pith. machine review for the scientific record. sign in

arxiv: 2605.12462 · v1 · submitted 2026-05-12 · 💻 cs.AI · cs.CY· cs.GT· cs.LG

Recognition: 2 theorem links

· Lean Theorem

Towards Affordable Energy: A Gymnasium Environment for Electric Utility Demand-Response Programs

Huazheng Wang, Jose E. Aguilar Escamilla, Lingdong Zhou, Xiangqi Zhu

Pith reviewed 2026-05-13 04:10 UTC · model grok-4.3

classification 💻 cs.AI cs.CYcs.GTcs.LG
keywords demand responsereinforcement learningGymnasium environmentelectric utilityenergy affordabilitywholesale electricity pricesbuilding energy simulation
0
0 comments X

The pith

A new open-source Gymnasium environment called DR-Gym lets utilities train reinforcement learning agents on demand-response decisions that account for customer responses to pricing.

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

The paper introduces DR-Gym to address the gap where offline historical smart meter and wholesale price data cannot model the live feedback loop between an electric utility issuing pricing signals and customers adjusting their usage. The environment simulates this interaction online using a regime-switching model for wholesale prices tuned to real extreme events and physics-based building demand profiles. A configurable multi-objective reward function allows specification of goals such as reducing customer costs or improving grid flexibility. Baseline strategies demonstrate that the simulator generates realistic and learnable scenarios focused on market-level utility decisions rather than individual devices. If the models hold, utilities gain a practical way to develop and test policies that issue credits during high-price periods to protect consumers from volatile markets.

Core claim

We introduce DR-Gym, an open-source, online Gymnasium-compatible environment designed to train and evaluate demand-response from the electric utility's perspective. Unlike existing device-level energy simulators, our environment focuses on the market-level electric utility setting and provides a rich observational space relevant to the electric utility. The simulator additionally features a regime-switching wholesale price model calibrated to real-world extreme events, alongside physics-based building demand profiles. For our learning signal, we use a configurable, multi-objective reward function for specifying diverse learning objectives. We demonstrate through baseline strategies and data

What carries the argument

DR-Gym, the Gymnasium-compatible simulator that models interactive feedback between utility pricing signals and customer demand using regime-switching wholesale prices and physics-based building profiles.

Load-bearing premise

The regime-switching wholesale price model calibrated to real-world extreme events together with physics-based building demand profiles sufficiently capture the interactive feedback between utility pricing signals and customer acceptance.

What would settle it

A direct comparison of demand reductions produced by the simulator under specific pricing policies against measured outcomes from an actual utility demand-response trial.

Figures

Figures reproduced from arXiv: 2605.12462 by Huazheng Wang, Jose E. Aguilar Escamilla, Lingdong Zhou, Xiangqi Zhu.

Figure 1
Figure 1. Figure 1: Overview of simulator architecture. From left to right, we outline the external data we [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Detailed simulator architecture. We outline each part of our simulator as well as the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Wholesale price dynamics produced by the [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Market model validation against ERCOT day-ahead market statistics. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Aggregate building demand profile: mean and standard deviation across 50 buildings [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Customer response model validation. (a) Acceptance probability curves for each of the four archetypes as a function of credit level (Equation 6); weighted average acceptance at c = $0.05/kWh is ≈ 0.65, consistent with empirical DR pilot ranges [10]. (b) Fatigue decay: acceptance factor over consecutive daily activations for each archetype, illustrating the declining￾participation dynamic captured by the fa… view at source ↗
Figure 7
Figure 7. Figure 7: PPO learning and final evaluation. (a) PPO episode reward during training (smoothed). The agent consistently improves over the heuristic baselines within 5 × 105 steps. (b) Final per￾formance comparison over 100 evaluation episodes. PPO achieves the highest mean reward [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Baseline and PPO policy comparison over 50 evaluation episodes. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Baseline policy definitions. Bt = budget remaining at step t; B0 = initial daily budget. Simulation Time Using our simulator to generate data often takes longer the more buildings are chosen to be simulated. Currently, we achieve a rate of 0.3 seconds per episode (24 steps) when simulating N = 50 buildings. In contrast, we achieve a rate of 13 seconds per episode (again, 24 steps) when simulating N = 500 b… view at source ↗
Figure 10
Figure 10. Figure 10: Scalability ablation: baseline policy performance at [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: CVaR–reward trade-off analysis. Sweeping a uniform credit policy from [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
read the original abstract

Extreme weather and volatile wholesale electricity markets expose residential consumers to catastrophic financial risks, yet demand response at the distribution level remains an underutilized tool for grid flexibility and energy affordability. While a demand-response program can shield consumers by issuing financial credits during high-price periods, optimizing this sequential decision-making process presents a unique challenge for reinforcement learning despite the plentiful offline historical smart meter and wholesale pricing data available publicly. Offline historical data fails to capture the dynamic, interactive feedback loop between an electric utility's pricing signals and customer acceptance and adaptation to a demand-response program. To address this, we introduce DR-Gym, an open-source, online Gymnasium-compatible environment designed to train and evaluate demand-response from the electric utility's perspective. Unlike existing device-level energy simulators, our environment focuses on the market-level electric utility setting and provides a rich observational space relevant to the electric utility. The simulator additionally features a regime-switching wholesale price model calibrated to real-world extreme events, alongside physics-based building demand profiles. For our learning signal, we use a configurable, multi-objective reward function for specifying diverse learning objectives. We demonstrate through baseline strategies and data snapshots the capability of our simulator to create realistic and learnable environments.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript presents DR-Gym, a new open-source Gymnasium environment for simulating and optimizing demand-response programs from the electric utility's perspective. It includes a regime-switching wholesale price model calibrated to extreme events, physics-based building demand profiles, and a configurable multi-objective reward function. The environment is intended to model the interactive feedback between utility pricing and customer responses, enabling RL training where offline data falls short. Baseline strategies are shown to illustrate its use.

Significance. Should the simulator prove realistic, it would offer a valuable platform for developing and benchmarking RL algorithms for demand response, potentially aiding in mitigating financial risks from volatile energy prices and enhancing grid flexibility. The open-source and Gymnasium-compatible design promotes accessibility and reproducibility in the field. The shift to market-level utility focus distinguishes it from device-level simulators.

major comments (2)
  1. [Abstract] The central claim that DR-Gym captures the dynamic interactive feedback loop between pricing signals and customer acceptance relies on the regime-switching price model and physics-based profiles, yet no quantitative validation metrics, error analysis, or comparisons to real-world customer response data (e.g., smart-meter data under DR programs) are provided. This leaves the fidelity of the modeled interactions unverified.
  2. [Demonstration of baseline strategies] The provided demonstrations consist only of baseline strategies and data snapshots without any quantitative assessment of how well the simulated load shifts or acceptance rates match empirical outcomes from actual demand-response programs.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We address each major comment below and describe the revisions we intend to incorporate to improve the manuscript.

read point-by-point responses

  1. Referee: [Abstract] The central claim that DR-Gym captures the dynamic interactive feedback loop between pricing signals and customer acceptance relies on the regime-switching price model and physics-based profiles, yet no quantitative validation metrics, error analysis, or comparisons to real-world customer response data (e.g., smart-meter data under DR programs) are provided. This leaves the fidelity of the modeled interactions unverified.

    Authors: We agree that the absence of quantitative validation metrics, error analysis, or direct comparisons to real-world smart-meter data from operational DR programs leaves the fidelity of the modeled pricing-customer interaction unverified. The manuscript grounds the price component in a regime-switching process calibrated to historical extreme events and uses physics-based building demand models, but does not provide empirical benchmarks for customer acceptance rates or load-shift magnitudes. In the revised manuscript we will add a dedicated limitations subsection that explicitly states these gaps, includes sensitivity analysis on the acceptance parameters, and outlines feasible validation pathways using publicly available DR program reports where direct smart-meter traces are unavailable. revision: yes

  2. Referee: [Demonstration of baseline strategies] The provided demonstrations consist only of baseline strategies and data snapshots without any quantitative assessment of how well the simulated load shifts or acceptance rates match empirical outcomes from actual demand-response programs.

    Authors: The baseline demonstrations are intended to illustrate environment usability and task learnability rather than to serve as empirical validation. We acknowledge that they lack quantitative metrics comparing simulated load shifts and acceptance rates to outcomes reported in the DR literature. In the revision we will expand the demonstration section with additional performance tables that report load-shift percentages and acceptance fractions under the baseline policies, together with a brief comparison to typical ranges cited in utility DR program evaluations, while clearly noting that these remain illustrative rather than statistically validated against proprietary smart-meter datasets. revision: yes

Circularity Check

0 steps flagged

No circularity: environment creation with inputs stated as such

full rationale

The paper's core contribution is the creation and release of the DR-Gym simulator rather than any derivation, theorem, or prediction. The regime-switching wholesale price model and physics-based building profiles are explicitly presented as calibrated inputs and features of the environment, not as outputs or predictions derived from the simulator itself. Baseline strategy demonstrations are described as illustrative snapshots without any claim that they constitute fitted predictions or self-referential results. No equations, uniqueness theorems, or self-citations are invoked in a load-bearing way that would reduce the central claim to its own inputs by construction. The work is therefore self-contained as an engineering artifact.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the new simulator itself; it introduces calibrated parameters for the price model and domain assumptions about customer behavior that are not independently verified in the provided abstract.

free parameters (2)
  • regime-switching wholesale price parameters
    Calibrated to real-world extreme events; these values are fitted inputs that define the price dynamics the RL agent interacts with.
  • multi-objective reward weights
    Configurable coefficients that shape the learning signal; their specific values determine which behaviors are reinforced.
axioms (1)
  • domain assumption Physics-based building demand profiles plus historical smart-meter patterns can approximate real customer responses to pricing signals
    Invoked to justify the simulator's ability to close the feedback loop missing from offline data.

pith-pipeline@v0.9.0 · 5528 in / 1205 out tokens · 69311 ms · 2026-05-13T04:10:20.118794+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

Reference graph

Works this paper leans on

32 extracted references · 32 canonical work pages

  1. [1]

    URLhttps://www.eia.gov/todayinenergy/detail.php?id=55139#

    Lori Antini, Jan 2023. URLhttps://www.eia.gov/todayinenergy/detail.php?id=55139#

    work page 2023

  2. [2]

    Artificial intelligence and machine learning approaches to energy demand-side response: A systematic review.Renewable and Sustainable Energy Reviews, 130:109899, 2020

    Ioannis Antonopoulos, Valentin Robu, Benoit Couraud, Desen Kirli, Sonam Norbu, Aris- tides Kiprakis, David Flynn, Salvador Elizondo-Gonzalez, Steve Wattam, and Merlinda An- doni. Artificial intelligence and machine learning approaches to energy demand-side response: A systematic review.Renewable and Sustainable Energy Reviews, 130:109899, 2020. doi: 10.10...

    work page doi:10.1016/j.rser.2020.109899 2020

  3. [3]

    Retail demand response in Southwest Power Pool

    Ranjit Bharvirkar, Charles Goldman, Grayson Heffner, and Richard Sedano. Retail demand response in Southwest Power Pool. Technical Report LBNL-1470E, Lawrence Berkeley Na- tional Laboratory, 2009. URLhttps://eta-publications.lbl.gov/sites/default/files/ report-lbnl-1470e.pdf

    work page 2009

  4. [4]

    Dynamic pricing, advanced meter- ing, and demand response in electricity markets

    Severin Borenstein, Michael Jaske, and Arthur Rosenfeld. Dynamic pricing, advanced meter- ing, and demand response in electricity markets. Technical Report CSEM WP 105, Center for the Study of Energy Markets, UC Berkeley, 2002. URLhttps://escholarship.org/uc/ item/11w8d6m4

    work page 2002

  5. [5]

    gridstatus: Extract data from isos and other energy grid sources,

    Fiona Burlig and Others. gridstatus: Extract data from isos and other energy grid sources,

    work page

  6. [6]

    URLhttps://github.com/gridstatus/gridstatus

    work page

  7. [7]

    Sinergym – a virtual testbed for building energy op- timization with reinforcement learning.Energy and Buildings, 327:115075, January 2025

    Alejandro Campoy-Nieves, Antonio Manjavacas, Javier Jim´ enez-Raboso, Miguel Molina- Solana, and Juan G´ omez-Romero. Sinergym – a virtual testbed for building energy op- timization with reinforcement learning.Energy and Buildings, 327:115075, January 2025. ISSN 0378-7788. doi: 10.1016/j.enbuild.2024.115075. URLhttp://dx.doi.org/10.1016/j. enbuild.2024.115075

    work page doi:10.1016/j.enbuild.2024.115075 2025

  8. [8]

    Crawley, Linda K

    Drury B. Crawley, Linda K. Lawrie, Frederick C. Winkelmann, W. F. Buhl, Y. Joe Huang, Curtis O. Pedersen, Richard K. Strand, Richard J. Liesen, Daniel E. Fisher, Michael J. Witte, and Jason Glazer. EnergyPlus: Creating a new-generation building energy simulation program. 14 Energy and Buildings, 33(4):319–331, 2001. doi: 10.1016/S0378-7788(00)00114-6. URL...

    work page doi:10.1016/s0378-7788(00)00114-6 2001

  9. [9]

    Bellemare, and R´ emi Munos

    Will Dabney, Mark Rowland, Marc G. Bellemare, and R´ emi Munos. Distributional reinforce- ment learning with quantile regression. InProceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, pages 2892–2901, 2018. URLhttps://arxiv.org/abs/1710.10044

    work page arXiv 2018

  10. [10]

    Gymnasium: A standard interface for reinforcement learning environ- ments.https://gymnasium.farama.org, 2023

    Farama Foundation. Gymnasium: A standard interface for reinforcement learning environ- ments.https://gymnasium.farama.org, 2023

    work page 2023

  11. [11]

    Household response to dynamic pricing of electric- ity: a survey of 15 experiments.Journal of Regulatory Economics, 38(2):193–225, 2010

    Ahmad Faruqui and Sanem Sergici. Household response to dynamic pricing of electric- ity: a survey of 15 experiments.Journal of Regulatory Economics, 38(2):193–225, 2010. doi: 10.1007/s11149-010-9127-y. URLhttps://link.springer.com/article/10.1007/ s11149-010-9127-y

    work page doi:10.1007/s11149-010-9127-y 2010

  12. [12]

    The February 2021 cold weather out- ages in Texas and the south central United States

    Federal Energy Regulatory Commission (FERC) and North American Elec- tric Reliability Corporation (NERC). The February 2021 cold weather out- ages in Texas and the south central United States. Technical report, FERC and NERC, November 2021. URLhttps://www.ferc.gov/news-events/news/ final-report-february-2021-freeze-underscores-winterization-recommendations

    work page 2021

  13. [13]

    A comprehensive survey on safe reinforcement learning

    Javier Garcia and Fernando Fernandez. A comprehensive survey on safe reinforcement learning. Journal of Machine Learning Research, 16(42):1437–1480, 2015. URLhttp://jmlr.org/ papers/v16/garcia15a.html

    work page 2015

  14. [14]

    Girish Ghatikar, Mary Ann Piette, and John H. Dudley. Demand response opportunities and enabling technologies for data centers: Findings from field studies. Technical Report LBNL- 1004492, Lawrence Berkeley National Laboratory, 2012. URLhttps://gridintegration. lbl.gov/publications/demand-response-opportunities-and

    work page 2012

  15. [15]

    A review of residen- tial demand response of smart grid.Renewable and Sustainable Energy Reviews, 59:166– 178, 2016

    Haider Tarish Haider, Ong Hang See, and Wilfried Elmenreich. A review of residen- tial demand response of smart grid.Renewable and Sustainable Energy Reviews, 59:166– 178, 2016. ISSN 1364-0321. doi: https://doi.org/10.1016/j.rser.2016.01.016. URLhttps: //www.sciencedirect.com/science/article/pii/S1364032116000447

    work page doi:10.1016/j.rser.2016.01.016 2016

  16. [16]

    pymgrid: An open-source python microgrid simulator for applied artificial intelligence research.CoRR, abs/2011.08004, 2020

    Gonzague Henri, Tanguy Levent, Avishai Halev, R´ eda Alami, and Philippe Cordier. pymgrid: An open-source python microgrid simulator for applied artificial intelligence research.CoRR, abs/2011.08004, 2020. URLhttps://arxiv.org/abs/2011.08004

    work page arXiv 2011

  17. [17]

    Regime jumps in electricity prices.Energy Economics, 25 (5):425–434, 2003

    Ronald Huisman and Ronald Mahieu. Regime jumps in electricity prices.Energy Economics, 25 (5):425–434, 2003. doi: 10.1016/S0140-9883(03)00041-0. URLhttps://www.sciencedirect. com/science/article/abs/pii/S0140988303000410

    work page doi:10.1016/s0140-9883(03)00041-0 2003

  18. [18]

    Kirschen and Goran Strbac.Fundamentals of Power System Economics

    Daniel S. Kirschen and Goran Strbac.Fundamentals of Power System Economics. John Wiley & Sons, 1st edition, 2004. ISBN 0-470-84572-4. URLhttps://onlinelibrary.wiley.com/ doi/book/10.1002/0470020598

    work page doi:10.1002/0470020598 2004

  19. [19]

    Kirschen

    D.S. Kirschen. Demand-side view of electricity markets.IEEE Transactions on Power Systems, 18(2):520–527, 2003. doi: 10.1109/TPWRS.2003.810692

    work page doi:10.1109/tpwrs.2003.810692 2003

  20. [20]

    Learning 15 to run a power network challenge: a retrospective analysis

    Antoine Marot, Benjamin Donnot, Gabriel Dulac-Arnold, Adrian Kelly, Aidan O’Sullivan, Jan Viebahn, Mariette Awad, Isabelle Guyon, Patrick Panciatici, and Camilo Romero. Learning 15 to run a power network challenge: a retrospective analysis. In Hugo Jair Escalante and Katja Hofmann, editors,Proceedings of the NeurIPS 2020 Competition and Demonstration Trac...

    work page 2020

  21. [21]

    URLhttps://proceedings.mlr.press/v133/marot21a.html

    work page

  22. [22]

    Flexible de- mand response programs modeling in competitive electricity markets.Applied Energy, 88 (9):3257–3269, 2011

    Mohsen Parsa Moghaddam, Abdolreza Abdollahi, and Masoud Rashidinejad. Flexible de- mand response programs modeling in competitive electricity markets.Applied Energy, 88 (9):3257–3269, 2011. URLhttps://www.sciencedirect.com/science/article/abs/pii/ S0306261911000638

    work page 2011

  23. [23]

    Market power and efficiency in a computational electricity market with discriminatory double-auction pricing.IEEE Transac- tions on Evolutionary Computation, 5(5):504–523, 2001

    James Nicolaisen, Valentin Petrov, and Leigh Tesfatsion. Market power and efficiency in a computational electricity market with discriminatory double-auction pricing.IEEE Transac- tions on Evolutionary Computation, 5(5):504–523, 2001. URLhttps://www.researchgate. net/publication/3418664

    work page arXiv 2001

  24. [24]

    Stable-Baselines3: Reliable reinforcement learning implementations.Journal of Ma- chine Learning Research, 22(268):1–8, 2021

    Antonin Raffin, Ashley Hill, Adam Gleave, Anssi Kanervisto, Maximilian Ernestus, and Noah Dormann. Stable-Baselines3: Reliable reinforcement learning implementations.Journal of Ma- chine Learning Research, 22(268):1–8, 2021. URLhttps://jmlr.org/papers/v22/20-1364. html

    work page 2021

  25. [25]

    Tyrrell Rockafellar and Stanislav Uryasev

    R. Tyrrell Rockafellar and Stanislav Uryasev. Optimization of conditional value-at-risk.Jour- nal of Risk, 2(3):21–41, 2000. doi: 10.21314/JOR.2000.038. URLhttps://sites.math. washington.edu/~rtr/papers/rtr179-CVaR1.pdf

    work page doi:10.21314/jor.2000.038 2000

  26. [26]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto.Reinforcement Learning: An Introduction. The MIT Press, Cambridge, MA, 2nd edition, 2018. URLhttp://incompleteideas.net/book/ the-book-2nd.html

    work page 2018

  27. [27]

    Optimizing the CVaR via sampling

    Aviv Tamar, Yonatan Glassner, and Shie Mannor. Optimizing the CVaR via sampling. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015. URLhttps: //arxiv.org/abs/1404.3862

    work page arXiv 2015

  28. [28]

    Siebert, and Jochen L

    Jasper van Tilburg, Luciano C. Siebert, and Jochen L. Cremer. Marl-idr: Multi-agent rein- forcement learning for incentive-based residential demand response. In2023 IEEE Belgrade PowerTech, pages 1–8, 2023. doi: 10.1109/PowerTech55446.2023.10202941

    work page doi:10.1109/powertech55446.2023.10202941 2023

  29. [29]

    V´ azquez-Canteli, J´ erˆ ome K¨ ampf, Gregor Henze, and Zoltan Nagy

    Jos´ e R. V´ azquez-Canteli, J´ erˆ ome K¨ ampf, Gregor Henze, and Zoltan Nagy. CityLearn v1.0: An OpenAI Gym environment for demand response with deep reinforcement learn- ing. InProceedings of the 6th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (BuildSys ’19), pages 356–357. ACM, 2019. doi: 10.1145/...

    work page doi:10.1145/3360322.3360998 2019

  30. [30]

    V´ azquez-Canteli and Zolt´ an Nagy

    Jos´ e R. V´ azquez-Canteli and Zolt´ an Nagy. Reinforcement learning for demand response: A review of algorithms and modeling techniques.Applied Energy, 235:1072–1089, 2019. ISSN 0306-2619. doi: https://doi.org/10.1016/j.apenergy.2018.11.002. URLhttps://www. sciencedirect.com/science/article/pii/S0306261918317082

    work page doi:10.1016/j.apenergy.2018.11.002 2019

  31. [31]

    Electricity price forecasting: A review of the state-of-the-art with a look into the future.International Journal of Forecasting, 30(4):1030–1081, 2014

    Rafa l Weron. Electricity price forecasting: A review of the state-of-the-art with a look into the future.International Journal of Forecasting, 30(4):1030–1081, 2014. doi: 10. 1016/j.ijforecast.2014.08.008. URLhttps://www.sciencedirect.com/science/article/ pii/S0169207014001083. 16

    work page 2014

  32. [32]

    End-use load profiles for the U.S

    Eric Wilson, Andrew Parker, Anthony Fontanini, Elaina Present, Janet Reyna, Rajendra Ad- hikari, Matthew Dahlhausen, Janghyun Kim, Amy LeBar, Liang Liu, et al. End-use load profiles for the U.S. building stock. Technical Report NREL/TP-5500-80889, National Renew- able Energy Laboratory, 2022. URLhttps://www.nrel.gov/docs/fy22osti/80889.pdf. 17 A Observati...

    work page 2022