Pith. sign in

REVIEW 4 major objections 5 minor 54 references

Behaviour-Conditioned Neural Processes for Adaptive Residential Short-Term Load Forecasting

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Inferred household behavioural classes, embedded as a decoder-conditioning variable in a Neural Process, improve residential short-term load forecasts: a 7.9% average MAE reduction and 6.9% CRPS reduction over a label-agnostic baseline on u

desk verdict A novel ANP extension for load forecasting, but the key ablation doesn't isolate behavioural conditioning—worth a careful referee. read the letter →

arxiv 2607.16168 v1 pith:LYEXLIHJ submitted 2026-07-17 cs.LG

Ramin Soleimani , Andrea Visentin , Dirk Pesch This is my paper · ORCID
classification cs.LG
keywords residentialloadforecastingneuralprocessesbehaviouralclusteringfeature-wiselinearmodulationprobabilisticsmartmeterdataconsumerheterogeneitysemi-supervisedlatentvariables
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 asks whether the behavioural routine behind a household's electricity use can improve near-term load forecasts if that behaviour is inferred from past data and used inside the forecasting model, rather than as an external grouping step. It proposes a behaviour-conditioned Attentive Neural Process: a continuous latent variable captures shared uncertainty across households, while a discrete behavioural latent variable, inferred from the observed context, conditions the decoder through feature-wise modulation and a prompt token. Clustering of complete load profiles supplies weak supervision during training only; at test time the class is inferred from context alone. On user-disjoint SGSC data, the soft-label variant reduces MAE by 7.9% and CRPS by 6.9% on average versus a label-agnostic ANP baseline, with the largest gains when context is short, and achieves lower RMSE than fixed-window deterministic baselines at every horizon.

What carries the argument

The central mechanism is the dual-latent Attentive Neural Process: a continuous latent variable z models shared functional uncertainty over plausible load trajectories, while a discrete behavioural latent c, drawn from a context-conditioned categorical prior pψ(c|DC), conditions a Transformer-style decoder through (i) HyperFiLM — a hypernetwork that maps the class embedding to per-layer (γℓ, βℓ) scale/shift parameters applied after LayerNorm, AdaLN(h|c)=(1+γ(c))LN(h)+β(c) — and (ii) a prompt token prepended to the decoder sequence. K-means soft memberships of complete profiles act as the teacher for c during training (with the mixing coefficient annealed from 1 to 0.15), and a semi-supervise

What would settle it

Retrain the soft-label variant with teacher labels randomly permuted across profiles, breaking any link between cluster and consumption behaviour, under the same splits and hyperparameters; if the MAE/CRPS improvements over the label-agnostic ANP persist, the reported gains are a capacity effect and not behaviour conditioning. Equivalently, measure the label-inference network's agreement with complete-profile K-means assignments on the user-disjoint test set: chance-level agreement would indicate the conditioning signal carries no behavioural information.

Watch

Extended reading notes

Core claim

The paper establishes that behavioural structure discovered offline by K-means clustering of complete two-day load profiles can be distilled into soft class memberships, learned by a label-inference network from partial context, and injected into an Attentive Neural Process decoder via HyperFiLM (per-layer scaling and shifting after layer normalisation) and a prepended prompt token. This conditional decoder, with the continuous latent z kept class-agnostic, yields consistent MAE and CRPS improvements over a label-agnostic ANP across horizons H=1–8 and context lengths C≥48, with the largest improvements (18.3% MAE, 18.8% CRPS at H=1) when context C<60; the soft-label variant outperforms the h

Load-bearing premise

The gains depend on the premise that K-means clusters of complete two-day load profiles define a behavioural structure that a label-inference network can reconstruct from a partial day of context, and that this inferred class adds forecasting information beyond what the continuous latent variable and the attention path already provide.

Editorial extensions

If this is right

  • A single Neural Process model can serve many households, varying contexts, and horizons without per-household retraining, shifting its predictive distribution with inferred consumption style.
  • Behavioural conditioning pays off most when history is short: at H=1 and C<60 the soft variant improves over ANP by 18.3% MAE and 18.8% CRPS, which is relevant for cold-start forecasting of new households.
  • Soft behavioural assignments outperform hard ones, implying that uncertainty in regime membership should be propagated into the forecast rather than collapsed to a single class.
  • The lower RMSE at all horizons versus deterministic baselines (13.4–18.4% below the strongest baselines) suggests fewer large prediction errors, which matters for operational decisions sensitive to large deviations.
  • Test-time conditioning relies on context-inferred class distributions rather than complete-profile clusters, avoiding label leakage from future observations.

Reading between the lines

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

  • The gains suggest behavioural archetypes learned from complete profiles are partially recoverable from a single day of history; a direct test would measure mutual information between the context-inferred class and the complete-profile cluster on held-out contexts.
  • The method offers a general recipe: cluster the training set on the full target variable, then train a latent-conditioned function approximator to re-infer the cluster from inputs; whether it helps depends on the cluster structure being predictive of future values beyond what a continuous latent captures.
  • The soft-versus-hard gap hints that a fully Bayesian treatment of the discrete regime, marginalising over c at inference rather than using the prior distribution's mode, could yield further gains.
  • A clean falsifier: randomise the teacher cluster assignments; if MAE/CRPS gains over the label-agnostic baseline persist, the benefit comes from added model capacity, not from behavioural semantics.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a behaviour-conditioned Attentive Neural Process (ANP) for residential short-term load forecasting. The model augments a standard ANP with a discrete behavioural latent variable c, inferred from the observed context by a label-inference network, and uses this variable to condition a Transformer decoder through HyperFiLM modulation and prompt-token injection. K-means soft labels computed from complete daily profiles provide weak supervision during training, while at test time only the context-inferred class distribution is used. Experiments on 400 SGSC households with user-disjoint splits, context lengths C in {48,...,95}, and horizons H=1,...,8 report average MAE/CRPS reductions of 7.9%/6.9% over a label-agnostic ANP baseline, and lower RMSE than fixed-window deterministic baselines at all horizons.

Significance. If the reported improvements are genuinely attributable to behavioural conditioning, this is a useful and timely contribution: it embeds inferred behavioural structure inside a single probabilistic forecasting model, rather than using clusters as an external grouping signal, and it supports uncertainty-aware, context-adaptive forecasting across heterogeneous households. The evaluation has good features: user-disjoint splits, variable context lengths and horizons, CRPS as a proper scoring rule, and explicit attention to avoiding future-label leakage at test time. However, the current experimental design does not isolate the behavioural signal from model capacity, auxiliary supervision, or input information, so the central claim is not yet established.

major comments (4)
  1. [Section 6.1] The ablation is not capacity-matched. The text says the variants 'share the same architecture... and differ only in how behavioural information is incorporated,' but the ANP baseline disables the label-inference network, the HyperFiLM/prompt conditioning branch, and the supervised cross-entropy term. The comparison therefore changes at least three variables: parameter count, auxiliary supervision, and the conditioning input. The 7.9%/6.9% gains could be a capacity/regularization artefact rather than evidence for behavioural conditioning. Please add controls such as (i) ANP with the same added parameters but using a constant or random conditioning vector, (ii) ANP with the auxiliary cluster-prediction loss but no FiLM conditioning, and (iii) test-time shuffling of the inferred class to verify that the behavioural signal itself matters.
  2. [Section 6.3, Table 4] The comparison with deterministic baselines is confounded by input length. The NP models are evaluated over context lengths C in {48,...,95} and averaged, while the fixed-window baselines receive exactly 48 half-hour observations. The statement that 'all models get the same input information' is therefore incorrect: the average NP context is substantially longer than 48 steps. This alone could explain the RMSE advantage. Please report NP results restricted to C=48 for the baseline comparison, or train/evaluate the deterministic baselines on the same variable context-length distribution (e.g., a direct model with 95-step inputs), so the comparison reflects modelling paradigm rather than information amount.
  3. [Section 5.2, Eq. (21)] There is a serious notation/semantics issue in the conditioning distribution. Eq. (21) defines pcond(x) = alpha pteach(x) + (1-alpha) qψ(c|DC), but qψ was defined as the target-conditioned posterior qψ(c|DT). If the decoder conditioning during training actually uses the target-conditioned posterior, then future target values leak into the conditioning signal and the reported gains may be an artefact of that leak, undermining the causal train/test consistency claim. If, as the text states, the context-conditioned prior pψ(c|DC) is intended, correct Eq. (21) and clarify that qψ(c|DT) is used only for the auxiliary loss in Eq. (16). This point is load-bearing and must be resolved.
  4. [Section 7, Table 2] The behavioural classes are K-means clusters of the same z-scored load profiles, with a silhouette value of only 0.042 and no external semantic validation. Since the label-inference network is trained to reconstruct these clusters from the same context that the deterministic attention path already encodes, it is unclear whether the inferred c contributes information beyond simple context statistics (e.g., mean, variance, peak timing). Please include a control that conditions on random labels or on continuous context summaries with matched capacity. Without such a control, the improvements cannot be attributed specifically to recoverable behavioural structure rather than to the extra conditioning mechanism.
minor comments (5)
  1. [Eq. (21)] Beyond the major issue, the notation qψ(c|DC) is undefined because qψ is a target-conditioned posterior; if the prior is intended, use pψ(c|DC) consistently throughout Section 5.2.
  2. [Section 4] The profiles are described as 'two-day load profiles' but W=104 half-hour steps equals 52 hours. Please harmonize the terminology ('two-day plus four hours' or '52-hour profiles').
  3. [Tables 3 and 4] All results are reported as single point estimates without error bars, confidence intervals, or significance tests. Given the small margins (e.g., FiLM-ANP-Soft vs. SVR at H=5 or vs. LSTM at H=2), multiple seeds or a paired test would materially strengthen the claims.
  4. [Section 6.3] There is a typo: 'ANP–Soft consistently achieves the lowest values' should read 'FiLM-ANP–Soft'.
  5. [References] References [1] and [47] appear to be the same Kong et al. paper; please consolidate.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the behavioural conditioning is not fitted to the forecast target and the headline improvements are held-out empirical outcomes.

full rationale

The paper's derivation chain is not circular. The forecast target YT is never used to fit the behavioural latent c at inference time: clustering-derived soft labels are computed from complete profiles and used only as weak supervision during training, while test-time decoder conditioning uses the context-inferred distribution pψ(c|DC), as stated in Sections 5.2 and 7. The likelihood pθ(YT | XT, {r*t}, z, c) is trained with a standard semi-supervised ELBO (Eqs. 12–18), and no equation reduces the predicted YT to the fitted labels or to the K-means assignment. The reported 7.9% MAE and 6.9% CRPS improvements are measured on user-disjoint held-out households and future horizons, so they are empirical outcomes, not fitted values renamed as predictions. The only self-citation ([14], an ANP load-forecasting paper) is background motivation and is not load-bearing for the central claim; the ANP architecture itself is independently established by [12]. The limitations section explicitly acknowledges that the K-means labels are 'proxy labels derived from K-means clustering, rather than externally validated semantic categories', which is an honest construct-validity caveat, not a circular definition. The skeptical concern that the ANP ablation is not capacity-matched is an experimental identification limitation, not a circularity: it does not exhibit an equation or fitted parameter that makes the claimed improvement true by construction. Therefore no circular step can be quoted and the score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 1 invented entities

The central result rests on variational NP machinery (standard), on K-means-derived soft behaviour labels being informative and causally inferable (domain-specific, not externally validated), and on hand-selected or unreported hyperparameters (K, tau, alpha schedule, beta, lambda_sup). The discrete behavioural latent is a model construct with no independent evidence.

free parameters (4)
  • Number of behavioural clusters K = 3
    Selected by composite rank sum over inertia, silhouette, Davies–Bouldin, and Calinski–Harabasz (Table 2) on the full 400-user dataset; data-dependent model selection.
  • Soft-label temperature tau = 0.15
    Eq. (19); chosen by hand to make K-means distance-derived soft labels 'informative but not degenerate'; evaluated on the same data.
  • Blending schedule alpha_min and warm-up epochs = 0.15 / 60
    Section 5.2, Eq. (21): anneals the clustering teacher from 1 to 0.15 over the first 60 epochs; hand-set and not ablated.
  • Objective weights beta, lambda_sup and Gumbel temperature schedule
    Appear in Eqs. (8), (17), (18) and the text, but their values are not reported, so the exact loss and sampling temperature cannot be reproduced.
assumptions (4)
  • standard math Standard variational NP/ANP assumptions: Gaussian likelihood, mean-field factorized posterior q(z,c)=q(z)q(c), and Jensen-based ELBOs in Eqs. (12) and (14).
    Used throughout Section 3; no formal proof or machine-checked verification is provided.
  • domain assumption K-means on z-normalized 104-step profiles yields stable behavioural archetypes, and the soft label transform (Eq. 19) preserves hard assignments while remaining informative.
    Section 4, Table 2; no external semantic labels or independent validation of the archetypes.
  • ad hoc to paper A label inference network trained on partial contexts generalizes to unseen households and recovers the complete-profile cluster structure needed for useful FiLM conditioning.
    Section 5.2 and the test-time protocol; this is the core 'behavioural inference' bet of the paper.
  • domain assumption The curated 400-user subset with complete 2013 data, z-score normalization, and user-disjoint splits represents the deployment population for residential STLF.
    Section 4 and Section 6.1; no missing-data or irregular-sampling robustness is evaluated (also acknowledged in Section 7).
invented entities (1)
  • Discrete behavioural latent variable c with K=3 classes
    purpose: Decoder conditioning via HyperFiLM and prompt token to adapt forecasts to inferred behavioural regime
    The classes are K-means clusters of the same load profiles used for training; the paper itself calls them proxy labels without external semantic validation, so there is no falsifiable handle outside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Behaviour-Conditioned Neural Processes for Adaptive Residential Short-Term Load Forecasting." pith.science (2026). https://pith.science/paper/LYEXLIHJ

@misc{pith2026260716168,
  author       = {Pith},
  title        = {Pith review of: Behaviour-Conditioned Neural Processes for Adaptive Residential Short-Term Load Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LYEXLIHJ}},
  note         = {Machine review of arXiv:2607.16168}
}
read the original abstract

Residential short-term load forecasting (STLF) is challenging because household demand is heterogeneous, temporally variable, and shaped by diverse behavioural routines. This work investigates whether inferred behavioural structure can be embedded within the forecasting mechanism of a Neural Process-based probabilistic model, rather than used only as an external grouping signal, for context-conditioned residential STLF. We propose a behaviour-conditioned Attentive Neural Process framework that treats each load profile as a forecasting task. Behavioural structure is represented by a discrete latent variable inferred from the available context and used for behaviour-conditioned decoder conditioning, while a continuous latent variable captures shared functional uncertainty across heterogeneous profiles. To enable conditioning without ground-truth behavioural labels, clustering-derived information provides weak supervision during training, whereas test-time conditioning relies only on context-inferred class distributions. Experiments on the Smart Grid, Smart City (SGSC) dataset use user-disjoint train/validation/test splits, variable context lengths, and multi-step forecast horizons, with comparisons against a label-agnostic ANP baseline and fixed-window deterministic STLF baselines. The proposed variants improve MAE and CRPS over ANP across horizons and context settings, with the largest gains under limited context. The best-performing variant achieves average reductions of 7.9% in MAE and 6.9% in CRPS relative to ANP. Compared with fixed-window baselines, this variant achieves lower RMSE across all evaluated horizons while maintaining competitive MAE, suggesting fewer large prediction deviations under heterogeneous consumption patterns. These results support single-model, uncertainty-aware forecasting across heterogeneous households, contexts, and horizons.

Figures

Figures reproduced from arXiv: 2607.16168 by the authors.

Figure 1
Figure 1. Overview of the proposed behaviour-conditioned Attentive Neural Process frame [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. K-means K-selection criteria (inertia, silhouette, Davies–Bouldin, and Calinski– Harabasz) evaluated on load profiles. profiles fall between clusters or share features from multiple archetypes. To reflect this, we turn the K-means results into a soft probabilistic label for each frame. For a profile x, we first compute its Euclidean distance to each centroid, dk(x) = [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗
Figure 3
Figure 3. Representative normalised load profiles given by K-means cluster centroids (52- [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Percentage reduction in MAE and CRPS of FiLM-ANP–Soft relative to ANP [PITH_FULL_IMAGE:figures/full_fig_p030_4.png]
Figure 5
Figure 5. Figure 5: Relative improvement of FiLM-ANP–Soft over ANP as a function of context [PITH_FULL_IMAGE:figures/full_fig_p031_5.png]
Figure 6
Figure 6. Figure 6: Context-driven forecasting progression for the FiLM-ANP–Soft model under [PITH_FULL_IMAGE:figures/full_fig_p034_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 9 canonical work pages

  1. [2]

    S., Chau, T

    Rubasinghe, O., Zhang, T., Zhang, X., Choi, S. S., Chau, T. K., Chow, Y., Fernando, T., Iu, H. H. C., Highly accurate peak and valley predic- tion short-term net load forecasting approach based on decomposition for power systems with high PV penetration, Applied Energy 333 (2023) 120641. https://doi.org/10.1016/j.apenergy.2023.120641

    arXiv 2023

  2. [3]

    J., Roy, P., Dey, P., Salsabil, N

    Hasan, M., Mifta, Z., Papiya, S. J., Roy, P., Dey, P., Salsabil, N. A., Chowdhury, N.-U.-R., Farrok, O., A state-of-the-art comparative re- view of load forecasting methods: Characteristics, perspectives, and ap- plications, Energy Conversion and Management: X 26 (2025) 100922. https://doi.org/10.1016/j.ecmx.2025.100922

    arXiv 2025

  3. [4]

    https://doi.org/10.1016/j.adapen.2023.100142

    Gao, J., Chen, Y., Hu, W., Zhang, D., An adaptive deep-learning load forecasting framework by integrating Transformer and domain knowl- edge (Adaptive-TgDLF), Advances in Applied Energy 10 (2023) 100142. https://doi.org/10.1016/j.adapen.2023.100142

    arXiv 2023

  4. [5]

    https://doi.org/10.1109/TSG.2024.3452490

    Wang, Z., Zhang, H., Yang, R., Chen, Y., Improving model gen- eralization for short-term customer load forecasting with causal in- ference, IEEE Transactions on Smart Grid 16 (1) (2025) 424–436. https://doi.org/10.1109/TSG.2024.3452490

    arXiv 2025

  5. [6]

    J., Viljanen, M., Modeling and fore- casting energy consumption for heterogeneous buildings using a physical–statistical approach, Applied Energy 144 (2015) 261–275

    Lü, X., Lu, T., Kibert, C. J., Viljanen, M., Modeling and fore- casting energy consumption for heterogeneous buildings using a physical–statistical approach, Applied Energy 144 (2015) 261–275. https://doi.org/10.1016/j.apenergy.2014.12.019 38

    work page doi:10.1016/j.apenergy.2014.12.019 2015

  6. [7]

    E., A data- driven methodology for deriving electricity consumption typolo- gies from smart meters, Energy Reports 14 (2025) 2420–2434

    Quesada, C., Montero-Manso, P., Pflugradt, N., Astigarraga, L., Merveille, C., Casado-Mansilla, D., Borges, C. E., A data- driven methodology for deriving electricity consumption typolo- gies from smart meters, Energy Reports 14 (2025) 2420–2434. https://doi.org/10.1016/j.egyr.2025.09.002

    work page doi:10.1016/j.egyr.2025.09.002 2025

  7. [8]

    A., Time se- ries clustering based on prediction accuracy of global fore- casting models, Knowledge-Based Systems 323 (2025) 113649

    López-Oriona, Á., Montero-Manso, P., Vilar, J. A., Time se- ries clustering based on prediction accuracy of global fore- casting models, Knowledge-Based Systems 323 (2025) 113649. https://doi.org/10.1016/j.knosys.2025.113649

    arXiv 2025

  8. [9]

    Peebles, W., Xie, S., Scalable diffusion models with transformers, In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) (2023) 4195–4205

    2023

Show all 54 references
  1. [10]

    J., Es- lami, S

    Garnelo, M., Schwarz, J., Rosenbaum, D., Viola, F., Rezende, D. J., Es- lami, S. M. A., Teh, Y. W., Neural processes, arXiv:1807.01622 (2018)

    2018 arXiv

  2. [11]

    C. K. I. Williams and C. E. Rasmussen,Gaussian Processes for Regres- sion, in *Advances in Neural Information Processing Systems 8 (NIPS 1995)*, pp. 514–520, 1995

    1995

  3. [12]

    Kim, H., Mnih, A., Schwarz, J., Garnelo, M., Eslami, S. M. A., Rosen- baum, D., Vinyals, O., Teh, Y. W., Attentive neural processes, In: In- ternational Conference on Learning Representations (ICLR) (2019)

    2019

  4. [13]

    https://doi.org/10.1609/aaai.v32i1.11671

    Perez, E., Strub, F., de Vries, H., Dumoulin, V., Courville, A., FiLM: Visual reasoning with a general conditioning layer, In: Proceed- ings of the AAAI Conference on Artificial Intelligence 32 (1) (2018). https://doi.org/10.1609/aaai.v32i1.11671

    2018 doi

  5. [14]

    https://doi.org/10.1109/SMARTCOMP65954.2025.00105

    Soleimani, R., Chung, S., Pesch, D., Short-term load fore- casting with attentive neural processes: Adaptivity and uncertainty estimation, In: IEEE International Confer- ence on Smart Computing (SMARTCOMP) (2025) 456–461. https://doi.org/10.1109/SMARTCOMP65954.2025.00105

    2025

  6. [15]

    B., A comprehensive review of deep learning-based short-term load forecasting for residential buildings, Renewable and Sustainable Energy Reviews 187 (2024) 114031

    Eren, S., Küçükdemiral, İ. B., A comprehensive review of deep learning-based short-term load forecasting for residential buildings, Renewable and Sustainable Energy Reviews 187 (2024) 114031. https://doi.org/10.1016/j.rser.2023.114031 39

    2024

  7. [16]

    https://doi.org/10.1109/TSG.2017.2686012

    Shi, H., Xu, M., Li, R., Deep learning for household load forecasting—A novel pooling deep recurrent neural network, IEEE Transactions on Smart Grid 9 (5) (2018) 5271–5280. https://doi.org/10.1109/TSG.2017.2686012

    2018

  8. [17]

    https://doi.org/10.24963/ijcai.2021/374

    Lin, W., Wu, D., Residential electric load forecasting via atten- tive transfer of graph neural networks, In: International Joint Conference on Artificial Intelligence (IJCAI) (2021) 2716–2722. https://doi.org/10.24963/ijcai.2021/374

    2021 doi

  9. [18]

    J., Genton, M

    Ben Taieb, S., Huser, R., Hyndman, R. J., Genton, M. G., Forecasting uncertainty in electricity smart meter data by boosting additive quantile regression, IEEE Transactions on Smart Grid 7 (5) (2016) 2448–2455. https://doi.org/10.1109/TSG.2016.2527820

    2016

  10. [19]

    https://doi.org/10.1016/j.apenergy.2018.10.078

    Wang, Y., Gan, D., Sun, M., Zhang, N., Lu, Z., Kang, C., Probabilistic individual load forecasting using pinball loss guided LSTM, Applied En- ergy 235 (2019) 10–20. https://doi.org/10.1016/j.apenergy.2018.10.078

    2019 doi

  11. [20]

    https://doi.org/10.1016/j.apenergy.2018.02.165

    Shepero, M., van der Meer, D., Munkhammar, J., Widén, J., Residen- tial probabilistic load forecasting: A method using Gaussian processes designed for electric load data, Applied Energy 218 (2018) 159–172. https://doi.org/10.1016/j.apenergy.2018.02.165

    2018 doi

  12. [21]

    https://doi.org/10.1109/TSG.2020.2972513

    Afrasiabi, M., Mohammadi, M., Rastegar, M., Stankovic, L., Afrasiabi, S., Khazaei, M., Deep-based conditional probability density function forecasting of residential loads, IEEE Transactions on Smart Grid 11 (4) (2020) 3746–3757. https://doi.org/10.1109/TSG.2020.2972513

    2020

  13. [23]

    https://doi.org/10.1016/j.epsr.2023.109804

    Li, B., Mo, Y., Gao, F., Bai, X., Short-term probabilistic load forecast- ing method based on uncertainty estimation and deep learning model considering meteorological factors, Electric Power Systems Research 225 (2023) 109804. https://doi.org/10.1016/j.epsr.2023.109804

    2023

  14. [24]

    https://doi.org/10.1016/j.apenergy.2020.116180

    Munkhammar, J., Shepero, M., van der Meer, D., Widén, J., Probabilistic forecasting of residential electricity consumption using 40 Markov-chain mixture distributions, Applied Energy 284 (2021) 116180. https://doi.org/10.1016/j.apenergy.2020.116180

    2021

  15. [25]

    https://doi.org/10.1016/j.enbuild.2012.01.037

    McLoughlin, F., Duffy, A., Conlon, M., Characterising domestic elec- tricity consumption patterns by dwelling and occupant socio-economic variables: An Irish case study, Energy and Buildings 48 (2012) 240–248. https://doi.org/10.1016/j.enbuild.2012.01.037

    2012 doi

  16. [26]

    https://doi.org/10.1109/TPWRS.2006.873122

    Chicco, G., Napoli, R., Piglione, F., Comparisons among clustering techniques for electricity customer classification, IEEE Transactions on Power Systems 21 (2) (2006) 933–940. https://doi.org/10.1109/TPWRS.2006.873122

    2006

  17. [27]

    Chicco, G., Overview and performance assessment of the clustering methods for electrical load pattern grouping, Energy 42 (1) (2012) 68–

    2012

  18. [28]

    https://doi.org/10.1016/j.knosys.2022.109236

    Kaur, R., Gabrijelčič, D., Behavior segmentation of electricity consump- tion patterns: A cluster analytical approach, Knowledge-Based Systems 251 (2022) 109236. https://doi.org/10.1016/j.knosys.2022.109236

    2022

  19. [29]

    L., Lee, W.-J., Huang, H., Wang, D

    Quilumba, F. L., Lee, W.-J., Huang, H., Wang, D. Y., Szaba- dos, R. L., Using smart meter data to improve the accuracy of intraday load forecasting considering customer behavior similar- ities, IEEE Transactions on Smart Grid 6 (2) (2015) 911–918. https://doi.org/10.1109/TSG.2...

    2015

  20. [30]

    https://doi.org/10.1016/j.energy.2018.09.156

    Fu, X., Zeng, X.-J., Feng, P., Cai, X., Clustering-based short-term load forecasting for residential electricity under the increasing-block pricing tariffs in China, Energy 165 (2018) 76–89. https://doi.org/10.1016/j.energy.2018.09.156

    2018 doi

  21. [31]

    https://doi.org/10.17775/CSEEJPES.2020.04060

    Han, F., Pu, T., Li, M., Taylor, G., Short-term forecasting of indi- vidual residential load based on deep learning and K-means cluster- ing, CSEE Journal of Power and Energy Systems 7 (2) (2021) 261–269. https://doi.org/10.17775/CSEEJPES.2020.04060

    2021

  22. [32]

    https://doi.org/10.1016/j.egyr.2023.03.042 41

    Kim, H., Park, S., Kim, S., Time-series clustering and forecasting house- hold electricity demand using smart meter data, Energy Reports 9 (2023) 4111–4121. https://doi.org/10.1016/j.egyr.2023.03.042 41

    2023 doi

  23. [33]

    https://doi.org/10.3389/fenrg.2022.1113733

    Yu, K., Cao, J., Chen, X., Yang, Z., Gan, L., Resi- dential load forecasting based on electricity consumption pat- tern clustering, Frontiers in Energy Research 10 (2023) 1113733. https://doi.org/10.3389/fenrg.2022.1113733

    2023

  24. [34]

    https://doi.org/10.1016/j.apenergy.2018.12.063

    Motlagh, O., Berry, A., O’Neill, L., Clustering of residential electric- ity customers using load time series, Applied Energy 237 (2019) 11–24. https://doi.org/10.1016/j.apenergy.2018.12.063

    2019 doi

  25. [35]

    https://doi.org/10.1016/j.enbuild.2023.113550

    Dab, K., Henao, N., Nagarsheth, S., Dubé, Y., Sansregret, S., Agbossou, K., Consensus-based time-series clustering approach to short-term load forecasting for residential electricity demand, Energy and Buildings 299 (2023) 113550. https://doi.org/10.1016/j.enbuild.2023.113550

    2023

  26. [36]

    Jang, E., Gu, S., Poole, B., Categorical reparameterization with Gumbel-Softmax, In: International Conference on Learning Represen- tations (ICLR) (2017)

    2017

  27. [37]

    J., Mnih, A., Teh, Y

    Maddison, C. J., Mnih, A., Teh, Y. W., The concrete distribution: A continuous relaxation of discrete random variables, In: International Conference on Learning Representations (ICLR) (2017)

    2017

  28. [38]

    M., Pattern recognition and machine learning, Springer, New York, 2006

    Bishop, C. M., Pattern recognition and machine learning, Springer, New York, 2006

    2006

  29. [39]

    Auto-encoding variational Bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational Bayes,” in Proc. International Conference on Learning Representations (ICLR), 2014

    2014

  30. [40]

    P., Welling, M., Auto-encoding variational Bayes, In: In- ternational Conference on Learning Representations (ICLR) (2014)

    Kingma, D. P., Welling, M., Auto-encoding variational Bayes, In: In- ternational Conference on Learning Representations (ICLR) (2014). https://doi.org/10.48550/arXiv.1312.6114

    2014 doi arXiv

  31. [41]

    Garnelo, M., Rosenbaum, D., Maddison, C. J., Ramalho, T., Saxton, D., Shanahan, M., Teh, Y.W., Rezende, D.J., Eslami, S.M.A., Conditional neural processes, In: Proceedings of Machine Learning Research 80, International Conference on Machine Learning (ICML) (2018) 1704– 1713. 42

    2018

  32. [42]

    R., Smola, A

    Zaheer, M., Kottur, S., Ravanbakhsh, S., Póczos, B., Salakhutdinov, R. R., Smola, A. J., Deep sets, In: Advances in Neural Information Processing Systems (NeurIPS) (2017) 3391–3401

    2017

  33. [43]

    https://doi.org/10.1109/TPAMI.2018.2889774

    Zhang, C., Bütepage, J., Kjellström, H., Mandt, S., Ad- vances in variational inference, IEEE Transactions on Pattern Analysis and Machine Intelligence 41 (8) (2019) 2008–2026. https://doi.org/10.1109/TPAMI.2018.2889774

    2019

  34. [44]

    P., Rezende, D

    Kingma, D. P., Rezende, D. J., Mohamed, S., Welling, M., Semi- supervised learning with deep generative models, In: Advances in Neural Information Processing Systems (NeurIPS) 27 (2014) 3581–3589

    2014

  35. [45]

    https://doi.org/10.18653/v1/2021.emnlp-main.243

    Lester, B., Al-Rfou, R., Constant, N., The power of scale for parameter- efficient prompt tuning, In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP) (2021) 3045–3059. https://doi.org/10.18653/v1/2021.emnlp-main.243

    2021 doi

  36. [46]

    https://doi.org/10.3390/en15218079

    Ibrahim, B., Rabelo, L., Gutierrez-Franco, E., Clavijo-Buritica, N., Ma- chine learning for short-term load forecasting in smart grids, Energies 15 (21) (2022) 8079. https://doi.org/10.3390/en15218079

    2022 doi

  37. [47]

    Y., Hill, D

    Kong, W., Dong, Z. Y., Hill, D. J., Luo, F., Xu, Y., Short- term residential load forecasting based on LSTM recurrent neural network, IEEE Transactions on Smart Grid 10 (1) (2019) 841–851. https://doi.org/10.1109/TSG.2017.2753802

    2019

  38. [48]

    Williams, C. K. I., Seeger, M., Using the Nyström method to speed up kernel machines, In: Advances in Neural Information Processing Sys- tems (NeurIPS) (2001) 682–688

    2001

  39. [49]

    https://researchdata.edu.au/smart-grid-smart- trial-data/3796213

    [dataset] Department of Industry, Science, Energy and Resources (Australia), Smart Grid, Smart City: customer trial data, Research Data Australia (2014). https://researchdata.edu.au/smart-grid-smart- trial-data/3796213

    2014

  40. [50]

    Birnbaum, S., Kuleshov, V., Enam, Z., Koh, P. W., Ermon, S., Tempo- ral FiLM: capturing long-range sequence dependencies with feature-wise modulations, In: Advances in Neural Information Processing Systems (NeurIPS) (2019) 10287–10298. 43

    2019

  41. [51]

    J., Principles and algorithms for forecasting groups of time series: locality and globality, International Journal of Forecasting 37 (4) (2021) 1632–1653

    Montero-Manso, P., Hyndman, R. J., Principles and algorithms for forecasting groups of time series: locality and globality, International Journal of Forecasting 37 (4) (2021) 1632–1653. https://doi.org/10.1016/j.ijforecast.2021.03.004

    2021 doi

  42. [52]

    https://doi.org/10.1145/3637528.3671961

    Dong, Z., Jiang, R., Gao, H., Liu, H., Deng, J., Wen, Q., Song, X., Heterogeneity-informed meta-parameter learning for spatiotempo- ral time series forecasting, In: Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (2024). https://doi.org/10...

    2024

  43. [53]

    https://doi.org/10.1016/j.ijforecast.2015.11.011

    Hong, T., Fan, S., Probabilistic electric load forecasting: a tutorial review, International Journal of Forecasting 32 (3) (2016) 914–938. https://doi.org/10.1016/j.ijforecast.2015.11.011

    2016 doi

  44. [54]

    M., Conejo, A

    Morales, J. M., Conejo, A. J., Madsen, H., Pinson, P., Zugno, M., Integrating renewables in electricity markets: operational problems, Springer, New York, 2014. https://doi.org/10.1007/978-1-4614-9411-9

    2014 doi

  45. [55]

    J., Miranda, V., Botterud, A., Wang, J., Con- stantinescu, E., Time-adaptive quantile-copula for wind power probabilistic forecasting, Renewable Energy 40 (1) (2012) 29–39

    Bessa, R. J., Miranda, V., Botterud, A., Wang, J., Con- stantinescu, E., Time-adaptive quantile-copula for wind power probabilistic forecasting, Renewable Energy 40 (1) (2012) 29–39. https://doi.org/10.1016/j.renene.2011.08.015 44

    2012 doi

  46. [80]

    https://doi.org/10.1016/j.energy.2011.12.031

    2011 doi

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.