand, when allowed to have that ability.  And when I didn't allow that ability, the process A nonetheless produced enough to align with the peak hour demand, and fully avoided any penalty of IMPDMD or IMPNRG (I didn't even have dummy imports enabled).  So, your conclusion may be valid for your model, but not in general.

[[email protected]]

Well, if you are asking me, I already described my viewpoints above, and so I tend to disagree with your broad “generally better” statement.  And I also disagree with your statement that modeling with ANNUAL timeslices would force activity to be distributed uniformly across seasons and hours: it doesn’t force such unless the commodity representing the activity is modelled on the DAYNITE level (and without a load curve specification). But I have no doubts that you are the supreme expert with respect to what is best for your own models.

> the seasonal supply doesn’t seem to align with the expected peak demand (the peak hour) multiplied by the total hours in each season [bcz for my understanding, if upstream process A is SEASON level and downstream B is DAYNITE level, the seasonal distribution decision for A should be made based on peak hour demand of B--to avoid penalty of IMPDMD or IMPNRG]. Why is that?

Well, it isn't so in general, it's just in your model. I just tested such with a small test model, and it certainly wasn't so.  The upstream process at SEASON level was even able to ramp up or down across hours to match the variability in downstream demand, when allowed to have that ability.  And when I didn't allow that ability, the process A nonetheless produced enough to align with the peak hour demand, and fully avoided any penalty of IMPDMD or IMPNRG (I didn't even have dummy imports enabled).  So, your conclusion may be valid for your model, but not in general.

Hi Antti,


Apologies—I may have described my question inaccurately and also may not have fully grasped your explanation. I've revised my wording below, and I’d greatly appreciate any further clarification you can provide.

Suppose we have two technologies, GAS-SpHeat and ELC-SpHeat, supplying a space heating demand whose load curve is defined using COM_FR (i.e., across different seasons and hours). If GAS-SpHeat is modeled at the seasonal level and ELC-SpHeat at the DAYNITE level, my understanding is that the model will optimize their seasonal activity levels and, additionally, optimize hourly activity for ELC-SpHeat. However, for GAS-SpHeat, the activity across hours will simply be distributed according to COM_FR, rather than optimized hour by hour. Is this interpretation correct?

Furthermore, if the upstream technology like MINGAS (natural gas mining) is also modeled at the seasonal level—and assuming gas is used across multiple downstream demands such as space cooling or industry—does the model aggregates downstream gas use and determine a seasonal-level activity without recognizing hourly variations?

I hope this time I described it right with the help of GPT.

Best,
Xiao