Pith. sign in

REVIEW 3 major objections 4 minor 9 references

Industrial Flexibility Investment Under Uncertainty: A Multi-Stage Stochastic Framework Considering Energy and Reserve Market Participation

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Reserve market participation raises optimal industrial storage capacity by 11.8% in 2025 and 48.8% in 2050.

desk verdict A coherent multi-stage investment framework for industrial flexibility, but its headline reserve-market gains rest on an unverified technology-qualification assumption and unreleased data. read the letter →

arxiv 2506.08638 v1 pith:2GDWMG24 submitted 2025-06-10 econ.GN q-fin.EC

Amund Norland , Lasse Skare , Ole Jakob Viken , Stian Backe This is my paper
classification econ.GNq-fin.EC MSC 90C1590C90
keywords multi-stagestochasticprogrammingindustrialflexibilitybalancingmarketsmulti-marketparticipationinvestmentdecisionsunderuncertaintyreservemarketthermalenergystorageNorwaycasestudy
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that industrial energy users should treat reserve-market revenue as part of the investment case for flexible storage, and it provides a multi-stage stochastic optimization model for sizing and operating that storage under price uncertainty. The model jointly decides capacity investments, day-ahead and intraday bids, reserve-market bids, and real-time operation, with uncertainty represented by a nodal scenario tree. In a Norwegian industrial case study, allowing reserve-market participation raises optimal storage capacity by 11.8% in 2025 and 48.8% in 2050 compared with energy-only participation. If correct, industrial prosumers who ignore reserve revenue when sizing assets will systematically under-invest in storage as renewable penetration grows.

What carries the argument

The carrying mechanism is a multi-stage stochastic optimization program formulated on a scenario tree, in which uncertainty about prices and operating conditions resolves node by node across stages. Bidding in the day-ahead, intraday, and capacity reserve markets is consolidated into one decision stage, matching the internal deadlines of industrial actors observed in discussions with Norwegian companies, while real-time operation and reserve activation are decided in later stages. Storage sizing is coupled to operation through a power-to-energy ratio $\theta_b$ and state-of-charge dynamics, and technology output is capped by installed plus added capacity scaled by availability factors.

What would settle it

Re-run the Norwegian case with technology-specific reserve-market qualification—for example, excluding thermal energy storage or flywheels from the mFRR market, or imposing minimum bid durations and ramp rates they cannot meet—and compare optimal storage capacity with and without reserve participation. If the 11.8% and 48.8% gaps shrink or reverse, the conclusion that reserve participation justifies larger storage depends on the uniform-qualification assumption.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single multi-stage stochastic programming framework can guide industrial flexibility investment by jointly modeling investment, consolidated market bidding in the day-ahead, intraday, and capacity reserve markets, and real-time operation under uncertainty. Applied to a large Norwegian industrial site with no pre-installed generation, the model finds that reserve-market participation changes the optimal design: total storage capacity increases 11.8% in 2025 and 48.8% in 2050, the preferred storage technology shifts from flywheel to lithium-ion as investment costs fall, thermal energy storage remains attractive across all cases, and rising CO2 costs shift conversion investments from gas boilers toward heat pumps and solar PV. The paper concludes that industries already planning flexible-asset investment should include reserve-market revenue when optimizing storage size.

Load-bearing premise

The load-bearing premise is that every flexible storage technology modeled is qualified to participate in the chosen reserve market, the manual Frequency Restoration Reserve (mFRR) market; if real qualification rules differ by technology, the reserve revenue available to each storage type changes and the optimal sizes and headline percentages would shift.

Editorial extensions

If this is right

  • Industrial prosumers should size storage roughly 11.8% larger in 2025 and 48.8% larger in 2050 when the site can sell reserve capacity, compared with arbitrage-only sizing.
  • In the 2050 scenario the optimal storage technology is lithium-ion battery storage, while flywheel dominates in 2025 as investment costs fall.
  • Thermal energy storage is profitable across all tested market and carbon-price scenarios and is only mildly sensitive to reserve-market participation.
  • Rising CO2 prices and falling investment costs make gas boilers less attractive and heat pumps and PV more attractive by 2050.
  • Consolidating day-ahead, intraday, and capacity-reserve bids into one stage reflects how industrial actors actually submit bids through their balancing responsible party.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Going beyond the paper, the 48.8% reserve-driven storage increase suggests that omitting ancillary-service revenue from investment models could under-size flexibility assets system-wide as renewable penetration rises; the same framework could quantify that bias in other price zones.
  • Because the paper leaves technology-specific prequalification for future work, a natural extension is to decompose the reserve value by storage technology to see whether flywheel's 2025 advantage or lithium-ion's 2050 advantage survives real market rules.
  • Reserve revenue enters the objective through capacity payments and activation costs, so the headline percentages would scale roughly with the level of mFRR prices; testing lower or higher reserve prices would show how sensitive the sizing decision is to assumptions about reserve scarcity.
  • The site-specific load-shifting window (10% up, 30% down, hours 8–17) and the single operational day with seasonal scenario branches are modeling choices; sites with different shiftable load profiles would likely see different technology mixes under the same framework.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a multi-stage stochastic optimization framework for investment decisions in industrial flexibility assets, jointly modeling investment, day-ahead/intraday/reserve bidding, and real-time operation under a nodal representation of uncertainty. A Norwegian industrial site case study compares optimal storage and conversion technology investments in 2025 and 2050, with and without reserve market participation. The headline quantitative results are that reserve market participation increases optimal storage capacity by 11.8% in 2025 and 48.8% in 2050 (Section V), and that the technology mix shifts from flywheel in 2025 to lithium-ion BESS in 2050. The framework itself is internally coherent, and the authors explicitly disclose that all storage technologies are assumed qualified for the chosen reserve market (mFRR).

Significance. If the results hold, the paper contributes a useful modeling template for industrial prosumers who must decide on flexibility investments while facing multiple electricity markets and uncertainty. The model is clearly formulated in the presented constraints, and the case-study conclusions follow from the stated assumptions. The authors also deserve credit for explicitly flagging the reserve-market qualification assumption and for stating what is left to future work. However, the quantitative significance is conditional: the headline storage-capacity gains are not robustly established because they depend on a single technology-qualification assumption and on fixed exogenous parameters, without sensitivity analysis or release of data/code. The paper does not fit parameters and then re-derive them as predictions, so circularity is not an issue; the main weakness is verifiability of the exact numerical claims.

major comments (3)
  1. [Section II and Section V] The central result—that reserve market participation increases optimal storage capacity by 11.8% in 2025 and 48.8% in 2050—rests on the stated assumption in Section II that 'all flexible storage technologies modeled are qualified to participate in the chosen market.' Since Section V selects flywheel as the favored storage technology in 2025, and mFRR has product-specific prequalification requirements such as sustained activation duration and ramp capability, it is not evident that a flywheel can earn mFRR revenue. If flywheel is not qualified, the 2025 headline gain could shrink or disappear. The 2050 result is less exposed because lithium-ion BESS is selected, but the percentage still depends on the same aggregate qualification assumption. Please add a sensitivity case that excludes non-qualifying technologies or introduces technology-specific qualification constraints, and discuss how the 11.8% and 48.8% figures change.
  2. [Section III] The formulation omits multiple constraints 'for brevity'—ramping across stages and the first time step, load-shift initialization/aggregation and permitted time steps, SoC continuity across stages and end-of-horizon, and peak-load propagation from parent to child nodes. Without these constraints, the model is not fully specified. Together with the absence of input data and code, this means the exact quantitative results of Section V cannot be independently verified or replicated. Please include the full formulation in an appendix or make the code and data available, at minimum for the case-study instances.
  3. [Section IV] The case-study results are single-point estimates. Key exogenous parameters—the 9% discount rate, load-shift limits of 10% above and 30% below reference demand, the reserve bid size cap, the intraday liquidity fraction, the technology cost projections from the Danish Energy Agency, and the CO2 price scenarios—are fixed, and no sensitivity analysis is reported. Since the optimal storage-capacity changes of 11.8% and 48.8% are driven by these parameters, the quantitative claims need a sensitivity or scenario analysis (for example, varying the discount rate, reserve price levels, and load-shift flexibility) to establish robustness.
minor comments (4)
  1. [Section III, Eq. (1) and Eq. (3)] The subscript 'w' in I_ID_wt and sigma_ID_wt appears to be a typo for omega; the same notation should be used consistently throughout.
  2. [Section III, Constraint (13)] Constraint (13) is motivated by monthly peak grid import, but the case study in Section IV uses one operational day per season. Please clarify how monthly peaks are evaluated in a single-day setup, or adjust the constraint description.
  3. [Section IV] The role of 'dummy-fuels' is introduced in Section IV but not formally defined in the mathematical formulation; a brief explanation of how they enter the energy balance equations would improve readability.
  4. [Throughout] There are several typographical issues, such as 'trepresents' in Section III and inconsistent spacing in 'L ¨ohndorf' in the Introduction. A thorough proofread is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the reported capacity gains are genuine model outputs driven by exogenously specified costs, prices, and constraints, with no fitted parameters renamed as predictions and no load-bearing self-citations.

full rationale

The paper's central results—notably the 11.8% and 48.8% storage capacity increases with reserve market participation—are obtained by solving a multi-stage stochastic optimization model with and without reserve market participation. Investment costs, load curves, price scenarios, discount rates, and market qualification assumptions are all exogenous inputs stated in Sections II and IV; the storage capacities are optimization outcomes, not quantities fitted to those outcomes. The key qualification assumption ('it is assumed that all flexible storage technologies modeled are qualified to participate in the chosen market') is a disclosed modeling premise and a potential limitation for external validity, but it is not circular: the model does not use the conclusion to define the input, and the reported percentages could change if the assumption were relaxed. The references are external prior work; none of the authors' own prior results are invoked as load-bearing justification. No equation in the paper defines a predicted quantity in terms of itself or renames a fitted parameter as a prediction. Thus there is no significant circularity.

Assumptions & free parameters 7 free parameters · 6 assumptions · 1 invented entities

The core results depend on a set of exogenous scenario assumptions and market simplifications. No parameters are fitted to produce the findings; instead, the headline percentages are outputs of an optimization model driven by these externally sourced inputs and stated modeling choices.

free parameters (7)
  • Discount rate for investment annualization = 9% (stated)
    Used to normalize investment costs over the planning horizon in Section IV; directly affects the relative cost of capital-intensive storage and hence optimal investment sizes.
  • Load shift limits = Up to 10% above and 30% below reference demand during hours 8 to 17
    Defines the temporal flexibility available to the industrial site and constrains how much reserve capacity and arbitrage the model can exploit, as described in Section IV.
  • Reserve bid size cap = X^Max, value not reported
    Upper bound on downward reserve bids in Constraint 8, reflecting indivisible bid sizes; directly limits reserve market revenue and affects storage sizing.
  • Intraday liquidity fraction = sigma_ID, fraction of historical average volumes, value not reported
    Limits intraday trading volume in Constraint 3; affects how much the model can use intraday markets beyond day-ahead positions.
  • Future technology cost projections = From Danish Energy Agency, values not reported in text
    Assumed decline in lithium-ion battery costs drives the 2050 shift from flywheel to BESS, as shown in Figures 2 and 5.
  • CO2 price scenarios = 2025 and 2050 levels not reported in text
    Higher assumed carbon prices drive the shift from gas boilers to heat pumps and PV by 2050, as summarized in Section V.
  • Grid export limit = G_export, value not reported
    Upper bound on electricity export from Constraint 12 based on Norwegian prosumer regulations; affects PV investment potential and overall system design.
assumptions (6)
  • domain assumption All flexible storage technologies modeled are qualified to participate in the chosen reserve market.
    Stated in Section II; this lets every storage asset earn reserve revenue, which directly drives the storage-sizing results. If qualification rules exclude some technologies, the headline percentages change.
  • domain assumption Day-ahead, intraday, and capacity market bidding can be consolidated into a single decision stage.
    Based on conversations with Norwegian industrial actors who follow internal BRP deadlines, as described in Section II. This removes sequential learning between markets from the model.
  • domain assumption A single merged reserve market represents all relevant reserve markets.
    Stated in Section II; multiple reserve markets and technology-specific qualification are left for future work.
  • domain assumption The case study site has no pre-installed flexible capacity.
    Assumed in Section IV so the model fully determines the optimal technology mix; real sites with existing assets would produce different investment sizes.
  • domain assumption Electricity and heat demand are exogenous and fixed apart from allowed load shifting.
    The model optimizes supply and storage, not demand; demand response is limited to the specified shifting window, as stated in Section IV.
  • standard math Uncertainty is represented through a scenario tree with node probabilities that sum to one for each stage.
    Required for the expected-value objective function in Equation (1), where each node probability pi_omega is defined and sums to one.
invented entities (1)
  • Dummy fuels
    purpose: Modeling devices that connect technologies and storage through an energy carrier even when no physical fuel exists, enabling multi-fuel flexibility in the case study.
    Introduced in Section IV ('the others are dummy-fuels to enable multi-fuel flexibility'). They are graph-theoretic glue in the optimization, not physical substances, so they have no independent empirical handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Industrial Flexibility Investment Under Uncertainty: A Multi-Stage Stochastic Framework Considering Energy and Reserve Market Participation." pith.science (2026). https://pith.science/paper/2GDWMG24

@misc{pith2026250608638,
  author       = {Pith},
  title        = {Pith review of: Industrial Flexibility Investment Under Uncertainty: A Multi-Stage Stochastic Framework Considering Energy and Reserve Market Participation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2GDWMG24}},
  note         = {Machine review of arXiv:2506.08638}
}
read the original abstract

The global energy transition toward net-zero emissions by 2050 is expected to increase the share of variable renewable energy sources (VRES) in the energy mix. As a result, industrial actors will encounter more complex market conditions, characterized by volatile electricity prices, rising carbon costs, and stricter regulations. This situation calls for the industry to capitalize on opportunities in both spot-price arbitrage and reserve market participation, while also meeting future regulatory demands. This paper presents a multi-stage optimization framework that supports investment decisions in flexible assets and enables reserve market participation by delivering ancillary services. The framework incorporates investment decisions, spot- and reserve-market bidding, and real-time operation. Uncertainty in market prices and operational conditions is handled through a nodal formulation. A case study of a large industrial site in Norway is performed, comparing the investment decisions with future technology- and carbon pricing scenarios under varying market conditions.

Figures

Figures reproduced from arXiv: 2506.08638 by the authors.

Figure 1
Figure 1. Modeling framework logic, visualizing how uncertainty is revealed and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Equivalent annual cost of technologies, using a discount rate of 9% [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. Equivalent annual cost of storage, using a discount rate of 9% [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Storage technology investments in 2025 and 2050 with and without [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

9 extracted references · 8 canonical work pages

  1. [1]

    Optimal bidding in sequential physical markets—a literature review and framework discussion,

    G. Klæboe and O. B. Fosso, “Optimal bidding in sequential physical markets—a literature review and framework discussion,” in2013 IEEE Grenoble Conference. IEEE, 2013, pp. 1–6. [Online]. Available: https://doi.org/10.1109/PTC.2013.6652371

    work page arXiv 2013

  2. [2]

    The value of coordination in multimarket bidding of grid energy storage,

    N. L ¨ohndorf and D. Wozabal, “The value of coordination in multimarket bidding of grid energy storage,”Operations research, vol. 71, no. 1, pp. 1–22, 2023. [Online]. Available: https://doi.org/10.1287/opre.2021.2247

    arXiv 2023

  3. [3]

    Bidding in sequential electricity markets: The nordic case,

    T. K. Boomsma, N. Juul, and S.-E. Fleten, “Bidding in sequential electricity markets: The nordic case,” European Journal of Operational Research, vol. 238, no. 3, pp. 797–809, 2014. [Online]. Available: https: //doi.org/10.1016/j.ejor.2014.04.027

    work page doi:10.1016/j.ejor.2014.04.027 2014

  4. [4]

    Multi market bidding strategies for demand side flexibility aggregators in electricity markets,

    S. Ø. Ottesen, A. Tomasgard, and S.-E. Fleten, “Multi market bidding strategies for demand side flexibility aggregators in electricity markets,”Energy, vol. 149, pp. 120–134, 2018. [Online]. Available: https://doi.org/10.1016/j.energy.2018.01.187

    work page doi:10.1016/j.energy.2018.01.187 2018

  5. [5]

    Trading strategy for a flexible factory participating in the german balancing and day-ahead market,

    M. B ¨ohringer, T. Pl¨oßer, J. Hanson, T. Weitzel, C. Glock, and N. Roloff, “Trading strategy for a flexible factory participating in the german balancing and day-ahead market,” in2019 54th International Universities Power Engineering Conference (UPEC). IEEE, 2019, pp. 1–6. [Online]. Available: https://doi.org/10.1109/UPEC.2019. 8893521

    work page doi:10.1109/upec.2019 2019

  6. [6]

    Energy-intense production-inventory planning with participation in sequential energy markets,

    M. Bohlayer, M. Fleschutz, M. Braun, and G. Z ¨ottl, “Energy-intense production-inventory planning with participation in sequential energy markets,” Applied Energy, vol. 258, 2020. [Online]. Available: https://doi.org/10.1016/j.apenergy.2019.113954

    work page arXiv 2020

  7. [7]

    Stacking revenues from flexible ders in multi-scale markets using tri-level optimization,

    ´A. Paredes, J. A. Aguado, C. Essayeh, Y . Xia, I. Savelli, and T. Morstyn, “Stacking revenues from flexible ders in multi-scale markets using tri-level optimization,”IEEE Transactions on Power Systems, vol. 39, no. 2, pp. 3949–3961, 2024. [Online]. Available: https://doi.org/10.1109/TPWRS.2023.3286178

    work page arXiv 2024

  8. [8]

    Comparative analyses of seven technologies to facilitate the integration of fluctu- ating renewable energy sources,

    B. V . Mathiesen and H. Lund, “Comparative analyses of seven technologies to facilitate the integration of fluctu- ating renewable energy sources,”IET Renewable Power Generation, vol. 3, no. 2, pp. 190–204, 2009. [Online]. Available: https://doi.org/10.1049/iet-rpg:20080049

    work page doi:10.1049/iet-rpg:20080049 2009

Show all 9 references
  1. [9]

    Optimal selection of thermal energy storage technology for fossil-free steam production in the processing industry,

    A. Beck, A. Sevault, G. Drexler-Schmid, M. Sch ¨ony, and H. Kauko, “Optimal selection of thermal energy storage technology for fossil-free steam production in the processing industry,”Applied Sciences, vol. 11, no. 3, p. 1063, 2021. [Online]. Available: https: //doi.org/10.339...

    2021 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.