pith. sign in

arxiv: 2602.18949 · v3 · submitted 2026-02-21 · ⚛️ physics.data-an

Symmetry-Constrained Forecasting of Periodically Correlated Energy Processes

Cyril Voyant , Candice Banes , Luis Garcia-Gutierrez , Gilles Notton , Milan Despotovic , Zaher Mundher Yaseen This is my paper

Pith reviewed 2026-05-15 21:00 UTC · model grok-4.3

classification ⚛️ physics.data-an
keywords cyclostationary processespersistence forecastingrenewable energytime seriesperiodic correlationsymmetry constraintsenergy forecasting
0
0 comments X

The pith

A closed-form operator forecasts cyclostationary energy series by embedding phase symmetry into persistence.

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

The paper introduces analytical forecasting operators for time series with strong diurnal and seasonal cycles, such as solar irradiance or wind speed. These operators extend classical persistence using a simple coefficient that incorporates the local correlation between an observation and its phase-aligned lag. The construction preserves periodic variance and covariance through symmetry, which reduces the effective degrees of freedom without any model training. A reader would care because the result supplies a reproducible, physically interpretable baseline that improves accuracy at multi-hour horizons while remaining computationally minimal. The approach is validated on both synthetic cyclostationary signals and real renewable-energy datasets.

Core claim

The paper claims that the forecasting operator defined by the closed-form coefficient λ̃(t,τ) = ½(1 + ρ(t,τ)), where ρ(t,τ) is the local correlation between the current value and its phase-aligned lag τ, constitutes a training-free analytical limit of persistence for cyclostationary processes; this operator preserves periodic variance and covariance by enforcing symmetry-induced reduction of effective degrees of freedom.

What carries the argument

The central mechanism is the analytical operator λ̃(t,τ) = ½(1 + ρ(t,τ)), where ρ(t,τ) denotes local correlation at time t with phase-aligned lag τ; it extends persistence while enforcing temporal symmetry to keep periodic statistics intact.

If this is right

  • Forecasts preserve periodic variance and covariance exactly, eliminating the need for separate variance modeling.
  • The operator delivers measurable accuracy gains over classical persistence on both synthetic and real renewable-energy data, especially beyond one hour.
  • No training step is required, yielding a reproducible, computationally minimal baseline for any cyclostationary process.
  • The symmetry constraint reduces effective degrees of freedom while remaining physically interpretable.

Where Pith is reading between the lines

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

  • The same symmetry construction could be tested on non-energy cyclostationary signals such as traffic flow or physiological rhythms.
  • Because the operator is closed-form, it could serve as an analytical prior inside larger hybrid forecasting ensembles without adding parameters.
  • If phase alignment is stable, pre-computed correlation tables might allow real-time deployment on embedded hardware with negligible latency.

Load-bearing premise

The local correlation between each observation and its phase-aligned lag can be obtained directly from the data without training or post-hoc fitting and still produces the claimed symmetry-induced reduction in degrees of freedom.

What would settle it

Apply the operator to a cyclostationary series whose measured local correlations violate the assumed phase symmetry; if forecast skill falls below that of classical persistence at multi-hour horizons, the central claim is falsified.

Figures

Figures reproduced from arXiv: 2602.18949 by Candice Banes, Cyril Voyant, Gilles Notton, Luis Garcia-Gutierrez, Milan Despotovic, Zaher Mundher Yaseen.

Figure 1
Figure 1. Figure 1: Flowcharts illustrating the forecasting process for the [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Geographical distribution of the 68 SIAR weather stations and kriging surface of daytime averaged I (2017-2020) gob.es/fr/outils/siar/). Spatial consistency is ensured by ordinary kriging, follow￾ing the empirical variogram of each subregion. Forecast horizons span 30 min to 6 h in half hour increments. An ad hoc quality control following [37] guarantees data integrity. Cyclo￾stationary statistics and lear… view at source ↗
Figure 3
Figure 3. Figure 3: Cyclic parameters (mean µ(t), standard deviation σ(t), and correlation coefficient ρ(t)) shown on the left and prediction model comparisons (right) are applied to a noisy periodic signal (T = 24, n = 1 and ∆t = 1). The left side illustrates the smoothed and original cyclic statistics, while the right side displays the performance of models. 3.2. Benchmark of Reference Models with Synthetic Data Among the f… view at source ↗
Figure 4
Figure 4. Figure 4: Normalized Root Mean Square Error (nRMSE) plotted against variability metrics for synthetic time series across multiple prediction horizons. Each row represents a specific horizon (1, 3, 6, and 9 hours), and each column corresponds to a variability metric: Coefficient of Variation (CV), Mean Absolute Return (MAR), Root Mean Square Error (RMSE), and Autocorrelation at Lag 1. Trend lines using third order po… view at source ↗
Figure 5
Figure 5. Figure 5: Boxplots of the Normalized Root Mean Square Error ( [PITH_FULL_IMAGE:figures/full_fig_p016_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: indicates that the machine learning model (EL) delivers the lowest average nRMSE and nMAE across all forecast horizons, outperforming all other deterministic models despite not using any clear sky input [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Distribution of normalized forecasting errors across stations for four selected horizons (30, 120, 240, [PITH_FULL_IMAGE:figures/full_fig_p019_7.png] view at source ↗
read the original abstract

Time series in energy systems, such as solar irradiance, wind speed, or electrical load, are characterized by strong diurnal and seasonal periodicities. Accurate forecasting requires accounting for time varying statistical properties that stationary or classical persistence models cannot capture. A family of analytical forecasting operators for cyclostationary processes is introduced, extending persistence through a closed form coefficient $\tilde{\lambda}(t,\tau)=\tfrac{1}{2}\bigl(1+\rho(t,\tau)\bigr)$, where $\rho(t,\tau)$ denotes the local correlation between the current observation and its phase aligned time lag ($\tau$). This formulation preserves periodic variance and covariance, achieving a symmetry induced reduction of effective degrees of freedom. The resulting operator defines a training free analytical limit of persistence under periodic non stationarity. Validation on synthetic cyclostationary signals and empirical renewable energy datasets demonstrates consistent accuracy gains over classical persistence, particularly at multi hour horizons. By embedding temporal symmetry into the prediction process, the framework provides a physically interpretable, reproducible, and computationally minimal baseline for forecasting periodic processes across energy and complex systems.

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 a family of analytical forecasting operators for cyclostationary processes in energy systems (e.g., solar irradiance, wind speed, electrical load) characterized by diurnal and seasonal periodicities. The central operator is given by the closed-form coefficient tilde lambda(t, tau) = 1/2 (1 + rho(t, tau)), where rho(t, tau) is the local correlation between the current observation and its phase-aligned lag tau. The paper claims this preserves periodic variance and covariance, achieves a symmetry-induced reduction of effective degrees of freedom, defines a training-free analytical limit of persistence under periodic non-stationarity, and yields consistent accuracy gains over classical persistence on synthetic cyclostationary signals and empirical renewable energy datasets.

Significance. If the training-free property and symmetry-based reduction in degrees of freedom can be rigorously established without implicit data estimation, the approach would supply a simple, reproducible, and physically interpretable baseline for forecasting periodically correlated processes, offering low computational overhead and clear interpretability advantages for energy and complex systems applications.

major comments (2)
  1. [Abstract] Abstract, central operator definition: The claim that the operator is a 'training free analytical limit' with 'zero-parameter' characterization is undermined by the dependence on rho(t, tau), the local phase-aligned correlation. In cyclostationary processes, such correlations are obtained via ensemble averaging over observed cycles, which is a data-dependent estimation step whose degrees of freedom are not shown to be eliminated by symmetry constraints alone. No explicit construction of rho(t, tau) from symmetry without averaging or fitting is provided.
  2. [Abstract] Validation description: The abstract states that validation 'demonstrates consistent accuracy gains over classical persistence, particularly at multi-hour horizons,' yet provides no quantitative error metrics, baseline comparisons, or details on the practical computation of rho(t, tau) from the datasets. This leaves the empirical support for the symmetry-induced reduction and training-free properties unverifiable.
minor comments (1)
  1. [Abstract] Abstract: Minor LaTeX rendering inconsistencies in the displayed equation (e.g., use of tfrac and bigl) that may affect readability in some formats.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript introducing symmetry-constrained analytical operators for cyclostationary energy processes. We address each major comment point by point below, clarifying the role of the symmetry constraint and proposing targeted revisions to the abstract for improved precision.

read point-by-point responses

  1. Referee: [Abstract] Abstract, central operator definition: The claim that the operator is a 'training free analytical limit' with 'zero-parameter' characterization is undermined by the dependence on rho(t, tau), the local phase-aligned correlation. In cyclostationary processes, such correlations are obtained via ensemble averaging over observed cycles, which is a data-dependent estimation step whose degrees of freedom are not shown to be eliminated by symmetry constraints alone. No explicit construction of rho(t, tau) from symmetry without averaging or fitting is provided.

    Authors: The symmetry constraint is realized precisely through the closed-form operator definition itself, which is analytically derived to enforce preservation of the periodic variance and covariance structure of the cyclostationary process; this yields the specific relation tilde lambda(t, tau) = 1/2 (1 + rho(t, tau)) without additional free parameters to optimize. While rho(t, tau) is obtained via standard ensemble averaging over phase-aligned cycles (the conventional way to characterize any cyclostationary process), this step is a one-time statistical pre-computation of the intrinsic periodic correlation function rather than iterative training or fitting of the forecasting operator. The 'training-free' and 'zero-parameter' descriptors therefore refer to the absence of any model optimization beyond this fixed periodic characterization. We agree the abstract would benefit from greater precision on this distinction and will revise it to explicitly note that rho(t, tau) is the precomputed periodic autocorrelation function under the cyclostationary assumption, thereby highlighting the symmetry-induced reduction in effective degrees of freedom. revision: yes

  2. Referee: [Abstract] Validation description: The abstract states that validation 'demonstrates consistent accuracy gains over classical persistence, particularly at multi-hour horizons,' yet provides no quantitative error metrics, baseline comparisons, or details on the practical computation of rho(t, tau) from the datasets. This leaves the empirical support for the symmetry-induced reduction and training-free properties unverifiable.

    Authors: The abstract is intentionally concise, but the full manuscript supplies the requested quantitative support: explicit RMSE and MAE comparisons against classical persistence on both synthetic cyclostationary signals and empirical renewable-energy datasets, together with the precise procedure for computing rho(t, tau) via phase-aligned ensemble averaging over multiple periods. We will revise the abstract to include a brief reference to these results (e.g., 'as quantified in the main text with consistent RMSE reductions at multi-hour horizons') so that the empirical backing for the symmetry-induced and training-free claims is apparent even from the abstract alone. revision: yes

Circularity Check

1 steps flagged

Training-free operator defined via data-estimated local correlation ρ(t,τ)

specific steps
  1. fitted input called prediction [Abstract (equation for λ̃(t,τ))]
    "extending persistence through a closed form coefficient λ̃(t,τ)=½(1+ρ(t,τ)), where ρ(t,τ) denotes the local correlation between the current observation and its phase aligned time lag (τ). This formulation preserves periodic variance and covariance, achieving a symmetry induced reduction of effective degrees of freedom. The resulting operator defines a training free analytical limit of persistence under periodic non stationarity."

    The operator is labeled closed-form and training-free, yet its definition incorporates ρ(t,τ) as an input. In the context of periodically correlated processes, ρ(t,τ) is not supplied by symmetry alone but requires explicit estimation from historical data (ensemble averaging over cycles). The symmetry reduction and training-free properties therefore hold only after this data-dependent step, making the forecast a direct function of a fitted quantity rather than an independent analytical limit.

full rationale

The central forecasting operator is introduced as a closed-form, training-free analytical limit whose coefficient is defined directly as λ̃(t,τ) = ½(1 + ρ(t,τ)). By the paper's own description, ρ(t,τ) is the local phase-aligned correlation, which in cyclostationary processes is obtained by ensemble averaging over observed cycles. No derivation is supplied showing that symmetry constraints alone determine ρ without any data estimation step; the symmetry-induced reduction in degrees of freedom therefore presupposes the very fitted quantity whose estimation the 'training-free' claim is asserted to avoid. This produces a moderate circularity in which the claimed analytical property reduces to a reparameterized persistence model whose inputs are data-dependent by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the target processes are cyclostationary and that local phase-aligned correlation can be treated as an observable that does not itself require fitting parameters.

axioms (1)
  • domain assumption Energy time series exhibit cyclostationarity with periodic variance and covariance structures.
    Invoked to justify preservation of periodic statistics under the proposed operator.

pith-pipeline@v0.9.0 · 5507 in / 1229 out tokens · 33119 ms · 2026-05-15T21:00:56.947565+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.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Complexity and persistence of price time series of the European electricity spot market,

    C. Han, H. Hilger, E. Mix,et al., “Complexity and persistence of price time series of the European electricity spot market,”PRX Energy, vol. 1, p. 013002, Apr 2022

    work page 2022

  2. [2]

    Solar forecasting methods for renewable energy integration,

    R. Inman, H. Pedro, and C. Coimbra, “Solar forecasting methods for renewable energy integration,”Prog. Energy Combust. Sci., vol. 39, no. 6, pp. 535–576, 2013

    work page 2013

  3. [3]

    Review of solar irradiance forecasting methods and a proposition for small-scale insular grids,

    M. Diagne, M. David, P. Lauret,et al., “Review of solar irradiance forecasting methods and a proposition for small-scale insular grids,”Renew. Sustain. Energy Rev., vol. 27, pp. 65–76, 2013

    work page 2013

  4. [4]

    Machine learning methods for solar radiation forecasting: A review,

    C. Voyant, G. Notton, S. Kalogirou,et al., “Machine learning methods for solar radiation forecasting: A review,”Renew. Energy, vol. 105, pp. 569–582, 2017

    work page 2017

  5. [5]

    Physics-informed machine learning for power grid frequency modeling,

    J. Kruse, E. Cramer, B. Schäfer,et al., “Physics-informed machine learning for power grid frequency modeling,”PRX Energy, vol. 2, p. 043003, Oct 2023

    work page 2023

  6. [6]

    Machine learning and deep learning in energy systems: A review,

    M. Forootan, I. Larki, R. Zahedi,et al., “Machine learning and deep learning in energy systems: A review,”Sustainability, vol. 14, no. 8, p. 4832, 2022

    work page 2022

  7. [7]

    A comparative study of machine learning-based methods for global horizontal irradiance forecasting,

    S. Gbemou, J. Eynard, S. Thil,et al., “A comparative study of machine learning-based methods for global horizontal irradiance forecasting,”Energies, vol. 14, no. 11, p. 3192, 2021

    work page 2021

  8. [8]

    A benchmarking of machine learning techniques for solar radiation forecasting in an insular context,

    P. Lauret, C. Voyant, T. Soubdhan,et al., “A benchmarking of machine learning techniques for solar radiation forecasting in an insular context,”Sol. Energy, vol. 112, pp. 446–457, 2015. 24

    work page 2015

  9. [9]

    Solar radiation forecasting using artificial neural networks and random forest methods: application to normal beam, horizontal diffuse and global components,

    L. Benali, G. Notton, A. Fouilloy,et al., “Solar radiation forecasting using artificial neural networks and random forest methods: application to normal beam, horizontal diffuse and global components,”Renew. Energy, vol. 132, pp. 871–884, 2019

    work page 2019

  10. [10]

    Probabilistic analysis of electricity load forecasting errors,

    C. Eduardo and J. Jesús, “Probabilistic analysis of electricity load forecasting errors,” Electr. Power Syst. Res., vol. 255, p. 112807, 2026

    work page 2026

  11. [11]

    Napolitano,Cyclostationary Processes and Time Series: Theory, Applications, and Generalizations

    A. Napolitano,Cyclostationary Processes and Time Series: Theory, Applications, and Generalizations. Academic Press, 2019

    work page 2019

  12. [12]

    The comparison of Holt-Winters method and multiple regres- sion method: A case study,

    L. Tratar and E. Strmčnik, “The comparison of Holt-Winters method and multiple regres- sion method: A case study,”Energy, vol. 109, pp. 266–276, 2016

    work page 2016

  13. [13]

    The theta model: a decomposition approach to forecasting,

    V. Assimakopoulos and K. Nikolopoulos, “The theta model: a decomposition approach to forecasting,”International Journal of Forecasting, vol. 16, no. 4, pp. 521–530, 2000

    work page 2000

  14. [14]

    Forecasting of Electrical Generation Us- ing Prophet and Multiple Seasonality of Holt–Winters Models: A Case Study of Kuwait,

    A. Almazrouee, A. Almeshal, A. Almutairi,et al., “Forecasting of Electrical Generation Us- ing Prophet and Multiple Seasonality of Holt–Winters Models: A Case Study of Kuwait,” Appl. Sci., vol. 10, no. 23, 2020

    work page 2020

  15. [15]

    Forecasting time series with complex seasonal patterns using exponential smoothing,

    A. De Livera, R. Hyndman, and R. Snyder, “Forecasting time series with complex seasonal patterns using exponential smoothing,”Journal of the American Statistical Association, vol. 106, no. 496, pp. 1513–1527, 2011

    work page 2011

  16. [16]

    The Holt-Winters forecasting procedure,

    C. Chatfield, “The Holt-Winters forecasting procedure,”Journal of the Royal Statistical Society: Series C (Applied Statistics), vol. 27, no. 3, pp. 264–279, 1978

    work page 1978

  17. [17]

    Hyndman, A

    R. Hyndman, A. Koehler, J. Ord,et al.,Forecasting With Exponential Smoothing: the State Space Approach. Berlin, Heidelberg: Springer, 2008

    work page 2008

  18. [18]

    Periodic autoregressive forecasting of global solar irradiation without knowledge-based model implementation,

    C. Voyant, J. De Gooijer, and G. Notton, “Periodic autoregressive forecasting of global solar irradiation without knowledge-based model implementation,”Solar Energy, vol. 174, pp. 121–129, 2018

    work page 2018

  19. [19]

    G. Box, G. Jenkins, G. Reinsel,et al.,Time series analysis: forecasting and control. Wiley, 5 ed., 2015

    work page 2015

  20. [20]

    STL: a seasonal-trend decomposition pro- cedure based on loess,

    R. Cleveland, W. Cleveland, J. McRae,et al., “STL: a seasonal-trend decomposition pro- cedure based on loess,”Journal of Official Statistics, vol. 6, no. 1, pp. 3–73, 1990

    work page 1990

  21. [21]

    N-BEATS: Neural basis expansion analysis for interpretable time series forecasting,

    B. Oreshkin, D. Carpov, N. Chapados,et al., “N-BEATS: Neural basis expansion analysis for interpretable time series forecasting,”International Conference on Learning Represen- tations (ICLR), 2020

    work page 2020

  22. [22]

    Load and renewable energy forecasting for a microgrid using persistence technique,

    S. Dutta, Y. Li, A. Venkataraman,et al., “Load and renewable energy forecasting for a microgrid using persistence technique,”Energy Procedia, vol. 143, pp. 617–622, 2017. Leveraging Energy Technologies and Policy Options for Low Carbon Cities

    work page 2017

  23. [23]

    Performance persistence,

    S. Brown and W. Goetzmann, “Performance persistence,”J. Finance, vol. 50, no. 2, pp. 679–698, 1995

    work page 1995

  24. [24]

    A critical review of wind power forecasting methods—past, present and future,

    S. Hanifi, X. Liu, Z. Lin,et al., “A critical review of wind power forecasting methods—past, present and future,”Energies, vol. 13, no. 15, p. 3764, 2020. 25

    work page 2020

  25. [25]

    Time series and renewable energy forecasting,

    M. Ghofrani and M. Alolayan, “Time series and renewable energy forecasting,” inTime Series Analysis and Applications, pp. 333–344, IntechOpen, 2018

    work page 2018

  26. [26]

    Making reference solar forecasts with climatology, persistence, and their optimal convex combination,

    D. Yang, “Making reference solar forecasts with climatology, persistence, and their optimal convex combination,”Sol. Energy, vol. 193, pp. 981–985, 2019

    work page 2019

  27. [27]

    Using smart persistence and random forests to predict photovoltaic energy production,

    J. Tato and M. Brito, “Using smart persistence and random forests to predict photovoltaic energy production,”Energies, vol. 12, no. 1, p. 100, 2019

    work page 2019

  28. [28]

    Use of physics to improve solar forecast: Physics- informed persistence models for simultaneously forecasting GHI, DNI, and DHI,

    W. Liu, Y. Liu, X. Zhou,et al., “Use of physics to improve solar forecast: Physics- informed persistence models for simultaneously forecasting GHI, DNI, and DHI,”Sol. Energy, vol. 215, pp. 252–265, 2021

    work page 2021

  29. [29]

    Solar forecasts based on the clear sky index or the clearness index: Which is better?,

    P. Lauret, R. Alonso-Suárez, J. Le Gal La Salle,et al., “Solar forecasts based on the clear sky index or the clearness index: Which is better?,”Solar, vol. 2, no. 4, pp. 432–444, 2022

    work page 2022

  30. [30]

    Robust day-ahead solar forecasting with endogenous data and sliding windows,

    Y. Kamarianakis, Y. Pantazis, E. Kalligiannaki,et al., “Robust day-ahead solar forecasting with endogenous data and sliding windows,”J. Renew. Sustain. Energy, vol. 16, no. 2, p. 026103, 2024

    work page 2024

  31. [31]

    Characterization of cyclostationary random signal processes,

    W. Gardner and F. Lewis, “Characterization of cyclostationary random signal processes,” IEEE Trans. Inf. Theory, vol. 21, pp. 4–14, 1975

    work page 1975

  32. [32]

    Cyclostationary analysis for fault detection in PV inverters,

    M. Telidjane and B. Bakhti, “Cyclostationary analysis for fault detection in PV inverters,” Sol. Energy, vol. 291, p. 113381, 2025

    work page 2025

  33. [33]

    A cyclostationary model for temporal forecast- ing and simulation of solar global horizontal irradiance,

    S. Das, M. Genton, Y. Alshehri,et al., “A cyclostationary model for temporal forecast- ing and simulation of solar global horizontal irradiance,”Environmetrics, vol. 32, no. 8, p. e2700, 2021

    work page 2021

  34. [34]

    Validation of wind and PV power generation using historical and forecast weather data,

    S. Lee, F. Safdarian, and T. Overbye, “Validation of wind and PV power generation using historical and forecast weather data,” in2025 IEEE Texas Power and Energy Conference (TPEC), pp. 1–6, 2025

    work page 2025

  35. [35]

    Benchmarks for solar radiation time series forecasting,

    C. Voyant, G. Notton, J. Duchaud,et al., “Benchmarks for solar radiation time series forecasting,”Renew. Energy, vol. 191, pp. 747–762, 2022

    work page 2022

  36. [36]

    Solar irradiance time series fore- casting using auto-regressive and extreme learning methods: Influence of transfer learning and clustering,

    M. Despotovic, C. Voyant, L. Garcia-Gutierrez,et al., “Solar irradiance time series fore- casting using auto-regressive and extreme learning methods: Influence of transfer learning and clustering,”Appl. Energy, vol. 365, p. 123215, 2024

    work page 2024

  37. [37]

    Evaluation and comparison of spatial clustering for solar irradiance time series,

    L. Garcia-Gutierrez, C. Voyant, G. Notton,et al., “Evaluation and comparison of spatial clustering for solar irradiance time series,”Appl. Sci., vol. 12, no. 17, 2022

    work page 2022

  38. [38]

    Forecasting renewable energy consumption with hydrogen integration: A comprehensive regression approach,

    M. Güçyetmez, S. Akkaya, M. Uyar,et al., “Forecasting renewable energy consumption with hydrogen integration: A comprehensive regression approach,”International Journal of Hydrogen Energy, vol. 142, pp. 981–993, 2025

    work page 2025

  39. [39]

    On the importance of Clearsky model in short-term solar radiation forecasting,

    C. Voyant, M. Despotovic, G. Notton,et al., “On the importance of Clearsky model in short-term solar radiation forecasting,”Sol. Energy, vol. 294, p. 113490, 2025

    work page 2025

  40. [40]

    Verification of deterministic solar fore- casts,

    D. Yang, S. Alessandrini, J. Antonanzas,et al., “Verification of deterministic solar fore- casts,”Sol. Energy, vol. 210, pp. 20–37, 2020. Special Issue on Grid Integration. 26

    work page 2020

  41. [41]

    Validating surface downwelling solar irradiances estimated by the McClear model under cloud-free skies in the United Arab Emirates,

    Y. Eissa, S. Munawwar, A. Oumbe,et al., “Validating surface downwelling solar irradiances estimated by the McClear model under cloud-free skies in the United Arab Emirates,”Sol. Energy, vol. 114, pp. 17–31, 2015. 27

    work page 2015