Veda2.0 Released!


Stochastic properties group and interpolation
#1
Hi! 

I am running a fairly simple long term investment model, which needs to meet a fixed electricity demand of 10 PJ, currently first introduced in 2030, and then increasing linearly from 2030 up to 2050. I have recently been running some 3 stage stochastic scenarios, using the uncertainty options 'Stochastic' and 'SPINES' under the properties groups. 

I have a 2nd stage introducing uncertainty on technology investment costs using S_NCAP_COST starting in 2025 with values being specified every 5 years (2025, 2030, 2035. etc.), and a third stage introducing production uncertainty using S_NCAP_AFS, with values in 2030, 2040, and 2050 (totalling 9 scenarios in the end). The program runs from 2025 to 2050, with milestoneyrs in every 5 years from then (2030, 2035 etc.)

I have found documentation briefly detailing the SPINES option in the 'Stochastic Programming and Tradeoff Analysis in TIMES', down in Appendix A, but I am unsure how to understand specifically the capacity investment decisions happening in SPINES. Since the investments are always equal across all scenarios in SPINES, what exactly does it use to determines what investments are optimal?

To rephrase if needed: I did read that during a stochastic run, the program cannot see the realization of the uncertain parameters ahead of time, so it has to make some investment decisions based only on the information currently available, however, once the parameters are revealed, what are the investment decisions in these periods, in SPINES, based on?

I am somewhat new to TIMES and VEDA, so do bear with me if the question is unclear. If you need anything else from me to make it easier to answer the question, do let me know.

Best regards,
Lucas
Reply
#2
> Since the investments are always equal across all scenarios in SPINES, what exactly does it use to determines what investments are optimal?

Appendix A does clearly refer to recurring uncertainties. So, while the standard multi-stage stochastic option hedges against the modeled uncertainties in the future periods, under SPINES the model hedges against the recurring short-term uncertainties modeled. While under the standard multi-stage stochastic option the investments give the maximum expected surplus under the future uncertainties, under SPINES the investments give the maximum expected surplus under the recurring uncertainties. The recurring uncertainties affecting any future investments are not revealed (resolved) to the investors before the investment decisions, but they are revealed only during the actual operation.
[+] 1 user Likes Antti-L's post
Reply
#3
Thank you very much for your response Antti,

I have a follow up question on Appendix A if that's ok, specifically about the following statement on page 42:

"When modeling these kinds of recurring uncertainties, all the investment decision variables should only have a single state-of-the-world index in all periods, and only the period-specific flow and activity variables should thus be split into the sets of states implied by the event tree."

Would that for example refer to my S_NCAP_COST, as it affects the investment decision variables? I.e. Does that mean that I cannot have 3 states of the world each with their own S_NCAP_COST value in a SPINES run, as these are not short term recurring uncertainties but long term? Or does it simply refer to the fact that it invests the same across all time periods, but still takes the uncertainty of the investment cost into account when doing so.

S_NCAP_AFS is on an hourly DAYNITE timeslice level in comparison to the yearly values of S_NCAP_COST, would that mean that this is actually unsuited for a standard stochastic run, as they are short term uncertainties and not related to some long term development? (Based on the sentence in : "In the multi-stage stochastic mode it is generally assumed that the uncertain parameters are related to some long-term development...")

Again, apologies if the questions are a little 'dense', but reading the documentation hasn't fully made it click for me yet.
Reply
#4
> Does that mean that I cannot have 3 states of the world each with their own S_NCAP_COST value in a SPINES run

Yes.  Capacity variables have only a single state-of-world (SOW) under SPINES, and therefore only the S_NCAP_COST value for SOW=1 would be taken into account, if such is defined.

> S_NCAP_AFS is on an hourly DAYNITE timeslice level in comparison to the yearly values of S_NCAP_COST, would that mean that this is actually unsuited for a standard stochastic run, as they are short term uncertainties and not related to some long term development?

I would say it depends on the modeller. If you consider it meaningful to analyze, for example, optimal strategies for hedging against the volatilities in wind availabilities in the future, then I guess it is suited.  And certainly it would be well suited at least insofar as the capacity investments made before the resolution time are still in operation during the stochastic branches.

Concerning S_NCAP_COST, what one can do with it under multi-stage stochastics is primarily to find an optimal hedging strategy against the uncertainties in the future investment costs. But besides that, S_NCAP_COST is also useful for sensitivity analysis and Monte-Carlo analyses.
[+] 1 user Likes Antti-L's post
Reply
#5
Perfect, thank you so much for your time and help. It's making more sense to me now.
Reply


Possibly Related Threads…
Thread Author Replies Views Last Post
  Optimizing solver-opt files for stochastic runs LucasRM 4 293 13-01-2025, 08:18 PM
Last Post: LucasRM
Question User define commodity group anik 3 1,000 14-03-2024, 06:52 PM
Last Post: AKanudia
  Scenario Group remove VEDA2 Victorn 9 3,769 12-12-2022, 03:52 PM
Last Post: Ravinder
  CPLEX-error with stochastic run [email protected] 1 1,698 07-02-2022, 02:22 AM
Last Post: Antti-L
  Removing Scenario Group marie 1 1,652 15-09-2021, 10:55 AM
Last Post: Ravinder
  Compound Annual growt rate constraint under stochastic mode Mahmoud 7 5,812 07-04-2021, 05:02 PM
Last Post: Mahmoud
  About the stochastic version of TIMES Mahmoud 5 4,907 27-03-2021, 03:52 AM
Last Post: Antti-L
  Empty Group in FLO_SUM/FLO_FUNC/FLO_SHAR canismajoris 19 17,187 28-11-2020, 02:24 AM
Last Post: Antti-L

Forum Jump:


Users browsing this thread: 1 Guest(s)