pith. sign in

arxiv: 2606.02852 · v1 · pith:OKSY2HXFnew · submitted 2026-06-01 · 💻 cs.LG

RESCAST-100K: A Comprehensive Dataset for Cross-Domain Residential Load and Indoor Temperature Forecasting

Jainam Dhruva , Yousaf Raza , A.B. Siddique , Simone Silvestri This is my paper

Pith reviewed 2026-06-28 15:33 UTC · model grok-4.3

classification 💻 cs.LG
keywords residential load forecastingindoor temperature forecastingdomain adaptationtransfer learningbenchmark datasetcross-domain generalizationEnergyPlus simulationzero-shot forecasting
0
0 comments X

The pith

RESCAST-100K supplies a configuration-driven benchmark of 100,000 homes to evaluate transfer learning and zero-shot forecasting across residential domain shifts.

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

The paper introduces RESCAST-100K to fill the gap in residential datasets that support structured tests of how forecasting models handle differences between homes. It supplies both a large collection of simulated homes and integrated real datasets under one schema, together with an interface that lets users set source and target domains by varying geography, climate zone, wall construction, and heating equipment. This controlled setup makes it possible to measure the effectiveness of domain adaptation and transfer methods on three coupled forecasting targets: total load, HVAC load, and indoor temperature. A reader would care because residential energy data is highly heterogeneous, and reliable cross-domain methods could reduce the amount of local data needed to deploy accurate short-term predictions for home management and grid demand response.

Core claim

RESCAST-100K provides approximately 100,000 EnergyPlus-simulated U.S. homes derived from ResStock, each with 15-minute time series for total load, HVAC load, and indoor temperature, paired with weather channels, HVAC setpoints, and over 40 static building covariates. It also integrates five real-world residential datasets under a unified schema. The central mechanism is a configuration-driven interface that instantiates source and target domains along interpretable axes including geography, climate zone, wall construction, and heating equipment, enabling systematic evaluation of transfer learning, domain adaptation, and zero-shot generalization under controlled domain shifts. Benchmark resul

What carries the argument

The configuration-driven interface that instantiates source and target domains along interpretable axes including geography, climate zone, wall construction, and heating equipment.

If this is right

  • Researchers can now run controlled experiments comparing recurrent, attention-based, and MLP-mixer models for zero-shot performance across domains and missing-input conditions.
  • Cross-attention and MLP-mixer models can be expected to deliver higher accuracy than recurrent and classical transformer baselines when domain shifts are present.
  • Direct sim-to-real evaluation becomes feasible on the same load and temperature forecasting tasks.
  • The benchmark supports studies at home, community, and grid scale by providing consistent data across many domain combinations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Improved zero-shot models developed with this benchmark could allow deployment in new regions with minimal or no local training data.
  • The same configuration approach might be extended to test generalization across additional building types or international locations.
  • Forecast accuracy gains could feed into optimization routines for home energy management systems that operate under varying climate and equipment conditions.

Load-bearing premise

The simulated homes and integrated real datasets generate domain shifts that are representative of real-world residential variability and that affect forecasting model performance in the intended ways.

What would settle it

An experiment in which models trained on one climate zone or construction type show no measurable drop in accuracy when tested on another zone or type within the benchmark would indicate that the controlled shifts do not produce the expected generalization challenge.

Figures

Figures reproduced from arXiv: 2606.02852 by A.B. Siddique, Jainam Dhruva, Simone Silvestri, Yousaf Raza.

Figure 1
Figure 1. Figure 1: Distribution of total load (kWh), HVAC load (kWh), indoor temperature ( [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Data Curation pipeline for RESCAST-100K. Home envelope files, weather data, and stochastic schedules, derived from Resstock are used as inputs to EnergyPlus simulations computed through high performance computing (HPC) cluster. The data can be used to create specific domain splits for cross domain adaptions. This dataset can be used to evaluate various domain adaptation methodologies for time series foreca… view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of total load, HVAC load, indoor and outdoor temperatures across state [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Distribution of total load, HVAC load, indoor and outdoor temperatures across climate zone [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of total load, HVAC load, indoor and outdoor temperatures across different [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of total load, HVAC load, indoor and outdoor temperatures across heating [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distributions of indoor temperature values ( [PITH_FULL_IMAGE:figures/full_fig_p023_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Distributions of total load values (y) in the top row and their consecutive differences (∆y) in the bottom row across real and their corresponding synthetic datasets. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_8.png] view at source ↗
read the original abstract

Accurate short-term forecasting of residential energy load and indoor temperature is essential for home energy management systems, grid-level demand response, and community energy efficiency efforts. Domain adaptation and transfer learning have shown promise for improving forecasting accuracy under data heterogeneity and scarcity commonly seen in residential settings. However, progress is limited by the lack of comprehensive residential datasets: existing benchmarks are narrow in target coverage and rarely support structured cross-domain evaluation. We introduce RESCAST-100K, a large-scale residential forecasting benchmark for studying cross-domain generalization. It provides a configuration-driven interface that instantiates source and target domains along interpretable axes, including geography, climate zone, wall construction, and heating equipment, enabling systematic evaluation of transfer learning, domain adaptation, and zero-shot generalization under controlled domain shifts. The benchmark covers approximately 100,000 EnergyPlus-simulated U.S. homes derived from ResStock, with 15-minute time series for three coupled targets per home: total load, HVAC load, and indoor temperature. These are paired with weather channels, HVAC setpoints, and over 40 static building covariates. RESCAST-100K also integrates five real-world residential datasets under a unified schema, supporting sim-to-real evaluation on the same tasks. We benchmark recurrent, attention-based, and MLP-mixer architectures for zero-shot performance across domains, missing-input conditions, and forecasting tasks. Cross-attention and MLP-mixer models consistently outperform recurrent and classical transformer baselines under domain shift. RESCAST-100K is intended to aid the machine learning and building analytics communities advance cross-domain residential forecasting at home, community, and grid scale.

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 / 1 minor

Summary. The manuscript introduces RESCAST-100K, a large-scale benchmark dataset for cross-domain residential load and indoor temperature forecasting. It comprises ~100k EnergyPlus simulations of U.S. homes derived from ResStock (with 15-minute time series for total load, HVAC load, and indoor temperature, plus weather and >40 static covariates) together with five integrated real-world datasets under a unified schema. A configuration-driven interface allows instantiation of source/target domains along interpretable axes (geography, climate zone, wall construction, heating equipment) to support systematic study of transfer learning, domain adaptation, and zero-shot generalization. The paper also reports benchmarks of recurrent, attention-based, and MLP-mixer architectures, claiming that cross-attention and MLP-mixer models consistently outperform recurrent and classical transformer baselines under domain shift.

Significance. If the simulated domain shifts prove representative of real residential variability and produce the expected systematic effects on forecasting performance, RESCAST-100K would address a clear gap in existing narrow benchmarks and provide a valuable standardized resource for the ML and building analytics communities working on data heterogeneity in residential energy applications.

major comments (2)
  1. [Abstract] Abstract / dataset description: The central claim that the configuration-driven interface enables 'systematic evaluation of transfer learning, domain adaptation, and zero-shot generalization under controlled domain shifts' is load-bearing for the contribution, yet the manuscript provides no quantitative validation of simulation fidelity, no error analysis comparing EnergyPlus/ResStock outputs to real measurements, and no demonstration that the resulting shifts in load/temperature series degrade model performance in an axis-aligned manner rather than as simulation artifacts.
  2. [Benchmarks] Benchmarks section: The statement that 'cross-attention and MLP-mixer models consistently outperform recurrent and classical transformer baselines under domain shift' is presented without accompanying performance numbers, tables, or analysis showing performance gaps along the claimed axes (e.g., climate zone or heating equipment); this leaves the utility of the controlled shifts for domain-adaptation research unverified.
minor comments (1)
  1. [Abstract] The abstract would be strengthened by inclusion of at least one or two key quantitative benchmark results to substantiate the outperformance claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive feedback on the RESCAST-100K manuscript. We address each major comment below and commit to revisions that strengthen the validation of the dataset and benchmarks.

read point-by-point responses

  1. Referee: [Abstract] Abstract / dataset description: The central claim that the configuration-driven interface enables 'systematic evaluation of transfer learning, domain adaptation, and zero-shot generalization under controlled domain shifts' is load-bearing for the contribution, yet the manuscript provides no quantitative validation of simulation fidelity, no error analysis comparing EnergyPlus/ResStock outputs to real measurements, and no demonstration that the resulting shifts in load/temperature series degrade model performance in an axis-aligned manner rather than as simulation artifacts.

    Authors: We agree that quantitative validation of simulation fidelity is essential to substantiate the controlled domain shifts. Although ResStock/EnergyPlus is a widely used and validated platform in the building science community, the manuscript does not include explicit error metrics against the integrated real-world datasets. In the revised manuscript, we will add a dedicated validation subsection that reports distribution-level comparisons (e.g., Kolmogorov-Smirnov statistics and mean absolute percentage errors) between simulated and measured load/temperature series for overlapping climate zones and building types. We will also include an axis-aligned degradation analysis showing forecasting error increases when shifting along individual covariates (climate zone, heating equipment) while holding others fixed, to distinguish systematic shifts from potential simulation artifacts. revision: yes

  2. Referee: [Benchmarks] Benchmarks section: The statement that 'cross-attention and MLP-mixer models consistently outperform recurrent and classical transformer baselines under domain shift' is presented without accompanying performance numbers, tables, or analysis showing performance gaps along the claimed axes (e.g., climate zone or heating equipment); this leaves the utility of the controlled shifts for domain-adaptation research unverified.

    Authors: We acknowledge that the benchmarks section as currently written does not provide the detailed numerical results or axis-specific breakdowns needed to fully verify the claims. The revised manuscript will expand the benchmarks section with complete tables reporting MAE, RMSE, and MAPE for all model families across source-target pairs, together with figures that disaggregate performance gaps by climate zone, heating equipment type, and wall construction. These additions will directly demonstrate the utility of the controlled shifts for domain-adaptation studies. revision: yes

Circularity Check

0 steps flagged

Dataset benchmark paper with no derivations or self-referential claims

full rationale

The manuscript describes the construction of RESCAST-100K from ResStock/EnergyPlus simulations plus five real datasets, a unified schema, and baseline model benchmarks. No equations, fitted parameters, uniqueness theorems, or ansatzes appear in the provided text. The central claim is the existence and utility of the released resource itself; it does not derive any quantity from prior results that could reduce to its own inputs. Self-citations, if present, are not load-bearing for any inference chain. This is the standard non-circular outcome for a data-release paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a dataset and benchmark paper with no mathematical derivations or models; the central contribution rests on data generation and schema unification rather than on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5837 in / 1220 out tokens · 32628 ms · 2026-06-28T15:33:02.642862+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

62 extracted references · 29 canonical work pages

  1. [1]

    URLhttps://www.pecanstreet.org/dataport/

    Pecan street dataport, Apr 2025. URLhttps://www.pecanstreet.org/dataport/

    2025

  2. [2]

    Alden, Huangjie Gong, Cristinel Ababei, and Dan M

    Rosemary E. Alden, Huangjie Gong, Cristinel Ababei, and Dan M. Ionel. Lstm forecasts for smart home electricity usage. In2020 9th International Conference on Renewable Energy Research and Application (ICRERA), pages 434–438, 2020. doi: 10.1109/ICRERA49962.2020. 9242804

    work page doi:10.1109/icrera49962.2020 2020

  3. [3]

    Clue: Safe model-based rl hvac control using epistemic uncertainty estimation

    Zhiyu An, Xianzhong Ding, Arya Rathee, and Wan Du. Clue: Safe model-based rl hvac control using epistemic uncertainty estimation. BuildSys ’23, page 149–158, New York, NY , USA, 2023. Association for Computing Machinery. ISBN 9798400702303. doi: 10.1145/3600100.3623742

    work page doi:10.1145/3600100.3623742 2023

  4. [4]

    Hierarchical transfer learning with applications to electricity load forecasting.International Journal of Forecasting, 40(2):641–660, 2024

    Anestis Antoniadis, Solenne Gaucher, and Yannig Goude. Hierarchical transfer learning with applications to electricity load forecasting.International Journal of Forecasting, 40(2):641–660, 2024

    2024

  5. [5]

    ANSI/ASHRAE Standard 140-2023: Method of Test for Evaluating Building Performance Simulation Software

    ASHRAE. ANSI/ASHRAE Standard 140-2023: Method of Test for Evaluating Building Performance Simulation Software. Technical report, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Atlanta, GA, 2023

    2023

  6. [6]

    A hot dataset: 150,000 buildings for hvac operations transfer research

    Anaïs Berkes, Yoshua Bengio, David Rolnick, and Donna Vakalis. A hot dataset: 150,000 buildings for hvac operations transfer research. InProceedings of the 12th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, BuildSys ’25, page 171–180, New York, NY , USA, 2025. Association for Computing Machinery. ISBN ...

    work page doi:10.1145/3736425.3770110 2025

  7. [7]

    Heapo – an open dataset for heat pump optimization with smart electricity meter data and on-site inspection protocols

    Tobias Brudermueller, Elgar Fleisch, Marina González Vayá, and Thorsten Staake. Heapo – an open dataset for heat pump optimization with smart electricity meter data and on-site inspection protocols. InProceedings of the 16th ACM International Conference on Future and Sustainable Energy Systems, E-Energy ’25, page 699–711, New York, NY , USA, 2025. Associa...

    work page doi:10.1145/3679240.3734637 2025

  8. [8]

    Two-layer transfer-learning-based architecture for short-term load forecasting.IEEE Transactions on Industrial Informatics, 16(3):1722–1732, 2019

    Long Cai, Jie Gu, and Zhijian Jin. Two-layer transfer-learning-based architecture for short-term load forecasting.IEEE Transactions on Industrial Informatics, 16(3):1722–1732, 2019

    2019

  9. [9]

    Enrico Casella, Simone Silvestri, D. A. Baker, and Sajal K. Das. A human-centered power conservation framework based on reverse auction theory and machine learning.ACM Trans. Cyber-Phys. Syst., 8(3), July 2024. ISSN 2378-962X. doi: 10.1145/3656348

    work page doi:10.1145/3656348 2024

  10. [10]

    Stochastic simulation of occupant-driven energy use in a bottom-up residential building stock model.Applied Energy, 325:119890, 2022

    Jianli Chen, Rajendra Adhikari, Eric Wilson, Joseph Robertson, Anthony Fontanini, Ben Polly, and Opeoluwa Olawale. Stochastic simulation of occupant-driven energy use in a bottom-up residential building stock model.Applied Energy, 325:119890, 2022. ISSN 0306-2619. doi: https://doi.org/10.1016/j.apenergy.2022.119890

    work page doi:10.1016/j.apenergy.2022.119890 2022

  11. [11]

    Yoder, Sercan O

    Si-An Chen, Chun-Liang Li, Nathanael C. Yoder, Sercan O. Arik, and Tomas Pfister. Tsmixer: An all-mlp architecture for time series forecasting.Transactions on Machine Learning Research (TMLR), 2023

    2023

  12. [12]

    The impact of extreme weather on peak electricity demand from homes heated by air source heat pumps.Energy Sources, Part B: Economics, Planning, and Policy, 16(8):707–718, 2021

    Michael Chesser, Padraic O’Reilly, Padraig Lyons, and Paula Carroll. The impact of extreme weather on peak electricity demand from homes heated by air source heat pumps.Energy Sources, Part B: Economics, Planning, and Policy, 16(8):707–718, 2021. 10

    2021

  13. [14]

    Fouda, and Mohammad G

    Shijon Das, Mostafa M. Fouda, and Mohammad G. Abdo. Short-term load forecasting using gru-lgbm fusion. In2024 International Conference on Smart Applications, Communications and Networking (SmartNets), pages 1–6, 2024. doi: 10.1109/SmartNets61466.2024.10577637

    work page doi:10.1109/smartnets61466.2024.10577637 2024

  14. [15]

    Eckelman

    Hengfang Deng, David Fannon, and Matthew J. Eckelman. Predictive modeling for us commer- cial building energy use: A comparison of existing statistical and machine learning algorithms using cbecs microdata.Energy and Buildings, 163:34–43, 2018. ISSN 0378-7788. doi: https://doi.org/10.1016/j.enbuild.2017.12.031

    work page doi:10.1016/j.enbuild.2017.12.031 2018

  15. [16]

    Dhruva and Simone Silvestri

    Jainam J. Dhruva and Simone Silvestri. Surrogate-assisted adaptive evolutionary algorithm for efficient hvac control in home energy management systems.IEEE Access, 14:56010–56025,

  16. [17]

    doi: 10.1109/ACCESS.2026.3682562

    work page doi:10.1109/access.2026.3682562 2026

  17. [18]

    Buildingsbench: A large-scale dataset of 900k buildings and benchmark for short-term load forecasting

    Patrick Emami, Abhijeet Sahu, and Peter Graf. Buildingsbench: A large-scale dataset of 900k buildings and benchmark for short-term load forecasting. InThirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2023

    2023

  18. [19]

    An efficient P300-based brain–computer interface for disabled subjects

    Xi Fang, Guangcai Gong, Guannan Li, Liang Chun, Pei Peng, and Xing Shi. Transferability investigation of a sim2real deep transfer learning framework for cross-building energy prediction. Energy and Buildings, 287:112968, 2023. ISSN 0378-7788. doi: https://doi.org/10.1016/j. enbuild.2023.112968

    work page doi:10.1016/j 2023

  19. [20]

    Multi-zone indoor temperature prediction with lstm-based sequence to sequence model.Energy and Buildings, 245:111053, 2021

    Zhen Fang, Nicolas Crimier, Lisa Scanu, Alphanie Midelet, Amr Alyafi, and Benoit Delinchant. Multi-zone indoor temperature prediction with lstm-based sequence to sequence model.Energy and Buildings, 245:111053, 2021. doi: 10.1016/j.enbuild.2021.111053

    work page doi:10.1016/j.enbuild.2021.111053 2021

  20. [21]

    Model-agnostic meta-learning for fast adap- tation of deep networks

    Chelsea Finn, Pieter Abbeel, and Sergey Levine. Model-agnostic meta-learning for fast adap- tation of deep networks. InInternational conference on machine learning, pages 1126–1135. PMLR, 2017

    2017

  21. [22]

    REFIT Smart Home dataset

    Steven Firth, Tom Kane, Vanda Dimitriou, Tarek Hassan, Farid Fouchal, Michael Cole- man, and Lynda Webb. REFIT Smart Home dataset. 6 2017. doi: 10.17028/ rd.lboro.2070091.v1. URL https://repository.lboro.ac.uk/articles/dataset/ REFIT_Smart_Home_dataset/2070091

    arXiv 2017

  22. [23]

    Nan Gao, Wei Shao, Mohammad Saiedur Rahaman, Jun Zhai, Klaus David, and Flora D. Salim. Transfer learning for thermal comfort prediction in multiple cities.Building and Environment, 195:107725, 2021. ISSN 0360-1323. doi: https://doi.org/10.1016/j.buildenv.2021.107725

    work page doi:10.1016/j.buildenv.2021.107725 2021

  23. [24]

    Transfer learning in transformer-based demand forecasting for home energy management system

    Gargya Gokhale, Jonas Van Gompel, Bert Claessens, and Chris Develder. Transfer learning in transformer-based demand forecasting for home energy management system. BuildSys ’23, page 458–462, New York, NY , USA, 2023. Association for Computing Machinery. ISBN 9798400702303. doi: 10.1145/3600100.3626635

    work page doi:10.1145/3600100.3626635 2023

  24. [25]

    Transferrable model-agnostic meta-learning for short-term household load forecasting with limited training data.IEEE Transactions on Power Systems, 37(4):3177–3180, 2022

    Yu He, Fengji Luo, and Gianluca Ranzi. Transferrable model-agnostic meta-learning for short-term household load forecasting with limited training data.IEEE Transactions on Power Systems, 37(4):3177–3180, 2022. doi: 10.1109/TPWRS.2022.3169389

    work page doi:10.1109/tpwrs.2022.3169389 2022

  25. [26]

    UCI Machine Learning Repository (2006), DOI: https://doi.org/10.24432/C58K54

    Georges Hebrail and Alice Berard. Individual Household Electric Power Consumption. UCI Machine Learning Repository, 2006. DOI: https://doi.org/10.24432/C58K54

    work page doi:10.24432/c58k54 2006

  26. [27]

    Comprehensive energy demand and usage data for building automation

    Philipp Heer, Curdin Derungs, Benjamin Huber, Felix Bünning, Reto Fricker, Sascha Stoller, and Björn Niesen. Comprehensive energy demand and usage data for building automation. Scientific Data, 11(1):469, May 2024. ISSN 2052-4463. doi: 10.1038/s41597-024-03292-2. URLhttps://doi.org/10.1038/s41597-024-03292-2

    work page doi:10.1038/s41597-024-03292-2 2024

  27. [28]

    Long short-term memory

    Sepp Hochreiter and Jürgen Schmidhuber. Long short-term memory.Neural Comput., 9(8): 1735–1780, November 1997. ISSN 0899-7667. doi: 10.1162/neco.1997.9.8.1735

    work page doi:10.1162/neco.1997.9.8.1735 1997

  28. [29]

    Building performance database

    Josh Kace, Travis Walter, and Earth Advantage. Building performance database. Open Energy Data Initiative (OEDI), Office of Energy Efficiency & Renewable En- ergy, https://data.openei.org/submissions/145, 2014. URL https://data.openei.org/ submissions/145. Accessed: 2026-04-29. 11

    2014

  29. [30]

    A synthetic energy dataset for non-intrusive load monitoring in households.Scientific data, 7(1): 108, 2020

    Christoph Klemenjak, Christoph Kovatsch, Manuel Herold, and Wilfried Elmenreich. A synthetic energy dataset for non-intrusive load monitoring in households.Scientific data, 7(1): 108, 2020

    2020

  30. [31]

    Forecasting residential heating and electricity demand with scalable, high-resolution, open-source models.arXiv preprint arXiv:2505.22873, 2025

    Stephen J Lee and Cailinn Drouin. Forecasting residential heating and electricity demand with scalable, high-resolution, open-source models.arXiv preprint arXiv:2505.22873, 2025

    Pith/arXiv arXiv 2025

  31. [32]

    Building thermal dynamics modeling with deep learning exploiting large residential smart thermostat dataset

    Han Li, Giuseppe Pinto, Alfonso Capozzoli, and Tianzhen Hong. Building thermal dynamics modeling with deep learning exploiting large residential smart thermostat dataset. InProceedings of the 9th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, pages 242–245, 2022

    2022

  32. [33]

    Transfer learning applied to drl-based heat pump control to leverage microgrid energy efficiency.Smart Energy, 3:100044, 2021

    Paulo Lissa, Michael Schukat, Marcus Keane, and Enda Barrett. Transfer learning applied to drl-based heat pump control to leverage microgrid energy efficiency.Smart Energy, 3:100044, 2021

    2021

  33. [34]

    Lawrence berkeley national lab building 59, 01 2021

    Na Luo and Tianzhen Hong. Lawrence berkeley national lab building 59, 01 2021. URL https://www.osti.gov/biblio/1762808

    arXiv 2021

  34. [35]

    Ecobee donate your data 1,000 homes in 2017, 03 2022

    Na Luo and Tianzhen Hong. Ecobee donate your data 1,000 homes in 2017, 03 2022. URL https://www.osti.gov/biblio/1854924

    arXiv 2017

  35. [36]

    The building data genome project: An open, public data set from non-residential building electrical meters.Energy Procedia, 122:439 – 444, 2017

    Clayton Miller and Forrest Meggers. The building data genome project: An open, public data set from non-residential building electrical meters.Energy Procedia, 122:439 – 444, 2017. ISSN 1876-6102. doi: https://doi.org/10.1016/j.egypro.2017.07.400. {CISBAT} 2017 International ConferenceFuture Buildings & Districts – Energy Efficiency from Nano to Urban Scale

    work page doi:10.1016/j.egypro.2017.07.400 2017

  36. [37]

    Hobson, Zixiao Shi, and Forrest Meggers

    Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, and Forrest Meggers. The building data genome project 2, energy meter data from the ashrae great energy predictor iii competition.Scientific Data, 7(1), Oct 2020. doi: 10.1038/s41597-020-00712-x

    work page doi:10.1038/s41597-020-00712-x 2020

  37. [38]

    Ozan Baris Mulayim, Pengrui Quan, Liying Han, Xiaomin Ouyang, Dezhi Hong, Mario Bergés, and Mani Srivastava. Are time series foundation models ready to revolutionize predictive building analytics? InProceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, pages 169–173, 2024

    2024

  38. [39]

    Transfer learning and explainable ai for heating load forecasting: A large-scale benchmark with shap-based static features.Energy and AI, page 100722, 2026

    Alexander Neubauer, Mengbo Yu, Pedram Babakhani, Stefan Brandt, and Martin Kriegel. Transfer learning and explainable ai for heating load forecasting: A large-scale benchmark with shap-based static features.Energy and AI, page 100722, 2026

    2026

  39. [40]

    Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam

    Yuqi Nie, Nam H. Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. InInternational Conference on Learning Representations (ICLR), 2023

    2023

  40. [41]

    Energyplus™, 09 2017

    National Renewable Energy Laboratory (NREL). Energyplus™, 09 2017. URL https: //www.osti.gov/biblio/1395882

    arXiv 2017

  41. [42]

    Comstock reference documentation (v.1)

    Andrew Parker, Henry Horsey, Matthew Dahlhausen, Marlena Praprost, Christopher CaraDonna, Amy LeBar, and Lauren Klun. Comstock reference documentation (v.1). Technical report, National Renewable Energy Laboratory (NREL), Golden, CO (United States), 03 2023. URL https://www.osti.gov/biblio/1967948

    arXiv 2023

  42. [43]

    Climate-adaptive energy forecasting in green buildings via attention-enhanced seq2seq transfer learning.Scientific Reports, 15(1): 31829, 2025

    Fang Peng, Tao Su, Qing Zeng, and Xiaojuan Han. Climate-adaptive energy forecasting in green buildings via attention-enhanced seq2seq transfer learning.Scientific Reports, 15(1): 31829, 2025

    2025

  43. [44]

    Building timeseries dataset: Empowering large-scale building analytics

    Arian Prabowo, Xiachong Lin, Imran Razzak, Hao Xue, Emily W Yap, Matthew Amos, and Flora D Salim. Building timeseries dataset: Empowering large-scale building analytics. Advances in Neural Information Processing Systems, 37:133180–133206, 2024

    2024

  44. [45]

    The ideal household energy dataset, electricity, gas, contextual sensor data and survey data for 255 uk homes.Scientific Data, 8(1):146, 2021

    Martin Pullinger, Jonathan Kilgour, Nigel Goddard, Niklas Berliner, Lynda Webb, Myroslava Dzikovska, Heather Lovell, Janek Mann, Charles Sutton, Janette Webb, et al. The ideal household energy dataset, electricity, gas, contextual sensor data and survey data for 255 uk homes.Scientific Data, 8(1):146, 2021

    2021

  45. [46]

    Sim-to-real transfer learning for large-scale short-term building energy forecasting in sustainable cities

    Heng Quan and Semiha Ergan. Sim-to-real transfer learning for large-scale short-term building energy forecasting in sustainable cities. BuildSys ’25, page 86–95, New York, NY , USA, 2025. Association for Computing Machinery. ISBN 9798400719455. doi: 10.1145/3736425.3772006. 12

    work page doi:10.1145/3736425.3772006 2025

  46. [47]

    Gentl: A general transfer learning model for building thermal dynamics

    Fabian Raisch, Thomas Krug, Christoph Goebel, and Benjamin Tischler. Gentl: A general transfer learning model for building thermal dynamics. InProceedings of the 16th ACM International Conference on Future and Sustainable Energy Systems (E-Energy), 2025. doi: 10.1145/3679240.3734589

    work page doi:10.1145/3679240.3734589 2025

  47. [48]

    London Low Carbon Dataset

    Sufiyan Rehman. London Low Carbon Dataset. 8 2025. doi: 10.6084/m9.figshare.29983723. v2. URL https://figshare.com/articles/dataset/London_Low_Carbon_Dataset/ 29983723

    work page doi:10.6084/m9.figshare.29983723 2025

  48. [49]

    Smart building temperature forecasting with probabilistic temporal fusion transformers

    Abhinav Sagar. Smart building temperature forecasting with probabilistic temporal fusion transformers. InUrbanAI: Harnessing Artificial Intelligence for Smart Cities, Dec 2025

    2025

  49. [50]

    doi: 10.1109/tnn.2008.2005605

    Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. The graph neural network model.IEEE Transactions on Neural Networks, 20(1):61–80, 2009. doi: 10.1109/TNN.2008.2005605

    work page doi:10.1109/tnn.2008.2005605 2009

  50. [51]

    Real-e: A foundation benchmark for advancing robust and generalizable electricity forecasting

    Chen Shao, Michael Färber, Sebastian Pütz, Benjamin Schäfer, Yue Wang, Tobias Käfer, Zhanbo Huang, and Zhenyi Zhu. Real-e: A foundation benchmark for advancing robust and generalizable electricity forecasting. InProceedings of the 34th ACM International Conference on Information and Knowledge Management, CIKM ’25, page 6523–6527, New York, NY , USA, 2025....

    arXiv 2025

  51. [52]

    Resstock: Annual baseline results with component loads

    Andrew Speake, Eric Wilson, Yueyue Zhou, and Scott Horowitz. Resstock: Annual baseline results with component loads. Open Energy Data Initiative (OEDI), National Renewable Energy Lab, https://doi.org/10.25984/2001060, 2023. URL https://data.openei.org/ submissions/5959. Accessed: 2026-04-28

    work page doi:10.25984/2001060 2023

  52. [53]

    High resolution synthetic residential energy use profiles for the united states.Scientific Data, 10(1):76, 2023

    Swapna Thorve, Young Yun Baek, Samarth Swarup, Henning Mortveit, Achla Marathe, Anil Vullikanti, and Madhav Marathe. High resolution synthetic residential energy use profiles for the united states.Scientific Data, 10(1):76, 2023

    2023

  53. [54]

    ElectricityLoadDiagrams20112014

    Artur Trindade. ElectricityLoadDiagrams20112014. UCI Machine Learning Repository, 2015. DOI: https://doi.org/10.24432/C58C86

    work page doi:10.24432/c58c86 2015

  54. [55]

    Department of Energy

    U.S. Department of Energy. Building america climate-specific guid- ance, n.d. URL https://www.energy.gov/cmei/buildings/ building-america-climate-specific-guidance. Accessed: 2026-05-06

    2026

  55. [56]

    Gomez, Łukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. InAdvances in Neural Informa- tion Processing Systems (NeurIPS), 2017

    2017

  56. [57]

    Timexer: Empowering transformers for time series forecasting with exogenous variables

    Yuxuan Wang, Haixu Wu, Jiaxiang Dong, Guo Qin, Haoran Zhang, Yong Liu, Yunzhong Qiu, Jianmin Wang, and Mingsheng Long. Timexer: Empowering transformers for time series forecasting with exogenous variables. InAdvances in Neural Information Processing Systems (NeurIPS), 2024

    2024

  57. [58]

    Efficient residential electric load forecasting via transfer learning and graph neural networks.IEEE Transactions on Smart Grid, 14(3):2423–2431, 2023

    Di Wu and Weixuan Lin. Efficient residential electric load forecasting via transfer learning and graph neural networks.IEEE Transactions on Smart Grid, 14(3):2423–2431, 2023. doi: 10.1109/TSG.2022.3208211

    work page doi:10.1109/tsg.2022.3208211 2023

  58. [59]

    Transfer learning integrating similarity analysis for short-term and long-term building energy consumption prediction.Applied energy, 365:123276, 2024

    Zhuoqun Xing, Yiqun Pan, Yiting Yang, Xiaolei Yuan, Yumin Liang, and Zhizhong Huang. Transfer learning integrating similarity analysis for short-term and long-term building energy consumption prediction.Applied energy, 365:123276, 2024

    2024

  59. [60]

    One for many: Transfer learning for building hvac control

    Shichao Xu, Yixuan Wang, Yanzhi Wang, Zheng O’Neill, and Qi Zhu. One for many: Transfer learning for building hvac control. BuildSys ’20, page 230–239, New York, NY , USA, 2020. Association for Computing Machinery. ISBN 9781450380614. doi: 10.1145/3408308.3427617

    work page doi:10.1145/3408308.3427617 2020

  60. [61]

    A physics-based domain adaptation framework for modeling and forecasting building energy systems.Data-Centric Engineering, 4: e10, 2023

    Zack Xuereb Conti, Ruchi Choudhary, and Luca Magri. A physics-based domain adaptation framework for modeling and forecasting building energy systems.Data-Centric Engineering, 4: e10, 2023. doi: 10.1017/dce.2023.8

    work page doi:10.1017/dce.2023.8 2023

  61. [62]

    An adaptive transfer learning framework for data-scarce hvac power consumption forecasting.IEEE Transactions on Sustainable Energy, 15(4):2815–2825, 2024

    Yanan Zhang, Gan Zhou, Zhan Liu, Li Huang, and Yucheng Ren. An adaptive transfer learning framework for data-scarce hvac power consumption forecasting.IEEE Transactions on Sustainable Energy, 15(4):2815–2825, 2024. doi: 10.1109/TSTE.2024.3444689. 13

    work page doi:10.1109/tste.2024.3444689 2024

  62. [63]

    FFN/token dim

    Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pages 11106–11115, 2021. Appendix A Static Home Properties Each home in RESCAST-100K is described by a ...

    2021